Methods, equipment, and storage media for predicting the service life of nylon materials
By constructing an empirical statistical lifetime model and a mass transfer mechanism lifetime model, and combining them with an order-preserving denoising algorithm to process the degradation data of nylon materials, accurate prediction of the lifetime of nylon materials was achieved. This solved the problem of unpredictable lifetime of materials in marine engineering service scenarios and improved the reliability and applicability of the prediction.
Patent Information
- Application Number
- CN202611351088.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-09-02
- Publication Date
- 2026-11-17
AI Technical Summary
In marine engineering applications, nylon materials suffer from dimensional expansion and mechanical property degradation due to the combined effects of high humidity, wide temperature range, and salt spray. Consequently, their service life is difficult to predict accurately.
By acquiring the original degradation time series data of nylon samples, the data is processed using an order-preserving denoising algorithm. Combined with linear interpolation based on failure criteria, an empirical statistical lifetime model and a mass transfer mechanism lifetime model are constructed. Based on these models, dual-path collaborative calculations are performed to predict the target lifetime of the material.
It improves the accuracy and physical interpretability of nylon material life prediction, solves the problem of difficult prediction of material replacement cycle, and enhances the reliability and engineering adaptability of prediction.
Smart Images

Figure CN123116602A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of lifespan prediction technology, and in particular to a method, apparatus and storage medium for predicting the lifespan of nylon materials. Background Technology
[0002] In marine engineering applications such as offshore wind power and marine oil and gas equipment, nylon materials are widely used in precision components such as insulating connectors, sealing gaskets, and cable clamps due to their excellent electrical insulation properties, chemical corrosion resistance, and lightweight advantages.
[0003] In related technologies, due to the long-term high humidity, wide temperature range and salt spray synergistic hot and humid conditions of the marine atmospheric environment, water molecules will continuously diffuse into the interior of nylon materials, causing the material size to expand and the mechanical properties to gradually decline, resulting in a short service life of nylon materials, and thus making it difficult to predict the replacement cycle of nylon materials.
[0004] The above content is only used to help understand the technical solution of this application and does not represent an admission that the above content is prior art. Summary of the Invention
[0005] The main objective of this application is to provide a method, device, and storage medium for predicting the service life of nylon materials, aiming to solve the technical problem of the difficulty in predicting the replacement cycle of nylon materials.
[0006] To achieve the above objectives, this application proposes a method for predicting the service life of nylon materials, the method comprising: Obtain the original degradation time series data of nylon samples, wherein the original degradation time series data includes the outer diameter, mass and tensile strength of each group of nylon samples arranged according to the time period under each group of temperature and humidity conditions. The original degraded time series data is processed by a sequence-preserving denoising algorithm to obtain a denoised degraded sequence. Combined with linear interpolation based on the failure criterion, the median characteristic lifetime corresponding to the working condition is calculated. Using the median characteristic lifetime and the denoised degradation sequence as the fitting basis, an empirical statistical lifetime model and a mass transfer mechanism lifetime model are constructed through a staged fitting strategy. Based on the empirical statistical lifetime model and the mass transfer mechanism lifetime model, the target predicted lifetime of the nylon sample is obtained by substituting the temperature and humidity parameters of the target service environment through dual-path collaborative calculation.
[0007] In one embodiment, the nylon sample and the standard tensile test strip are placed in a vacuum oven for constant temperature drying. When the mass change rate of two consecutive weighings is lower than the preset mass change rate, the dried sample is obtained. The dried samples were placed in batches in a constant temperature and humidity test chamber with multiple temperature and humidity gradients and a constant temperature water bath with multiple temperature gradients to obtain aged samples through an aging process. The aging samples were obtained in batches according to the time period of denser samples at the beginning and sparser samples at the end, and the samples were wiped and left to stand to obtain the test samples. The outer diameter of the test sample was measured at three points, the mass was weighed with high precision, and the tensile strength was tested. The parameter values at each time point were recorded to obtain the original degradation time series data.
[0008] In one embodiment, the original degradation time series data is split into groups according to the test conditions and sample numbers to generate multiple independent degradation subsequences; The degraded subsequence is processed according to the order-preserving denoising algorithm, and adjacent data blocks that do not satisfy monotonicity are merged by the adjacent violation point pooling algorithm to obtain the denoised degraded sequence. Based on the relative expansion failure coefficient and the strength retention rate failure coefficient, the data points before and after the failure threshold are located in the denoising degradation sequence, and the failure time is calculated by linear interpolation. Discrete outliers are removed from the failure times of all samples under the same operating condition. The median of the target failure times is taken as the characteristic lifetime of the operating condition, and the median characteristic lifetime is obtained by summing them up.
[0009] In one embodiment, the time-series data points in the degenerate subsequence are initialized as independent data blocks, and the sample size and average value of the independent data blocks are recorded to construct an initial ordered block set arranged in chronological order. The initial ordered block set is traversed sequentially, and the mean values of adjacent independent data blocks are compared to locate the first monotonic violation adjacent block pair where the mean of the later block is less than the mean of the earlier block. The monotonicity violation adjacent block pairs are merged into a single data block to obtain the merged data block; After replacing the monotonicity violation adjacent block pair based on the merged data block, backtrack one bit and re-verify the monotonicity of the preceding block until the entire sequence block set satisfies the non-decreasing constraint, thus obtaining the global order-preserving block set; The mean of each data block in the global order-preserving block set is mapped one by one back to the data points in the corresponding time interval of the degenerate subsequence to obtain the denoised degenerate sequence.
[0010] In one embodiment, based on the initial outer diameter and initial tensile strength reference value of the nylon sample, and combined with the relative expansion failure coefficient and the strength retention failure coefficient, the absolute threshold for dimensional failure and the absolute threshold for strength failure corresponding to the initial state of the nylon sample are calculated and generated. Based on the outer diameter data subsequence and intensity data subsequence in the denoised degradation sequence, the first failure sampling node whose value is greater than or equal to the corresponding absolute threshold for size failure or the absolute threshold for intensity failure is retrieved and determined. Based on the time and performance parameters of the failed sampling node and its preceding adjacent node, a failure crossing node pair is constructed. Based on the assumption that the performance of the nodes between the failure cross-node pairs changes linearly with time, a proportional conversion relationship is constructed. The time coordinates corresponding to when the performance value is equal to the absolute threshold of the size failure or the absolute threshold of the strength failure are solved in reverse. The failure time is obtained by linear interpolation.
[0011] In one embodiment, the median characteristic lifetime and the denoised degradation sequence are grouped and matched according to the temperature and humidity dimension to determine the lifetime statistical subset with multiple humidity at the same temperature and the moisture absorption degradation sequence subset across the entire temperature range. The humidity acceleration index is calculated based on the aforementioned lifetime statistical subset, and the humidity coupling effect of the median characteristic lifetime is corrected by the humidity acceleration index to obtain the logarithmic lifetime. The activation energy and the precondition constant are obtained by performing least squares linear regression on the logarithmic lifetime and the reciprocal of the absolute temperature, so as to construct the empirical statistical lifetime model. Based on the aforementioned moisture absorption degradation sequence subset, the moisture diffusion coefficient and equilibrium moisture absorption under each temperature and humidity condition are obtained by fitting each working condition, and the physical parameters of the mass transfer process are determined. Based on the linear conversion relationship between size and strength and moisture absorption, and coupled with the physical parameters of the mass transfer process, a mass transfer mechanism lifetime model is constructed.
[0012] In one embodiment, the temperature and humidity parameters of the target service environment are normalized in dimensions and their value ranges are verified to generate an environmental parameter set that adapts to the empirical statistical lifetime model and the mass transfer mechanism lifetime model. The environmental parameter set is substituted into the empirical statistical lifetime model and the mass transfer mechanism lifetime model respectively for independent calculation to obtain the initial value of the mechanism path lifetime and the dynamic moisture absorption evolution parameters. Based on the degree of deviation between the target operating condition and the test operating condition, as well as the time span of the prediction period, the initial value of the mechanism path lifetime and the dynamic moisture absorption evolution parameter are assigned confidence weights, and the baseline predicted lifetime is obtained by weighted calculation. By applying the physical constraints and applicable boundaries of the empirical statistical lifetime model and the mass transfer mechanism lifetime model, the rationality verification and boundary correction of the baseline predicted lifetime are performed to obtain the target predicted lifetime of the nylon sample.
[0013] In one embodiment, the set of environmental parameters is substituted into the empirical statistical lifetime model, and the lifetime is solved by pre-identified activation energy and humidity acceleration index. The initial value of empirical path lifetime and the power function degradation rate coefficient under the corresponding working condition are output to form the empirical side calculation result. The temperature and humidity mapping parameters of the environmental parameter set are input into the isothermal adsorption calculation unit of the mass transfer mechanism lifetime model, and the equilibrium moisture absorption under the target working condition is obtained by combining the monolayer saturated moisture absorption parameters. The absolute temperature parameter in the environmental parameter set is input into the diffusion kinetics calculation unit of the mass transfer mechanism lifetime model. Based on the prefitted diffusion activation energy and the pre-exponential factor, an exponential operation is performed to obtain the moisture diffusion coefficient at the target temperature. Based on the equilibrium moisture absorption and the moisture diffusion coefficient as boundary conditions of the mass transfer process, the failure critical moisture absorption and the performance moisture absorption conversion relationship are substituted into the two-dimensional moisture diffusion equation for numerical root finding, and the initial value of the mechanism path lifetime is obtained. Based on the degradation rate coefficient of the function, combined with the degradation time exponent and the initial performance benchmark value, a full-cycle extrapolation is performed to generate the dynamic moisture absorption evolution parameters.
[0014] Furthermore, to achieve the above objectives, this application also proposes a device for predicting the service life of nylon materials. The device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. The computer program is configured to implement the steps of the method for predicting the service life of nylon materials as described above.
[0015] In addition, to achieve the above objectives, this application also proposes a storage medium, which is a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements the steps of the method for predicting the service life of nylon materials as described above.
[0016] This application provides a method for predicting the service life of nylon materials. The method involves subjecting nylon samples to accelerated aging treatment under multi-gradient temperature and humidity conditions. The method collects sample outer diameter, sample mass, and tensile strength data at preset time intervals to determine the original degradation time series data. An order-preserving denoising algorithm is used to process the original degradation time series data to obtain a denoised degradation sequence. Combined with failure criteria, the median characteristic life corresponding to each operating condition is calculated via linear interpolation. Then, using the median characteristic life and the denoised degradation sequence as the fitting basis, a staged fitting strategy is employed to decouple the temperature and humidity coupling effect. An empirical statistical life model and a mass transfer mechanism life model are simultaneously constructed. Finally, the temperature and humidity parameters of the target service environment are substituted into the model, and the target predicted life of the nylon sample is obtained through dual-path collaborative calculation. This method solves the technical problems in existing nylon material aging assessments where measurement noise causes degradation data to violate physical monotonicity, the difficulty in separating the temperature and humidity coupling effect leads to ambiguity in the physical meaning of model parameters, and the limited extrapolation accuracy and applicability of a single prediction model. This method improves the accuracy, physical interpretability, and engineering scenario adaptability of the nylon material size and mechanical property life prediction results.
[0017] In summary, this application uses multi-gradient temperature, humidity, and heat to accelerate aging and collect degradation data. After order-preserving denoising preprocessing and staged fitting, a dual-lifetime model is constructed. The results are output through dual-path collaborative calculation, which solves the technical problem of unpredictable replacement cycle of nylon materials and improves prediction reliability and physical interpretability. Attached Figure Description
[0018] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.
[0019] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0020] Figure 1 This is a flowchart illustrating the first embodiment of the method for predicting the service life of nylon materials according to this application. Figure 2 This is a flowchart illustrating the lifespan prediction process for the nylon material in this application. Figure 3 This is a flowchart illustrating the fourth embodiment of the method for predicting the service life of nylon materials in this application; Figure 4 This is a flowchart illustrating the fifth embodiment of the method for predicting the service life of nylon materials in this application; Figure 5 This is a flowchart illustrating the eighth embodiment of the method for predicting the service life of nylon materials in this application; Figure 6 The following is a fitting diagram of the four stages of the model in this application; Figure 7 This is a typical operating condition prediction diagram for this application; Figure 8 This is a schematic diagram of the electron microscopy results of this application; Figure 9 This is a schematic diagram of the device for predicting the service life of nylon materials in this application.
[0021] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation
[0022] It should be understood that the specific embodiments described herein are merely illustrative of the technical solutions of this application and are not intended to limit this application.
[0023] In related technologies, due to the long-term high humidity, wide temperature range and salt spray synergistic hot and humid conditions of the marine atmospheric environment, water molecules will continuously diffuse into the interior of nylon materials, causing the material size to expand and the mechanical properties to gradually decline, resulting in a short service life of nylon materials, and thus making it difficult to predict the replacement cycle of nylon materials.
[0024] This application provides a solution: First, the original degradation time series data of nylon samples is obtained, wherein the original degradation time series data includes the outer diameter, mass, and tensile strength of each group of nylon samples arranged according to the time period under each group of temperature and humidity conditions. Then, the original degradation time series data is processed by an order-preserving denoising algorithm to obtain a denoised degradation sequence. Combined with linear interpolation of the failure criterion, the median characteristic lifetime corresponding to the working condition is calculated. Then, the median characteristic lifetime and the denoised degradation sequence are used as the fitting basis, and an empirical statistical lifetime model and a mass transfer mechanism lifetime model are constructed through a staged fitting strategy. Finally, based on the empirical statistical lifetime model and the mass transfer mechanism lifetime model, the temperature and humidity parameters of the target service environment are substituted, and the target predicted lifetime of the nylon sample is obtained through dual-path collaborative calculation.
[0025] It should be noted that the executing entity in this embodiment can be a computing service device with data processing, network communication, and program execution functions, such as a tablet computer, personal computer, or mobile phone, or an electronic device capable of performing the above functions, such as a nylon material lifespan prediction device. The following description uses a nylon material lifespan prediction device as an example to illustrate this embodiment and the subsequent embodiments.
[0026] To better understand the technical solution of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific implementation methods.
[0027] This application provides a method for predicting the service life of nylon materials, referring to... Figure 1 , Figure 1 This is a flowchart illustrating the first embodiment of the method for predicting the service life of nylon materials according to this application.
[0028] In this embodiment, the method for predicting the service life of the nylon material includes steps S10 to S40: Step S10: Obtain the original degradation time series data of the nylon samples, wherein the original degradation time series data includes the outer diameter, mass and tensile strength of each group of nylon samples arranged according to the time period under each group of temperature and humidity conditions.
[0029] Accelerated aging treatment with humidity is a reliability testing method that accelerates the moisture absorption, expansion, and performance degradation process of nylon materials by precisely controlling ambient temperature and relative humidity. The preset time period is a sequence of measurement time nodes set according to the characteristic that the moisture absorption rate of nylon is fast at the beginning and slow at the end, with denser measurements at the beginning and sparser measurements at the end. The original degradation time series data is a multi-dimensional degradation data set of the sample's outer diameter, mass water absorption rate, and tensile strength, recorded synchronously in chronological order. For example, the outer diameter data corresponds to the dimensional expansion degradation, the mass data corresponds to the moisture absorption diffusion progress, and the tensile strength corresponds to the mechanical property degradation.
[0030] In this embodiment, the aforementioned accelerated aging test under damp heat can be constructed and initiated in four ways. First, a constant-gradient group initiation method: Following multiple preset fixed temperature and humidity conditions, dried nylon samples are grouped and placed into constant-temperature and humidity test chambers and constant-temperature water baths with corresponding parameters, simultaneously initiating the aging process for all conditions to obtain a complete degradation dataset of the full-condition matrix. Second, a gradient-progressive single-chamber initiation method: Using a single programmable temperature and humidity test chamber, conditions are switched sequentially according to temperature from low to high and humidity from low to high. Samples from the same batch undergo different gradient aging processes sequentially, reducing equipment usage and enabling progressive data acquisition across multiple conditions. Third, a service environment mapping initiation method: Based on the measured temperature and humidity spectrum of the target service scenario, a proportional accelerated conversion is performed, setting aging conditions that match the stress ratio of the actual service environment to obtain degradation data with higher accuracy to the actual conditions. Fourth, a failure prediction dynamic initiation method: A preliminary degradation rate is obtained through pre-testing, and the subsequent temperature and humidity gradient settings and measurement cycles are dynamically adjusted to shorten the total test cycle while ensuring data integrity.
[0031] Once the aging test is started and the preset measurement time point is reached, the sample parameter acquisition begins to obtain degradation data in various dimensions.
[0032] For example, there are two methods for collecting the outer diameter, mass, and tensile strength of the specimen. The first is non-destructive in-situ synchronous acquisition. At each measurement time point, the circular specimen corresponding to the working condition is quickly removed, the surface condensation is gently wiped off, and it is allowed to stand at room temperature. Then, the three-point average measurement of the outer diameter and high-precision mass weighing are completed sequentially. After the measurement is completed, it is immediately returned to the test environment for continued aging. At the same time point, a parallel strip is taken out to complete the tensile strength test. After all parameters are collected, they are entered into the data system to form complete degradation data for a single time point. This method has a clear acquisition logic, strong data correspondence, and allows for continuous tracking of the same sample for dimensional and mass data, providing continuous and reliable time-series data for subsequent degradation analysis.
[0033] The second method involves offline sampling in batches. All samples are pre-divided into multiple batches according to measurement time points, with each batch corresponding to a specific measurement time point. Once the corresponding time is reached, all samples in the batch are retrieved, and the outer diameter, mass, and tensile strength of all samples in that batch are tested in one go. After testing, the samples are not returned to the aging environment. This method avoids interference from repeated removal and return to the sample's moisture absorption state, resulting in higher data stability. It also allows for simultaneous completion of supporting tests such as cross-sectional morphology observation.
[0034] Once the measurement data for all time points is obtained, the raw degradation time series data can be generated by organizing the data in the following way.
[0035] In one alternative approach, data is archived at two levels: operating conditions and samples. First, all measured data are grouped according to temperature and humidity conditions. Then, under each operating condition, the measured values at each time point are organized by sample number, generating independent time-series data for outer diameter, mass, and strength for each sample. Finally, these are aggregated to form a complete raw degradation time-series dataset covering all operating conditions. This archiving method has an intuitive logic, clear data indexing, and facilitates subsequent retrieval and analysis according to operating conditions.
[0036] In another alternative approach, all raw measurements are first standardized in terms of dimensions and outliers are initially screened to remove outliers that clearly exceed the physically reasonable range. Then, the measurement nodes of all samples under the same operating condition are aligned along the time axis to construct a multi-sample degradation data matrix at the operating condition level, ultimately outputting a standardized raw degradation time-series dataset. This method completes data standardization and preliminary cleaning in advance, allowing direct integration with subsequent denoising processes and improving overall processing efficiency.
[0037] In another alternative approach, a nylon material lifespan prediction system is used. This system includes a sample information management module, a multi-station environmental condition control module, a sampling cycle intelligent scheduling module, a multi-parameter parallel measurement module, and a raw data archiving and verification module. Each module collaboratively executes the complete data acquisition process according to preset logic: The sample information management module pre-assigns, groups, and enters initial reference parameters for all nylon disc samples and standard tensile specimens, assigning each group of samples a unique identifier and corresponding test condition label. The multi-station environmental condition control module interfaces with multiple constant temperature and humidity test chambers and constant temperature water baths, independently controlling the temperature and humidity parameters of each station to maintain long-term stability of the temperature and humidity conditions for each group. The grouped, dried samples are then placed in their respective stations to initiate the damp heat aging process. The intelligent sampling cycle scheduling module automatically triggers sampling tasks at each station according to a preset time cycle with higher frequency sampling followed by lower frequency, outputting sampling prompts and corresponding sample lists. The multi-parameter parallel measurement module integrates three measurement units: a digital micrometer, a high-precision analytical balance, and an electronic universal testing machine. It sequentially performs three-point average measurement of the outer diameter, high-precision weighing of the mass, and tensile strength testing on the extracted test samples, automatically collecting the values of each measurement and associating them with the corresponding sample identifier and sampling timestamp. The raw data archiving and verification module categorizes, sorts, and verifies the integrity of all measurement data according to test conditions, sample number, and sampling time. Finally, it generates raw degradation time-series data containing the outer diameter, mass, and tensile strength of each group of nylon samples under each temperature and humidity condition, arranged in an orderly manner according to time periods.
[0038] Before acquiring the raw degradation time series data of nylon samples, the optional aging test data acquisition schemes in the nylon material service life prediction system are as follows: all multi-dimensional degradation data generated by aging tests are collected in real time by the multi-source degradation data acquisition unit and then fully collected and archived to the raw degradation data full-archiving module. This module will simultaneously associate the sample outer diameter, sample mass, and tensile strength measurements of all nylon samples throughout the entire aging cycle at each temperature and humidity station with sample identification, operating parameters, sampling timestamps, environmental operation records, and other metadata to complete structured storage and full-cycle traceability. When formally retrieving the raw degradation time series data used for service life calculation, the system supports multiple... The flexible data acquisition methods allow for targeted sampling based on user-selected target temperature and humidity conditions and specified sampling time points. This enables precise extraction of degradation data for samples in the corresponding group at specific times to adapt to single-condition rapid verification scenarios. Alternatively, full data acquisition can be performed, retrieving degradation data for all conditions and their entire lifecycle from the archive library to support the complete fitting of the global multi-parameter model. Furthermore, conditional retrieval can be performed based on preset temperature ranges, humidity ranges, degradation degree thresholds, sample production batches, and other custom filtering conditions. This automatically matches and filters out a subset of degradation time-series data that meets the constraints, thus balancing the data accuracy and computational efficiency requirements under different prediction scenarios.
[0039] In an exemplary scheme for determining the original degradation time series data, a pre-set test condition matrix and sample number list are first loaded. This list includes four temperature gradients, three humidity gradients, and a water bath control condition, as well as the number of parallel samples and numbering rules for each condition. Next, initial data verification is performed to check whether the number of measured samples at each time point matches the pre-set number, and to verify whether the numerical ranges of outer diameter, mass, and strength data are within the reasonable range for the material, eliminating abnormal data points caused by operational errors. Then, time series alignment is performed, arranging the measurement data of each sample in chronological order, filling in missing time point markers, and generating a single-sample degradation time series sequence. Next, condition-dimensional aggregation is performed, integrating the time series sequences of all samples under the same condition to form a condition-level multi-sample degradation data set. Finally, data format standardization is performed, unifying time units, size units, and strength units, adding condition parameter labels and sample attribute labels, generating the final original degradation time series dataset for subsequent order-preserving denoising and lifetime calculation.
[0040] It should be noted that in some special cases, if the number of effective samples under a single working condition is lower than the preset minimum sample size, the time series data of that working condition will be supplemented by interpolation of data from adjacent working conditions, or the working condition will be marked as a verification working condition and will not participate in subsequent model fitting.
[0041] Step S20: The original degraded time series data is processed by the order-preserving denoising algorithm to obtain the denoised degraded sequence. Combined with the failure criterion linear interpolation, the median characteristic lifetime corresponding to the working condition is calculated.
[0042] Order-preserving denoising algorithms are data processing algorithms that eliminate measurement random noise while strictly preserving the monotonic trend of the data. Denoising degradation sequences are degraded data sequences that conform to the monotonic law of hygroscopic expansion after eliminating non-physical fallback anomalies. Failure criteria are quantitative thresholds for determining whether a sample has reached a service failure state. Median characteristic lifetime is the median failure time of all samples under the same operating condition, used to characterize the overall lifespan level of the material under that condition. For example, the dimensional failure criterion is a relative expansion of 0.5%, and the strength failure criterion is a decrease in tensile strength retention to 80% of the initial value.
[0043] In this embodiment, the above-mentioned order-preserving denoising process can be performed in three ways. The first is global order-preserving processing using adjacent violation point pooling. Each data point in the original degraded time series data is treated as an independent block. The mean of adjacent blocks is compared sequentially from left to right. If the mean of a later block is less than that of a previous block, the two blocks are merged and the weighted mean is recalculated. This iterative backtracking verification continues until the entire sequence meets the non-decreasing constraint, ultimately mapping back to a denoised sequence of equal length. This method requires no preset hyperparameters, has a high degree of adaptability, can completely eliminate data fallback, and ensures the physical consistency of the degradation process. The second is sliding window order-preserving fitting processing. A fixed-width sliding window is set, and order-preserving regression fitting is performed on the data within each window. The fitted value replaces the original measurement value. The window gradually slides through the entire sequence, ultimately obtaining a smooth denoised degraded sequence. This method can simultaneously suppress small random fluctuations, resulting in a smoother output sequence, and is suitable for measurement data with high noise levels. Thirdly, a hierarchical order-preserving processing method is used. First, coarse-grained order-preserving denoising is performed on the working condition-level mean sequence to determine the baseline of the overall degradation trend. Then, fine-grained order-preserving correction is performed on the single-sample sequence using the baseline as a constraint, finally obtaining the single-sample denoised sequence. This method takes into account both the overall degradation law and the discrete characteristics of individuals, and the denoising result is closer to the actual degradation trend of the material.
[0044] After the denoising process is completed and the denoised degradation sequence is obtained, the failure time calculation and median characteristic lifetime statistics are performed.
[0045] For example, there are two methods for calculating the failure time and median characteristic lifetime. The first method uses single-sample interpolation and median statistics for each operating condition. First, for each sample's denoised degradation sequence, the two measurement nodes before and after the failure threshold are located. Based on these two points, a linear degradation relationship is constructed, and the precise time corresponding to reaching the failure threshold is solved to obtain the single-sample failure time. Then, the failure times of all valid samples under the same operating condition are sorted, and the median is taken as the median characteristic lifetime for that operating condition. This method has a clear calculation logic, high interpolation accuracy, and strong resistance to extreme values in median statistics, conforming to the statistical characteristics of lifetime data.
[0046] The second method involves Weibull distribution fitting and characteristic lifetime calculation. First, the failure times of all samples under the same operating condition are fitted with Weibull distribution parameters to obtain shape and scale parameters. Then, the 50th percentile lifetime of the distribution is used as the median characteristic lifetime for that operating condition. This method, based on a probabilistic statistical model, can fully utilize the distribution information of the sample data and still obtain robust lifetime estimates when the sample size is small.
[0047] Once the failure time of a single sample is obtained, the median characteristic lifetime under operating conditions can be generated based on the following method.
[0048] In one alternative approach, outlier sample removal is first performed. The Grubbs test is used to identify and remove outliers in failure time under the same operating condition. Then, the failure times of the remaining valid samples are sorted in ascending order, and the median value is taken as the median characteristic lifetime. If the sample size is even, the average of the two median values is used. This method is simple to operate, provides intuitive results, requires no additional distribution assumptions, and has strong engineering applicability.
[0049] In another alternative approach, a goodness-of-fit test is first performed on the failure times of all samples to confirm that they conform to a Weibull distribution. Then, the maximum likelihood estimation method is used to solve for the distribution parameters, directly calculating the median characteristic lifetime corresponding to a 50% failure probability, and simultaneously outputting the lifetime confidence interval. This method is supported by statistical theory and can also provide the range of lifetime fluctuations.
[0050] In another alternative scheme, the data preprocessing and characteristic life calculation unit of the nylon material life prediction system performs the calculation. The unit has four built-in functional modules: data splitting and sequence construction module, order-preserving denoising calculation module, failure threshold matching and interpolation calculation module, and working condition-level life statistics module. Each module works together to complete the initial life value calculation in the order of data flow: the data splitting and sequence construction module receives the raw degradation time series data from the upstream acquisition and outputs it. According to the three-level rules of test condition grouping, sample number indexing, and performance dimension splitting, the full amount of raw data is split into multiple independent degradation subsequences corresponding to a single sample and a single dimension. After completing the time series alignment and missing value marking, it is output to the next module. The order-preserving denoising module embeds the Pool Adjacent Violators Algorithm (PAVA). First, it initializes the temporal measurement points of each degraded subsequence as independent data blocks and records the sample size and mean. Then, it iterates from left to right to verify the monotonicity of adjacent blocks. It locates monotonicity violations where the mean of the next block is less than that of the previous block, performs weighted merging, and automatically backslides one position to re-examine the preceding sequence. After iterating until the entire sequence satisfies the non-decreasing constraint, it maps the block mean back to the original measurement point position, outputting a globally denoised and degraded sequence that conforms to physical laws. The failure threshold matching and interpolation calculation module pre-stores the relative expansion failure coefficient and the strength retention rate failure coefficient. Combined with the initial baseline value of the sample, it calculates the absolute failure threshold for the corresponding size and strength. Then, it retrieves the first sampling node in each denoised and degraded sequence whose value reaches or exceeds the failure threshold, locks this node, and constructs a failure-crossing node pair with its preceding adjacent node. Based on the assumption that performance changes linearly with time, it performs a proportional inverse solution and calculates the precise failure time of a single sample through linear interpolation. The operating condition-level life statistics module performs discrete outlier removal on the failure time of all samples under the same operating condition, and after filtering the valid failure times, takes the median as the characteristic life of the operating condition. Finally, it summarizes the calculation results of all temperature and humidity operating conditions and outputs the median characteristic life dataset of the whole operating condition for subsequent model fitting and life prediction.
[0051] In an exemplary scheme for determining the median characteristic lifetime, firstly, denoised degradation sequences corresponding to all operating conditions and preset failure criterion thresholds are loaded, including two types: a failure threshold of 0.5% relative size expansion and a failure threshold of 80% strength retention. Then, single-sample failure time calculation is performed. The denoised sequence of each sample is traversed one by one, locating the first data point greater than or equal to the failure threshold. The time and performance parameters of this point relative to the previous node are recorded, and substituted into a linear interpolation formula to calculate the accurate single-sample failure time. Next, outlier detection is performed. A Grubbs test is performed on all single-sample failure times under the same operating condition, and outlier samples with a significance level lower than a preset threshold are removed. Finally, median calculation is performed. The remaining valid failure times are sorted in ascending order to calculate the median characteristic lifetime value.
[0052] It should be noted that in some special cases, if all samples under a certain working condition do not reach the failure threshold within the test period, then that working condition will be treated as truncated data and used only for model validation, and will not be used for parameter fitting.
[0053] Step S30: Using the median characteristic lifetime and the denoised degradation sequence as the fitting basis, an empirical statistical lifetime model and a mass transfer mechanism lifetime model are constructed through a staged fitting strategy.
[0054] The phased fitting strategy is a step-by-step parameter identification method that first separates the humidity effect and then solves for the temperature effect, used to decouple the effects of temperature and humidity coupling on lifetime. The empirical statistical lifetime model is a semi-empirical model that describes the quantitative relationship between temperature, humidity, and characteristic lifetime, based on fitting experimental data. The mass transfer mechanism lifetime model is a physical lifetime model derived from the microscopic mass transfer process based on the moisture diffusion and adsorption mechanism. For example, the empirical statistical lifetime model adopts the Peck Accelerated Life Model, while the mass transfer mechanism lifetime model couples Fick's Law of Diffusion, the Guggenheim-Anderson-de Boer Isothermal Adsorption Model (GAB), and the Arrhenius Temperature Relation.
[0055] In this embodiment, the aforementioned phased fitting and model building can be performed in three ways. The first is a two-stage construction method: humidity decoupling followed by temperature fitting. First, using lifetime data from multiple humidity conditions at the same temperature, the humidity acceleration index is calculated. After fixing the humidity index, the lifetime data is converted into a humidity-corrected logarithmic lifetime. Then, a linear regression is performed with the inverse of the absolute temperature to solve for the activation energy and precondition constant, completing the empirical statistical lifetime model construction. Simultaneously, based on a denoised degradation sequence, the diffusion coefficient and equilibrium moisture absorption are fitted condition-by-condition, respectively constructing the temperature dependence of the diffusion coefficient and the temperature and humidity dependence of the equilibrium moisture absorption. The coupling performance and the linear relationship of moisture absorption complete the mass transfer mechanism lifetime model construction. This method has clear physical meanings for the parameters, effectively decoupling the temperature and humidity coupling effect and avoiding parameter distortion caused by multicollinearity. The second method is an iterative construction method using single-factor iteration and global constraints. Initial values for the humidity index and activation energy are first initialized. The humidity and temperature parameters are iteratively optimized alternately. Physical interval constraints are applied to the parameters during the iteration process until the fitting error converges, simultaneously completing the parameter solutions for both types of models. This method offers higher fitting accuracy and fully utilizes data from all operating conditions, making it suitable for scenarios with a large number of operating conditions. Thirdly, it employs a hierarchical, cascaded dual-model collaborative construction approach. First, the temperature sensitivity of the macroscopic lifetime is obtained through an empirical statistical model. Then, this temperature sensitivity is used as a constraint in the mass transfer mechanism model to correct the fitting range of the diffusion activation energy, achieving parameter feedback between the empirical and mechanistic models. This method ensures consistency in the temperature response patterns of the two models, reducing prediction bias between them.
[0056] After the parameters of the two types of models are solved, the model validity is verified and solidified.
[0057] For example, the implementation methods for model validation and solidification include two types. The first is leave-one-condition cross-validation, where one condition is reserved as the validation set each time. The model parameters are fitted using the remaining condition data, and the fitted model is used to predict the lifetime value of the reserved condition. The relative error between the predicted value and the experimental value is calculated. Cross-validation is completed by traversing all conditions. If the average relative error is lower than a preset threshold, the model is deemed valid, and the model parameters are solidified. This method can comprehensively verify the extrapolation stability of the model, avoid the risk of overfitting, and ensure the predictive reliability of the model within the condition range.
[0058] The second method is full-cycle degradation trajectory verification. This not only verifies the prediction accuracy of the failure lifetime point, but also generates full-cycle degradation prediction curves based on the two models, compares them with the complete trajectory of the denoised degradation sequence, and calculates the goodness of fit for the entire cycle. If the goodness of fit meets the standard, the model is deemed effective. This method verifies the model's rationality from the entire degradation process, avoiding process bias caused by fitting only a single point of the lifetime, and improving the accuracy of the model's dynamic prediction.
[0059] After completing the model parameter fitting and validity verification, the final model system can be output in the following way.
[0060] In one optional approach, separate parameter files are output for the empirical statistical lifetime model and the mass transfer mechanism lifetime model, respectively. These files include model formulas, parameter values, physical meanings of the parameters, applicable temperature and humidity ranges, and fitting accuracy indices. The two models are independent and can be used separately. This method offers a clear output structure, flexible calling options, and allows selection of the appropriate model based on different scenario requirements.
[0061] In another alternative approach, the two types of models are encapsulated and integrated to establish a unified input-output interface. After inputting temperature and humidity parameters and failure criteria, the prediction results and confidence weights of both models can be output simultaneously, forming an integrated dual-model prediction system. This method is convenient to use and can directly output the collaborative results of the two models.
[0062] In another alternative approach, the dual-model phased fitting construction unit within the nylon material lifespan prediction system performs the task. This unit incorporates four functional modules: a working condition data grouping and matching module, an empirical statistical model phased fitting module, a mass transfer mechanism model hierarchical fitting module, and a model parameter physical verification module. Each module works collaboratively according to decoupled fitting logic to construct two types of lifespan models. The working condition data grouping and matching module receives the median characteristic lifespan dataset and the full denoised degradation sequence from the upstream output. It performs data splitting and matching according to the dimensions of temperature gradient and humidity level, and filters to generate a lifespan statistical subset with multiple humidity levels at the same temperature and a moisture absorption degradation sequence subset covering the entire temperature range. These subsets serve as the fitting input data sources for the two types of models, providing a standardized data foundation for phased decoupled fitting. The empirical statistical model's phased fitting module embeds the Peck Accelerated Life Model, employing a phased fitting strategy that decouples temperature and humidity to perform parameter identification: In the first stage, it calls a statistical subset of lifetimes with different humidity levels at the same temperature, and eliminates the temperature term and the precondition constant through power-law division to obtain the humidity acceleration index; In the second stage, it fixes the humidity acceleration index and performs least-squares linear regression on the humidity-corrected logarithmic lifetime and the inverse of the absolute temperature to obtain the activation energy and the precondition constant, thus completing the full parameter fitting of the empirical statistical lifetime model. The mass transfer mechanism model is fitted in a hierarchical module that couples Fick's Law of Diffusion, the Guggenheim-Anderson-deBoer Isothermal Adsorption Model (GAB), and the Arrhenius Temperature Relation. Fitting is performed in three steps according to the mass transfer hierarchy: First, based on a subset of the hygroscopic degradation sequence, a two-dimensional Fick diffusion equation is fitted for each operating condition, outputting the equilibrium hygroscopic amount and moisture diffusion coefficient under each temperature and humidity condition. Second, the GAB isothermal adsorption relationship between equilibrium hygroscopic amount and relative humidity, and the Arrhenius kinetic relationship between diffusion coefficient and temperature are fitted respectively, solving for a complete set of mass transfer physical parameters, including monolayer hygroscopic parameters, adsorption constant, pre-exponential diffusion factor, and diffusion activation energy. Third, linear transformation relationships between size and hygroscopicity, and strength and hygroscopicity are embedded, combined with the failure critical threshold to complete model coupling and construct a complete mass transfer mechanism lifetime model. The physical verification module for model parameters retrieves the reasonable range of values for polymer material degradation parameters and the goodness-of-fit threshold. It performs physical meaning verification and fitting accuracy verification on all fitting parameters of the two types of models. After confirming that the parameter values are within a reasonable range and the goodness-of-fit meets the preset requirements, it archives and stores the model parameters, fitting curves and model call interfaces in a unified manner, providing a callable model basis for subsequent lifetime prediction calculations.
[0063] In an exemplary scheme defining a dual-model system, the median characteristic lifetime dataset under all operating conditions and the denoised degradation sequence dataset are first loaded as the basic data sources for model fitting. Then, an empirical statistical lifetime model is fitted in stages. In the first stage, lifetime data under two different humidity levels at 60 degrees Celsius are extracted, and the humidity acceleration exponent is solved using a power-law relationship. In the second stage, the humidity exponent is fixed, and after humidity correction of the lifetime data for all operating conditions, a least-squares linear regression is performed with the inverse of absolute temperature to obtain the activation energy and pre-exponential constant, completing the parameter identification of the empirical statistical lifetime model. Next, a mass transfer mechanism lifetime model is fitted in stages. In the first stage, the moisture diffusion coefficient at each temperature is obtained based on the hygroscopic mass sequence for each operating condition, and the diffusion activation energy and pre-exponential factor are obtained through fitting using the Arrhenius equation. In the second stage, the parameters of the GAB isothermal adsorption model are fitted based on the equilibrium moisture absorption data for each operating condition to establish the temperature and humidity calculation relationship for equilibrium moisture absorption. In the third stage, the linear conversion coefficient between size / strength and moisture absorption is embedded, and combined with the failure threshold, the mass transfer mechanism lifetime model is constructed. Next, model accuracy verification was performed, reserving a 30-degree water bath condition as a validation set. The characteristic lifespan of this condition was predicted using both types of models, and the prediction error was calculated and confirmed to be within a reasonable range. Finally, the model was encapsulated, outputting a complete dual-model system containing model formulas, parameter sets, applicable scope, and accuracy indicators for subsequent service environment lifespan prediction.
[0064] It should be noted that in some special cases, if the target service environment exceeds the test temperature and humidity range, it will be marked as an extrapolation prediction scenario, and the prediction results of the mass transfer mechanism lifetime model will be the main prediction by default, with the results of the empirical statistical model as a reference.
[0065] Step S40: Based on the empirical statistical lifetime model and the mass transfer mechanism lifetime model, the temperature and humidity parameters of the target service environment are substituted into the model, and the target predicted lifetime of the nylon sample is obtained through dual-path collaborative calculation.
[0066] The temperature and humidity parameters of the target service environment are the actual temperature and relative humidity values of the service scenario for which the predicted lifetime is to be determined. Dual-path collaborative computation is a calculation mode that simultaneously solves for lifetime using both empirical statistical and mass transfer mechanism paths, and dynamically fuses the results from both paths. The target predicted lifetime is the final output, corresponding to the material dimensional failure lifetime and performance degradation lifetime under the target service environment. For example, if the target service environment is a normal temperature and high humidity scenario, the output results include the median characteristic lifetime and the full-cycle dimensional change curve.
[0067] In this embodiment, the aforementioned dual-path collaborative lifetime prediction can be performed in three ways. The first is dynamic weighted fusion prediction. Based on the deviation between the target operating condition and the experimental operating condition interval, and the prediction time span, dynamic confidence weights are assigned to the empirical statistical path and the mass transfer mechanism path, respectively. The smaller the deviation and the shorter the prediction period, the higher the weight of the empirical model; conversely, the higher the weight of the mechanism model. The weighted sum of the two results yields the final target predicted lifetime. This method can adaptively match different prediction scenarios, fully leveraging the advantages of both models while balancing prediction accuracy and extrapolation reliability. The second is upper and lower bound interval envelope prediction. Using the empirical model prediction result as the lower bound and the mechanism model prediction result as the upper bound, the lifetime interval including the upper and lower boundaries is output as the prediction result, along with the confidence level of the interval. This method can quantify the uncertainty of the prediction, is suitable for reliability margin design scenarios, and provides a conservative and optimistic dual-dimensional reference for the lifetime design of structural components. Thirdly, it employs multi-dimensional collaborative verification and prediction. Two models output full-cycle dynamic degradation curves, and the trends and deviations of the two curves are compared. In segments with deviations less than a preset threshold, the average value is taken as the prediction result; in segments with larger deviations, the result from the mechanistic model is used, and deviation warnings are issued simultaneously. This method enables self-verification of the prediction process, timely identification of high-deviation prediction segments, and improves the reliability of the results.
[0068] After inputting the temperature and humidity parameters of the target service environment, the dual-path parallel computing process is initiated.
[0069] For example, the dual-path collaborative computation can be implemented in two ways. The first is a serial, step-by-step computation. First, an empirical statistical lifetime model is called, and temperature and humidity parameters are input to obtain the initial value of the empirical path lifetime and the power function degradation rate coefficient. Then, a mass transfer mechanism lifetime model is called to calculate the equilibrium moisture absorption and diffusion coefficient in sequence, and these are substituted into the diffusion equation to obtain the initial value of the mechanism path lifetime and the dynamic moisture absorption evolution parameters. Finally, the results from both paths are fused to obtain the final target predicted lifetime. This method has a clear computational logic, strong process controllability, and facilitates step-by-step error checking.
[0070] The second approach is parallel distributed computing, which simultaneously launches empirical model calculation threads and mechanistic model calculation threads. These two threads solve the problem independently in parallel, each outputting its own lifetime results and degradation process data. Once both calculations are complete, they are fed into a fusion processing module to output the combined target predicted lifetime. This method offers higher computational efficiency, fully utilizes computing resources, and shortens prediction time.
[0071] Once the initial values for the two paths are obtained, the final target predicted lifetime can be generated based on the following method.
[0072] In one alternative approach, a preset weighting rule is invoked, and the deviation is calculated based on the position of the target temperature and humidity parameters within the test condition range. The two weights are then dynamically allocated based on the deviation, and a weighted average is used to calculate the baseline predicted lifetime. This baseline lifetime is then corrected by incorporating model fitting errors, and the target predicted lifetime is output in single-point numerical form. This method offers a simple and intuitive output with minimal computational complexity.
[0073] In another optional approach, the full-cycle degradation prediction curves of the two models are output separately. The corresponding failure times are then marked using failure criteria, and the deviation range and confidence level of the two results are provided, forming a complete prediction result package containing lifetime values, degradation curves, and confidence intervals. This method provides rich output information and can intuitively display the full-cycle degradation process.
[0074] In another optional scheme, the life prediction is performed by a dual-path collaborative life prediction unit within the nylon material life prediction system. This unit comprises five functional modules: an environmental parameter standardization preprocessing module, an empirical statistical path solving module, a mass transfer mechanism path deduction module, a confidence weight dynamic allocation module, and a prediction result verification and output module. Each module collaboratively completes the life calculation and output under the target service environment according to the prediction data flow. The environmental parameter standardization preprocessing module receives the target service environment temperature and humidity parameters input by the user. It first performs dimensional normalization conversion, converting Celsius temperature to thermodynamic absolute temperature and relative humidity percentage to decimal form. Then, it compares and verifies the normalized parameters with the test condition intervals used for model fitting, marking interpolation or extrapolation attributes and quantifying the deviation from the test conditions. Finally, it generates a standardized environmental parameter set that simultaneously adapts to the input formats of two types of life prediction models, and synchronously outputs the test condition deviation index for subsequent confidence weight allocation. The empirical statistical path solving module calls the Peck Accelerated Life Model, which has already undergone parameter identification. It substitutes the standardized environmental parameter set into the model interface and directly solves for the initial empirical path lifetime based on pre-stored activation energy, humidity acceleration index, and preconditions. Simultaneously, it combines the material's intrinsic degradation time index and failure critical expansion to inversely solve for the power-law degradation rate coefficient under the target operating condition, supporting rapid extrapolation of the full-cycle dimensional degradation process. The mass transfer mechanism path extrapolation module couples the Fick's Law of Diffusion, the Guggenheim-Anderson-de Boer Isothermal Adsorption Model (GAB), and the Arrhenius Temperature Relation, which have already undergone hierarchical fitting. It first calls the isothermal adsorption calculation unit to input temperature and humidity parameters and calculates the equilibrium moisture absorption under the target operating condition. Next, the absolute temperature parameter is input into the diffusion kinetics calculation unit to calculate the moisture diffusion coefficient at the corresponding temperature. Then, using the failure critical moisture absorption as the solution objective, and combining the performance-moisture absorption linear transformation relationship, the result is substituted into the two-dimensional moisture diffusion equation to perform numerical root finding, outputting the initial value of the mechanism path lifetime. Simultaneously, dynamic moisture absorption evolution parameters of moisture absorption, outer diameter expansion, and strength retention over the entire cycle are generated. The confidence weight dynamic allocation module dynamically calculates the confidence weights of the two prediction results based on two core dimensions: the degree of deviation between the target condition and the experimental condition, and the time span of the prediction period. When the target condition is within the coverage area of the experimental condition and the prediction period is short, the weight of the empirical statistical path is increased to ensure fitting accuracy. When the extrapolation range of the target condition is large and the prediction time span is long, the weight of the mass transfer mechanism path is increased to enhance the reliability of physical constraints.The initial lifetime values of the two paths are weighted and fused based on the assigned weights to obtain the baseline predicted lifetime. The prediction result verification output module retrieves the preset physical constraint rules and applicable boundary conditions of the two types of models to perform a rationality check on the baseline predicted lifetime and dynamic moisture absorption evolution parameters, correcting abnormal prediction results that exceed the non-decreasing degradation law and material physical limits. Finally, the target predicted lifetime results, including dimensional failure lifetime, strength failure lifetime, and full-cycle degradation evolution curves, are output, which can provide quantitative data support for the service life design, environmental adaptability assessment, and operation and maintenance strategy formulation of nylon precision components.
[0075] In an exemplary scheme for determining the target predicted lifetime, the system first receives the input parameters of the target service environment temperature and relative humidity. Parameter dimension normalization and value range verification are then performed to confirm that the parameter format meets the model input requirements. Subsequently, dual-path parallel computation is initiated. The empirical path calls the empirical statistical lifetime model, directly solving for the initial empirical lifetime value by substituting the temperature and humidity parameters, and generating a full-cycle size prediction curve based on a power function degradation model. The mechanistic path sequentially calls the GAB isothermal adsorption model to calculate the equilibrium moisture absorption, and the Arrhenius model to calculate the diffusion coefficient. Substituting these values into the two-dimensional Fick diffusion equation, the failure time is numerically solved to obtain the initial mechanistic lifetime value, while simultaneously generating a dynamic moisture absorption evolution curve. Next, weight allocation is performed, calculating the relative deviation between the target operating condition and the experimental operating condition intervals, and dynamically calculating the confidence weights of the two models based on the prediction time span. Finally, weighted fusion is performed, weighting the two initial lifetime values to obtain the baseline predicted lifetime, and simultaneously fusing the two degradation curves to obtain a synergistic full-cycle degradation curve. Finally, the reasonableness of the execution result is verified to confirm that the prediction result is within the reasonable range of material physics, and the output includes the complete target predicted lifetime, which includes the median characteristic lifetime, the full-cycle degradation curve, and the confidence interval.
[0076] It should be noted that in some special cases, if the target service environment has temperature and humidity cycles, it should be converted into constant temperature and humidity stress before the prediction is performed, or the cycle spectrum can be directly input to carry out segment-by-segment cumulative life calculation.
[0077] Specifically, to address the need for assessing the damp-heat service life of nylon insulating sealing components inside offshore wind turbine towers, a nylon material service life prediction system is used for full-process calculation. This system consists of a multi-source degradation data acquisition unit, a sequence-preserving noise reduction and service life statistics unit, a dual-model phased fitting unit, and a dual-path collaborative prediction unit, all operating collaboratively in the order of data flow. The multi-source degradation data acquisition unit manages multiple sets of constant temperature and humidity test chambers and constant temperature water baths. Nylon 66 disc samples and standard tensile strips, vacuum-dried at 80 degrees Celsius to a mass change rate of less than 0.01%, are placed in environments covered by 30°C, 50°C, and 60°C water baths. In multiple simulated marine humid and hot conditions with temperature gradients of 0°C and 70°C, and 75%RH, 85%RH, 95%RH, and water bath saturation humidity, aged samples were taken out in batches according to a preset sampling cycle with denser sampling at the beginning and sparser sampling at the end. After being wiped and left to stand, the samples were prepared as test samples. The outer diameter was measured at three points, the mass was weighed with high precision, and the tensile strength was tested in sequence. The data was automatically archived and generated as original degradation time sequence data of each group of nylon samples under each temperature and humidity condition, arranged in an orderly manner according to the time cycle, including the outer diameter, mass, and tensile strength of the samples. This solved the problems of limited coverage of conventional test conditions and single data dimension.The order-preserving denoising and lifetime statistics unit uses the embedded Pool Adjacent Violators Algorithm (PAVA) to process the original degradation time series data, eliminating numerical drop anomalies caused by condensation fluctuations and measurement errors in the simulation test. This yields a denoised degradation sequence that conforms to the non-decreasing law of hygroscopic expansion. Combined with the failure criterion of 0.5% relative expansion of precision marine components, the failure crossing point is located. Linear interpolation is used to solve for the precise failure time of a single sample. After outlier removal, the median characteristic lifetime corresponding to each working condition is statistically obtained, effectively improving the stability and physical consistency of the initial lifetime calculation. The dual-model phased fitting unit uses the median characteristic lifetime and the denoised degradation sequence as the fitting basis, employing a phased fitting strategy that decouples temperature and humidity. On one hand, it uses a two-step method—separating the humidity effect under the same temperature condition and regressing the temperature effect from the full temperature range data—to construct a Peck accelerated life model with concise parameters and high computational efficiency. The model is an empirical statistical lifetime model. On the other hand, it fits diffusion parameters under each working condition, identifies isothermal adsorption laws and kinetic laws in layers, and couples Fick's Law of Diffusion, the Guggenheim-Anderson-de Boer Isothermal Adsorption Model (GAB), and the Arrhenius Temperature Relation to construct a mass transfer mechanism lifetime model with well-defined physical constraints. The parameter values of both types of models are within the reasonable range for polymer materials, which solves the problem that single empirical models are difficult to decouple the temperature and humidity coupling effect and have insufficient extrapolation accuracy under wide working conditions. The dual-path collaborative prediction unit takes the temperature and humidity parameters of the marine target service environment into account, calls two types of models to solve independently to obtain two initial lifetime values and dynamic degradation parameters, and then dynamically assigns confidence weights according to the deviation between the target operating conditions and the test operating conditions. After weighted fusion and physical boundary rationality verification, it finally outputs the target predicted lifetime and full-cycle dimensional evolution law of the nylon sample. It can fully adapt to the full-cycle lifetime prediction needs of precision nylon components in complex humid and hot marine environments, and provide quantitative data support for the service cycle design, operation and maintenance strategy formulation and reliability management of offshore wind power equipment.
[0078] Further, please refer to Figure 2 , Figure 2This is a flowchart illustrating the lifespan prediction process for nylon materials used in this application. The lifespan prediction was conducted using nylon 66 as the research object. First, the test materials underwent pretreatment: Nylon 66 samples from the same batch of injection-molded materials were used, including thin circular samples and standard tensile test strips conforming to GB / T1040.2 / 1A / 50. A unified molding process ensured the consistency of material crystallinity, internal stress, and initial state, eliminating the interference of material dispersion on the test results from the source. Before the formal test, the samples were placed in a vacuum oven with an accuracy of ±0.1 degrees and continuously dried at 80 degrees Celsius for at least 24 hours until the mass change between two consecutive weighings was less than 0.01%, confirming the complete removal of residual moisture inside the sample. After drying, the samples were immediately transferred to a desiccator to cool to room temperature to avoid secondary adsorption of environmental moisture. Immediately after pretreatment, initial outer diameter and initial mass measurements, along with environmental exposure tests, were conducted to ensure the uniformity of the initial state to the greatest extent possible. The experiment used constant temperature and humidity chambers with an accuracy of ±1 degree and ±2%RH, constant temperature water baths with an accuracy of ±0.1 degree, analytical balances with an accuracy of 0.1 mg, digital micrometers with an accuracy of 0.01 mm, electronic universal testing machines, and electron microscopes as core experimental equipment. Combining the principle of consistency between the actual service environment of Nylon 66 and the accelerated testing mechanism, four temperature gradients of 30 degrees, 50 degrees, 60 degrees, and 70 degrees were set up, along with three humidity gradients of 75%RH, 85%RH, and 95%RH, as well as water bath immersion saturated humidity conditions at corresponding temperatures. This constructed a condition matrix covering medium-temperature and high-humidity to high-temperature and high-humidity conditions, which not only met the decoupling requirements of staged parameter fitting but also provided independent condition samples for subsequent model verification. In the initial measurement phase, all circular specimens and tensile strips were uniformly numbered and weighed. Three different measurement positions were marked on the outer diameter surface of the circular specimens with a thin thread to ensure that subsequent measurements were conducted at the same positions, eliminating systematic errors caused by positional deviations. After cooling to room temperature, the initial mass and initial outer diameter of each specimen were measured sequentially. Each set of dimensional measurements was repeated three times, and the average value was recorded as the initial baseline data. Subsequently, the accelerated aging test phase involving environmental exposure was initiated. The pretreated specimens were quickly transferred to a constant temperature and humidity chamber and a constant temperature water bath with preset parameters. Sufficient gaps were maintained between specimens to ensure full contact between each surface and the environment, avoiding mutual obstruction that could affect the uniformity of water absorption. Periodic measurements were performed using a preset time period with higher frequency at the beginning and lower frequency at the end. The measurement frequency was appropriately increased when the water absorption rate was high in the initial stage, and the measurement interval was gradually increased as the degradation rate slowed down in the later stage, balancing data density and test efficiency.During each measurement, the sample is quickly removed, and the surface condensation and dust are gently wiped away with a soft cloth. Excessive pressure or wiping force should be avoided to prevent the internal moisture from being squeezed out. After the sample temperature returns to room temperature, the mass and outer diameter are measured within 1-2 minutes using an analytical balance and a digital micrometer. Simultaneously, five samples are taken and weighed quickly for each cycle, and then tensile strength tests are performed using an electronic universal testing machine. After measurement, the circular sample is immediately returned to the testing environment to minimize interference from the measurement operation on the sample's moisture absorption state. When the mass change rate after three consecutive measurements is less than 0.01% / day, the sample is considered to have reached moisture absorption equilibrium. The mass and outer diameter data at the equilibrium point are recorded. Simultaneously, samples with different water absorption rates are subjected to tensile testing, and their microscopic cross-sectional morphology is observed using an electron microscope to analyze the aging mechanism. After completing the full cycle of data acquisition, the original degradation timeline data of the nylon sample is determined. Subsequently, data preprocessing was performed, employing the Pool Adjacent Violators Algorithm (PAVA) for order-preserving denoising. This algorithm processed the original degradation time-series data to obtain a denoised degradation sequence, and failure criteria were simultaneously established. The median characteristic lifetime for each operating condition was calculated using linear interpolation based on these failure criteria. Then, the median characteristic lifetime and the denoised degradation sequence were used as the basis for model parameter fitting. A staged fitting strategy was employed, sequentially completing diffusion fitting based on Fick's Law of Diffusion, isothermal adsorption fitting based on the Guggenheim-Anderson-deBoer Isothermal Adsorption Model (GAB), and parameter fitting and performance-hygroscopic linear model construction based on the Arrhenius Temperature Relation. Simultaneously, an empirical statistical lifetime model and a mass transfer mechanism lifetime model were constructed. After the model is built, model verification is performed. If the verification is successful, lifetime prediction is carried out. Based on the empirical statistical lifetime model and the mass transfer mechanism lifetime model, the temperature and humidity parameters of the target service environment, namely the temperature T and relative humidity RH of the target operating condition, are substituted. The failure threshold is solved with respect to time t. The dimensional lifetime and strength lifetime of the nylon sample, i.e. the target predicted lifetime, are obtained through dual-path collaborative calculation.
[0079] Second Embodiment This embodiment provides an exemplary scheme for acquiring original degradation time series data under multi-gradient damp heat aging. In this example, nylon samples and standard tensile strips are first subjected to vacuum isothermal drying treatment to unify the initial moisture content. Then, the dried samples are placed in multi-gradient temperature and humidity environments to initiate a damp heat accelerated aging process. The aged samples are taken out according to a preset cycle of denser initial aging followed by sparser aging to complete the state adjustment. Finally, the original degradation time series data is constructed through synchronous testing of multi-dimensional parameters. Step S10 includes steps A11 to A14: Step A11: Place the nylon sample and the standard tensile test strip in a vacuum oven for constant temperature drying. When the mass change rate of two consecutive weighings is lower than the preset mass change rate, the dried sample is obtained.
[0080] Step A12: The dried samples are placed in batches in a constant temperature and humidity test chamber with multiple temperature and humidity gradients and a constant temperature water bath with multiple temperature gradients to obtain aged samples through the aging process.
[0081] Step A13: Obtain the aging samples at the corresponding workstations in batches according to the time cycle of denser samples at the beginning and sparser samples at the end, wipe and let stand to obtain the test samples.
[0082] Step A14: Perform three-point average measurement of outer diameter, high-precision weighing of mass, and tensile strength test on the test sample, and record the parameter values at each time point to obtain the original degradation time series data.
[0083] Vacuum oven constant temperature drying is a pretreatment process that continuously removes residual moisture from samples under negative pressure and constant temperature, unifying the initial state of the materials. It is a core foundational step for eliminating initial moisture content differences and ensuring the parallelism of tests. For example, continuous drying under 80-degree vacuum is used as the drying endpoint when the mass change rate between two consecutive weighings is less than 0.01%. The dried samples are nylon discs and standard tensile strips that have undergone full-process drying and whose internal moisture content has been reduced to a uniform benchmark level; they are standardized test subjects for accelerated aging tests. The constant temperature and humidity test chamber is a reliability testing device that can precisely control the internal ambient temperature and relative humidity parameters, used to simulate aging conditions under medium to high humidity environments at different gradients. The constant temperature water bath is a saturated water immersion test device that can precisely control the water temperature, used to provide extreme aging control conditions with 100% relative humidity. Aging samples are test samples placed in a humid and hot environment and undergoing moisture absorption aging for a certain period; they are the direct subjects of periodic parameter testing.
[0084] The pre-set time period of denser initial measurements followed by sparser later measurements is a non-equidistant measurement sequence scheme set according to the characteristic that nylon's moisture absorption rate is fast at first and then slows down. It is a time node planning rule that takes into account both the density of degradation data and the efficiency of test execution. For example, measurements are performed once a day in the early stage of aging, once every three days in the middle stage, and once a week in the later stage. The test sample is an aged sample that has undergone surface condensation wiping and temperature recovery to room temperature, and is in a standard test state. It is the direct object of measurement for dimensional, mass, and mechanical property testing. The three-point average measurement of the outer diameter is a dimensional testing method in which the values are measured sequentially at three pre-set marked points on the outer diameter circumference of the sample and the arithmetic mean is calculated. It is used to eliminate systematic measurement errors caused by sample roundness deviation and placement angle differences.
[0085] High-precision weighing is a testing procedure that uses an analytical balance with an accuracy of 0.0001 to accurately measure the mass of a sample, and is used to calculate the instantaneous moisture absorption rate and moisture diffusion progress of the sample.
[0086] In this example, when conducting damp heat aging tests and collecting raw degradation time-series data, the process can be carried out in parallel according to a working condition matrix. A sample allocation scheme for all temperature and humidity conditions can be pre-planned, and the aging process for all conditions can be started synchronously. Sampling tests and data entry can be completed synchronously at preset time points, thus achieving parallel acquisition of time-series data for all working conditions. Alternatively, a single-chamber gradient progression method can be used, conducting aging tests sequentially in the same testing equipment according to a gradually increasing temperature and humidity gradient. The same batch of samples undergoes different gradients of damp heat treatment sequentially, completing data acquisition for corresponding working conditions in stages, and accumulating raw degradation time-series data for all working conditions, thereby reducing the equipment requirements.
[0087] After the preparation of the dried samples and the initiation of the wet heat aging process, a periodic data acquisition process is initiated. According to the preset time nodes with denser data at the beginning and sparser data at the end, the aging samples at the corresponding workstations are taken out in batches. After gently wiping away the condensation and dust adhering to the surface, they are left to stand at room temperature to prepare test samples with consistent conditions. Then, the test samples are subjected to three-point average measurement of outer diameter, high-precision weighing of mass, and tensile strength testing in sequence. The test values of each parameter and the corresponding aging time are recorded simultaneously. The data are classified and archived according to the working condition number and sample number. Finally, a complete original degradation time series dataset is constructed. In this way, the consistency and comparability of the data of each working condition are ensured through standardized sampling and testing procedures, and additional random measurement noise introduced by operational differences is avoided.
[0088] For example, there are two methods for acquiring raw degradation time-series data. The first is continuous tracking acquisition of the same sample. From the start to the end of the aging test, the same batch of disc samples and tensile strip samples are tracked. At each measurement time point, the target sample is quickly taken out, and after all parameter tests are completed, it is immediately returned to the corresponding aging environment to continue service. The full-cycle degradation time-series data of a single sample is accumulated sequentially. After each time point test is completed, the physical rationality of the data is verified simultaneously, and abnormal measurement points are marked. After all preset time points have been tested, the time-series data of all samples are summarized to obtain the raw degradation time-series dataset. This method adopts the test logic of continuous tracking of the same sample throughout the entire cycle. Through the strong correlation between the data of a single sample before and after, the continuous degradation process of a single sample is completely restored, eliminating the interference of sample discreteness on the degradation trend, and the temporal continuity of the data is stronger.
[0089] The second method involves batch sampling at fixed points. All samples are pre-divided into multiple independent test batches based on measurement time points. Each batch corresponds to a preset measurement time point. Upon reaching the corresponding time point, all samples in the batch are extracted, and all parameters of the outer diameter, mass, and tensile strength of all samples in that batch are tested in one go. After testing, the samples are not returned to the aging environment. After each batch is tested, the data for that batch is archived and preliminarily verified simultaneously. Once all batches have been tested, the test data from all batches are integrated along a timeline to obtain the original degradation time-series dataset. This method employs the logic of independent batch sampling and one-time full-parameter testing. By avoiding repeated entry and exit of samples from the aging environment, it eliminates interference from sampling operations on the moisture absorption steady state inside the sample, resulting in higher stability and accuracy of the test data. Simultaneously, destructive tests such as cross-sectional morphology observation can be carried out, effectively enriching the dimensions of the experimental data.
[0090] Furthermore, taking the evaluation of the damp heat service life of nylon 66 insulating seals for high-voltage connectors in new energy vehicles as a specific application scenario, the material pretreatment process is first completed: thin circular samples and dumbbell-shaped standard tensile specimens conforming to GB / T1040.2 / 1A / 50 are injection molded from the same batch. A unified molding process ensures the consistency of material crystallinity, internal stress, and initial state, eliminating material dispersion interference from the source. The nylon samples and standard tensile specimens are placed in a vacuum oven with an accuracy of ±0.1 degrees for constant temperature vacuum drying at 80 degrees Celsius for more than 24 hours. When the mass change rate of two consecutive weighings is less than the preset mass change rate threshold of 0.01%, it is confirmed that the internal residual moisture has been completely removed, resulting in a dried sample. After drying, the sample is immediately transferred to a desiccator to cool to room temperature, and the initial outer diameter and initial mass are measured immediately to minimize the test error caused by differences in initial moisture content. Subsequently, accelerated aging tests were initiated. The dried samples were placed in batches in constant temperature and humidity test chambers with multiple temperature and humidity gradients and constant temperature water baths with multiple temperature gradients, constructing a complete test matrix covering constant temperature and humidity conditions with temperature gradients of 50°C, 60°C, and 70°C and relative humidity of 85% and 95%, as well as water bath immersion conditions with saturated humidity at temperature gradients of 30°C, 50°C, 60°C, and 70°C. Sufficient gaps were maintained between samples to ensure full contact between each surface and the environment. Aged samples were obtained through a continuous damp heat aging process. During the test cycle, according to the time period of denser initial aging followed by sparser aging, the aged samples were taken out in batches from the corresponding workstations. After gently wiping the condensation and dust adhering to the surface with a lint-free soft cloth, they were allowed to stand at room temperature to prepare test samples with consistent conditions. Then, the test samples were subjected to three-point average outer diameter measurement, high-precision mass weighing, and tensile strength testing. The parameter values at each time point were recorded simultaneously, and the original degradation time series data of the nylon samples were obtained. Next, data preprocessing is performed, and the Pool Adjacent Violators Algorithm (PAVA) is used to carry out order-preserving denoising processing to eliminate random measurement noise and numerical drop anomalies introduced by instrument fluctuations and operational differences. This results in a denoised degradation sequence that conforms to the non-decreasing physical law of hygroscopic expansion. Simultaneously, two types of failure criteria, relative expansion amount and strength retention rate, are established. The median characteristic lifetime corresponding to each working condition is obtained by combining linear interpolation.Subsequently, using the median characteristic lifetime and the denoised degradation sequence as the fitting basis, a staged fitting strategy was employed to sequentially complete diffusion fitting based on Fick's Law of Diffusion, isothermal adsorption fitting based on the Guggenheim-Anderson-deBoer Isothermal Adsorption Model (GAB), and kinetic parameter fitting and performance-hygroscopic linear model construction based on the Arrhenius Temperature Relation. Simultaneously, an empirical statistical lifetime model and a mass transfer mechanism lifetime model were constructed. After model construction, model validation was performed to confirm prediction accuracy. Upon successful validation, lifetime prediction was conducted. The temperature and humidity parameters of the target service environment were input, and the failure threshold was solved against time. Through dual-path collaborative calculation, the dimensional lifetime and strength lifetime of the nylon sample, i.e., the target predicted lifetime, were obtained. This provides quantitative data support for the service life design, environmental adaptability assessment, and tolerance margin design of high-pressure connector seals.
[0091] Third Embodiment This embodiment provides an exemplary scheme for calculating median characteristic lifetime based on order-preserving denoising. In this example, the original degradation time series data is first split into multiple independent degradation subsequences according to the test conditions and sample numbers. Then, an order-preserving denoising algorithm with adjacent violation point pooling is used to process each degradation subsequence to eliminate non-physical fallback noise. Subsequently, based on two types of failure criteria, relative expansion and intensity retention, the threshold crossing node is located, and the failure time of a single sample is solved by linear interpolation. Finally, after removing discrete outliers, the median of the remaining failure times is taken, and the median characteristic lifetime of the entire test conditions is obtained. Step S20 includes steps B11 to B14: Step B11: The original degradation time series data is split into groups according to the test conditions and sample numbers to generate multiple independent degradation subsequences.
[0092] Step B12: Process the degraded subsequence according to the order-preserving denoising algorithm, and merge adjacent data blocks that do not satisfy monotonicity through the adjacent violation point pooling algorithm to obtain the denoised degraded sequence.
[0093] Step B13: Based on the relative expansion failure coefficient and the strength retention rate failure coefficient, locate the data points before and after the failure threshold in the denoising degradation sequence, and calculate the failure time through linear interpolation.
[0094] Step B14: For all samples under the same working condition, perform discrete outlier removal on the failure time, take the median of the target failure time as the characteristic lifetime of the working condition, and summarize to obtain the median characteristic lifetime.
[0095] A degraded subsequence is an independent time-series data unit corresponding to a single sample's single-dimensional degradation process, obtained by decomposing the original degraded time-series data under all operating conditions according to operating condition attributes and sample identity. It is the smallest processing unit of the order-preserving denoising algorithm. Examples include the outer diameter degradation subsequence, mass degradation subsequence, and tensile strength degradation subsequence for a single sample. The order-preserving denoising algorithm, under the constraint of strictly preserving the non-decreasing physical trend of the degraded data, eliminates noise from measurement random fluctuations and abnormal drops, and is the core processing method to ensure the physical consistency of the degraded data. The PoolAdjacentViolators Algorithm (PAVA) is the core operational logic for implementing the order-preserving constraint. It iteratively merges adjacent data blocks that do not satisfy monotonicity and recalculates the weighted mean, making the output sequence globally satisfy the non-decreasing constraint. For example, when the mean of a subsequent data block is less than that of a previous data block, the two blocks are immediately merged, and a new block mean is calculated using the sample size as the weight.
[0096] The denoised degradation sequence is the degradation time series data that globally satisfies non-decreasing physical laws after order-preserving denoising processing. It serves as the standardized data foundation for calculating the failure time and fitting subsequent models. The relative expansion failure coefficient is the critical expansion ratio parameter for determining dimensional failure. Combined with the initial outer diameter of the sample, the absolute threshold for dimensional failure can be calculated, serving as a quantitative indicator characterizing the service limit of material dimensional deviations. For example, a relative expansion failure coefficient of 0.5% means that dimensional failure is defined as an expansion of the outer diameter by 0.5% relative to the initial value. The strength retention rate failure coefficient is the critical strength retention ratio parameter for determining mechanical property failure. Combined with the initial tensile strength of the sample, the absolute threshold for strength failure can be calculated, serving as a quantitative indicator characterizing the failure limit of material mechanical properties. For example, a strength retention rate failure coefficient of 80% means that strength failure is defined as a decrease in tensile strength to 80% of the initial value. The failure time is the aging time corresponding to the first time the degradation index of the sample reaches the failure threshold, serving as a core quantitative indicator characterizing the service life of a single sample.
[0097] In this example, when splitting the original degraded time-series data to generate degraded subsequences, a hierarchical splitting approach prioritizing operating conditions can be used. First, the original data is grouped into primary groups based on temperature and humidity conditions. Then, within each operating condition, secondary splitting is performed according to sample number, generating single-sample, single-dimensional degraded subsequences one by one, thus completing the standardized splitting of the entire dataset. Alternatively, a parallel splitting approach can be used, simultaneously initiating the splitting process for the three data dimensions of outer diameter, mass, and intensity. This parallel processing completes the subsequence segmentation of all operating conditions and samples, generating multiple sets of independent degraded subsequences in batches, thereby improving the processing efficiency of data splitting.
[0098] After splitting the degraded subsequences, a denoising process preserving order is initiated. For each group of degraded subsequences, an adjacent violation point pooling algorithm is used to iteratively merge adjacent data blocks that do not satisfy monotonicity until the entire sequence satisfies the strict non-decreasing constraint. This process maps and restores the sequence to a denoised degraded sequence of the same length as the original sequence. Subsequently, based on preset relative expansion failure coefficients and strength retention rate failure coefficients, and combined with the initial baseline value of the sample, the corresponding absolute failure threshold is calculated. The first data point greater than or equal to the failure threshold and its preceding adjacent nodes are located in the denoised degraded sequence, constructing failure-crossing node pairs. The failure time of a single sample is calculated using linear proportional inverse kinematics. Finally, discrete outlier removal is performed on the failure times of all samples under the same working condition. The median of the remaining valid failure times is taken as the characteristic lifetime of that working condition. The results of all working conditions are summarized to obtain the median characteristic lifetime for all working conditions. By using denoising processing based on physical constraints and statistically robust lifetime calculation, the accuracy and reliability of the median characteristic lifetime calculation are improved, avoiding lifetime assessment bias caused by measurement noise.
[0099] For example, there are two methods for calculating the median characteristic lifetime from degraded subsequences. The first method is single-sample serial iteration with order preservation and point-by-point interpolation statistics. Starting from the first degraded subsequence of the first working condition, adjacent violation point pooling order preservation processing is performed sequentially for each single-sample degraded subsequence. After denoising of each subsequence is completed, failure threshold location and linear interpolation are performed simultaneously to calculate the failure time of the sample and temporarily store it. After all failure times of all samples under the same working condition have been calculated, the Grubbs test is used to identify and remove discrete outliers. The remaining valid failure times are sorted in ascending order and the median is taken to obtain the median characteristic lifetime of the working condition. The above process is executed serially for each working condition. After all working conditions have been processed, the median characteristic lifetime of the entire working condition is obtained by summing the results. This method adopts the calculation logic of single-sample serial iteration and progressive statistics for each working condition. Through independent denoising and interpolation closed-loop processing of a single sample, the physical consistency of the degradation trend of each sample is ensured. The calculation process is highly controllable and the intermediate results are traceable.
[0100] The second approach involves batch parallel denoising and median lifetime calculation based on load conditions. First, all degraded subsequences are grouped in batches according to load condition dimensions. Simultaneously, parallel order-preserving denoising processing for multiple load conditions is initiated. Within each load condition, adjacent violation point pooling operations are performed in parallel for all sample degraded subsequences, generating denoised degraded sequences for all samples under the same load condition in batches. After full-load condition denoising is completed, failure time interpolation calculations for all samples are performed simultaneously. Then, Weibull distribution parameters are fitted to all failure times under the same load condition. The quantile value corresponding to the 50% failure probability is directly calculated using the distribution function as the median characteristic lifetime. Outlier distribution verification and removal are performed simultaneously. After all load conditions are calculated in parallel, the median characteristic lifetime for all load conditions is obtained by summing the results. This method employs a load condition-level batch parallel processing and probability distribution fitting calculation logic. Parallel computation improves processing efficiency under large sample sizes, and the robustness of median lifetime estimation is enhanced by relying on the statistical properties of the Weibull distribution.
[0101] Furthermore, to address the issue of measurement noise during the water absorption and swelling process of nylon materials causing data to violate physical monotonicity, data preprocessing and characteristic lifetime calculation were performed. The water absorption and swelling of nylon materials is a physical process of continuous diffusion of water molecules inwards, macroscopically manifested as a strictly non-decreasing change in outer diameter and mass. However, in actual measurements, factors such as fluctuations in instrument accuracy, deviations in sample placement angles, and differences in the degree of surface moisture wiping introduce random noise, leading to abnormal phenomena such as numerical declines in the measurement sequence. This violates physical laws and reduces the accuracy of subsequent lifetime calculations. Conventional methods such as moving averages and wavelet thresholding can only smooth noise statistically and cannot guarantee the monotonicity constraint of the output sequence, potentially resulting in physically unrealizable size shrinkage. Therefore, an adjacent violation point pooling algorithm (PAVA) is introduced. The algorithm performs order-preserving denoising. Under the constraint of ensuring a strictly non-decreasing output sequence, it minimizes the squared error between the filtered value and the original value, preserving the true degradation trend while eliminating measurement noise. During execution, each data point in the original degraded time series data is treated as an independent "block," recording the mean and number of samples in each block. Then, the block sequence is traversed sequentially from left to right. If the mean of the current block is less than the mean of the previous block, violating the non-decreasing constraint, the two blocks are merged into a new block. The mean of the new block is calculated using the number of samples as the weight. After each merge, the previous sequence is checked again from one position back to ensure that the merge operation does not introduce new violations, until the entire block sequence fully satisfies the non-decreasing condition. Finally, the block structure is restored to a denoised degraded sequence of the same length as the original sequence. Mathematically, it uses the original outer diameter sequence as input, constrains the output sequence to satisfy a non-decreasing relationship, and optimizes to minimize the sum of the squared errors between the output value and the original value. This method does not require preset window length and threshold parameters, has a high degree of adaptability, and is particularly suitable for data preprocessing of monotonous degradation processes such as water absorption and expansion. After noise reduction, failure criteria and life calculations were performed. Combining the assembly tolerance requirements of precision structural components with dimensional failure scenarios in engineering applications, relative expansion was adopted as the dimensional failure criterion. Specifically, dimensional failure was defined as when the outer diameter of the sample reached 1.005 times its initial value. A relative expansion threshold of 0.5% corresponds to the upper limit of tolerance in precision assembly scenarios. Exceeding this threshold will lead to engineering failures such as excessive clearance and decreased sealing performance, demonstrating clear practical engineering significance. The failure threshold was calculated using the initial outer diameter and the relative expansion failure coefficient, using the following formula: In the formula: D0: initial outer diameter of the sample (measured at time 0), in mm; r: relative expansion failure threshold, which is taken as r=0.005 (i.e. 0.5%) in this case. Using relative values as the judgment criterion for samples with different initial sizes can eliminate the differences in criteria caused by discrete initial sizes and ensure the consistency of failure judgment criteria under different samples and working conditions. Since the experimental data are measurement results at discrete time points, the failure time usually falls between two adjacent measurement time points. Therefore, a precise single-sample failure time needs to be obtained through linear interpolation. First, locate the first sampling point with a value greater than or equal to the failure threshold in the denoised degradation sequence. Extract the time parameters and performance parameters of this point and its preceding adjacent nodes to construct a failure-crossing node pair. Based on the assumption that the performance between the two points changes linearly with time, construct a proportional conversion relationship. The time corresponding to when the performance value equals the failure threshold is the single-sample failure time. The characteristic lifetime under operating conditions is characterized by the median failure time. First, discrete abnormal samples caused by defects in sample preparation or operational errors are removed. Then, the median of the failure times of all valid samples under the operating condition is taken as the median characteristic lifetime of the operating condition. Compared with the arithmetic mean, the median has a stronger resistance to interference from extreme values and is more consistent with the statistical characteristics of the lifetime data following the Weibull distribution. It can more robustly represent the overall lifetime level of the material under the operating condition. Finally, the process of processing the original degradation time series data to obtain the denoised degradation sequence through the order-preserving denoising algorithm, and calculating the median characteristic lifetime corresponding to the operating condition by combining the failure criterion with linear interpolation is fully realized.
[0102] Specifically, the linear interpolation method calculates the failure time: in the denoised outer diameter sequence, find the first condition that satisfies... Time point t i Take the previous time point and corresponding outer diameter Two points determine a straight line; the time when the outer diameter equals the failure threshold is the single-sample failure time. : In the formula :iso is an abbreviation for isotonic (order preservation, order preservation regression), where i represents the i-th measurement time point; physical meaning: after order preservation denoising, the outer diameter value of the sample corresponding to the i-th time point, that is, the outer diameter measurement value of the i-th position in the denoised degradation sequence. :D corresponds to Diameter (outer diameter), and fail represents failure; physical meaning: absolute threshold for dimensional failure, calculated from the initial outer diameter of the sample and the relative expansion failure coefficient, is the critical outer diameter value for determining whether the sample has experienced dimensional failure.
[0103] Fourth embodiment This embodiment provides an exemplary scheme for generating order-preserving denoised sequences based on adjacent violation point pooling. In this example, each time-series data point of the degraded subsequence is first initialized as an independent data block, and an initial ordered block set arranged by time is constructed. Then, by traversing and comparing the mean of adjacent blocks, monotonic violation adjacent block pairs are located. After performing weighted merging on the violation block pairs, the preceding sequence is re-examined. This process is repeated until the entire sequence satisfies the non-decreasing constraint to obtain a global order-preserving block set. Finally, the block mean is mapped back to the original time-series point to generate a denoised degraded sequence that conforms to physical monotonicity. Please refer to... Figure 3 , Figure 3 This is a flowchart illustrating the fourth embodiment of the method for predicting the service life of nylon materials according to this application. Step B12 includes steps C11 to C15: Step C11: Initialize the time-series data points in the degenerate subsequence as independent data blocks, and record the sample count and average value of the independent data blocks to construct an initial ordered block set arranged in chronological order.
[0104] Step C12: Iterate through the initial ordered block set sequentially, compare the mean values of adjacent independent data blocks, and locate the first monotonic violation adjacent block pair where the mean of the subsequent block is less than the mean of the preceding block.
[0105] Step C13: Merge the monotonicity violation adjacent block pairs into a single data block to obtain a merged data block.
[0106] Step C14: After replacing the monotonicity violation adjacent block pair based on the merged data block, backtrack one bit and re-verify the monotonicity of the preceding block until the entire sequence block set satisfies the non-decreasing constraint, thus obtaining the global order-preserving block set.
[0107] Step C15: Map the mean of each data block in the global order-preserving block set to the data points in the corresponding time interval of the degenerate subsequence one by one to obtain the denoised degenerate sequence.
[0108] Independent data blocks are the smallest processing units for order-preserving denoising operations. They are aggregated from one or more time-series data points. Each data block synchronously records the number of samples and the weighted average value it contains, serving as the fundamental object for monotonicity verification and merging operations. For example, initially, each time-series measurement point corresponds to an independent data block with one sample and the average value being the original measurement value of that point. The initial ordered block set is the initial set of all independent data blocks arranged in chronological order. It is the starting operation object for order-preserving iterative processing, completely preserving the temporal order and numerical information of the original data. Monotonicity violation adjacent block pairs are two adjacent data blocks in the time-series arrangement whose latter mean is less than the former mean, violating the non-decreasing physical constraint. They are the direct processing objects for iterative merging operations. The merged data block is a new data block generated by merging the violation adjacent block pairs. Its sample size is the sum of the sample sizes of the original two blocks, and its average value is the weighted average value calculated with the sample size as the weight. It is used to replace the original violation block pairs in subsequent monotonicity verification. The global order-preserving block set is the set of blocks after the full iterative merging is completed, in which all adjacent data blocks satisfy the non-decreasing constraint. It is the intermediate operation result of the order-preserving denoising process and fully preserves the monotonic trend characteristics of the degradation process.
[0109] In this example, when constructing the initial ordered block set, it can be done using a single-point independent initialization method. Each time-series data point in the degenerate subsequence is initialized as an independent data block, with each block corresponding to a measurement time point. After arranging them in ascending time order, the initial ordered block set is directly obtained, thus completing the standardized construction of the computational objects. Alternatively, a stationary segment pre-aggregation initialization method can be used. First, continuous stationary measurement segments in the degenerate subsequence with numerical fluctuations less than a preset threshold are pre-aggregated, merging multiple consecutive points into initial data blocks. Then, the initial ordered block set is constructed by arranging them in time order. This reduces the number of subsequent iterations and merges, improving the processing efficiency of long sequences.
[0110] After constructing the initial ordered block set, the monotonicity iterative verification process is initiated. The block set is traversed from left to right, comparing the mean values of two adjacent independent data blocks to identify the first monotonicity violation pair where the mean of the subsequent block is less than the mean of the preceding block. This violation pair is merged into a single merged data block. The weighted mean of the merged block is calculated using the sample size of each block as weights. The merged data block replaces the original violation pair, and the process is backtracked by one position to re-verify the monotonicity of the preceding adjacent blocks, preventing the merging operation from introducing new violations. This process of violation location, block merging, and backtracking is iteratively executed until all adjacent blocks in the entire block set satisfy the non-decreasing constraint, resulting in a globally ordered block set. Finally, the mean of each data block in the global order-preserving block set is mapped one by one back to all the original data points in the corresponding time interval of the degenerate subsequence, generating a denoised degenerate sequence of the same length as the original sequence. In this way, through iterative block merging and backtracking verification mechanism, the square error between the denoising result and the original data is minimized under the premise of ensuring non-decreasing physical constraints, thereby improving the physical rationality of denoising processing and data fidelity.
[0111] For example, there are two ways to obtain a denoised degenerate sequence from a degenerate subsequence. The first is a serial single-point iterative processing method. Starting from the beginning of the initial ordered block set, it traverses backwards one by one. Each time a pair of adjacent blocks with monotonicity violations is found, a merge operation is immediately performed to generate a merged data block. After merging, it automatically backs up one block position to re-verify the monotonicity of the preceding sequence, ensuring that the merging operation does not destroy the order preservation of the previously verified regions. After each backtracking re-verification, it continues to traverse backwards until the end of the sequence is reached and no new pairs of violating blocks appear. After confirming that the entire sequence satisfies the non-decreasing constraint, point mapping restoration is performed to obtain the denoised degenerate sequence. This method uses a point-by-point serial traversal and backtracking dynamic verification operation logic. The operation process does not require additional global scanning, the logic is simple and stable, and each step of the merging operation can be precisely controlled. The intermediate process is fully traceable.
[0112] The second approach employs a parallel reduction-based inter-regional processing method. This method splits the degenerate subsequence into multiple non-overlapping, equal-length sub-intervals. Simultaneously, iterative order-preserving denoising processing is performed independently on each sub-interval, resulting in a local order-preserving block set for each sub-interval. Then, monotonicity checks are performed on the boundary blocks of adjacent sub-intervals. If a pair of non-compliant blocks appears across interval boundaries, interval boundary merging and local re-checking are performed until all interval boundaries satisfy the non-decreasing constraint. Finally, a global order-preserving block set is obtained, and point mapping is performed to restore the denoised degenerate sequence. This method utilizes a computational logic of parallel inter-regional computation and boundary reduction checks. The parallel processing of the initial interval splitting significantly improves the computational efficiency of long sequences, while boundary reduction checks ensure global monotonicity.
[0113] Furthermore, the degenerate subsequence is processed using the Pool Adjacent Violators Algorithm (PAVA) to preserve order and denoise the data. First, the temporal data points in the degenerate subsequence are initialized as independent data blocks, and the sample count and mean value of each independent data block are recorded, constructing an initial ordered block set arranged chronologically. Then, the initial ordered block set is traversed from left to right, comparing the mean values of adjacent independent data blocks to locate the first monotonic violating adjacent block pair whose mean value is less than that of the preceding block. Next, the monotonic violating adjacent block pairs are merged into a single data block, resulting in a merged data block. The weighted mean of the merged data block is calculated using the sample count of the original violating block pair as the weight. After replacing the monotonic violating adjacent block pair with the merged data block, the monotonicity of the preceding block is re-verified by backtracking by one position. This process of violator location, block merging, and backtracking is repeated until the entire sequence block set satisfies the non-decreasing constraint, resulting in a global order-preserving block set. Finally, the mean of each data block in the global sequence-preserving block set is mapped one by one back to the data points in the corresponding time interval of the degraded subsequence to obtain the denoised degraded sequence. Under the premise of ensuring that the degraded data conforms to the non-decreasing physical law of water absorption and expansion, the square error between the denoised result and the original measurement value is minimized, effectively eliminating the numerical drop anomaly caused by measurement random noise, and providing a physically consistent and trend-accurate data foundation for subsequent failure moment calculation.
[0114] Fifth Embodiment This embodiment provides an exemplary scheme for calculating the single-sample failure time based on linear interpolation. In this example, the absolute thresholds for two types of failures, size and strength, are first calculated by combining the initial performance benchmark of the sample and the failure coefficient. Then, the first failure sampling node that exceeds the threshold is located in the denoised degradation sequence. Based on this node and its preceding adjacent nodes, a failure crossing node pair is constructed. Finally, the accurate failure time is obtained by inverse solution through linear proportional conversion. Please refer to... Figure 4 , Figure 4 This is a flowchart illustrating the fifth embodiment of the method for predicting the service life of nylon materials according to this application. Step B13 includes steps D11 to D14: Step D11: Based on the initial outer diameter and initial tensile strength reference value of the nylon sample, and combined with the relative expansion failure coefficient and the strength retention rate failure coefficient, calculate and generate the absolute threshold for dimensional failure and the absolute threshold for strength failure corresponding to the initial state of the nylon sample.
[0115] Step D12: Based on the outer diameter data subsequence and intensity data subsequence in the denoised degradation sequence, retrieve and determine the first failure sampling node whose value is greater than or equal to the corresponding absolute threshold for size failure or the absolute threshold for intensity failure.
[0116] Step D13: Based on the time parameters and performance parameters of the failed sampling node and its preceding adjacent nodes, construct a failed cross-node pair.
[0117] Step D14: Based on the assumption that the performance of the nodes in the failure-crossing node pair changes linearly with time, construct a proportional conversion relationship, solve the time coordinates corresponding to when the performance value is equal to the absolute threshold of the size failure or the absolute threshold of the strength failure, and obtain the failure time by linear interpolation.
[0118] The initial outer diameter reference value is the initial outer diameter of the specimen measured before the start of the damp heat aging test and after the completion of the drying pretreatment. It is the reference parameter for calculating the dimensional failure threshold. The initial tensile strength reference value is the initial mechanical property value of the specimen measured before aging degradation. It is the reference parameter for calculating the strength failure threshold. The dimensional failure absolute threshold is the critical outer diameter value for determining whether the specimen has experienced dimensional failure. It is calculated by converting the initial outer diameter and the relative expansion failure coefficient and is the direct basis for failure node retrieval. The strength failure absolute threshold is the critical tensile strength value for determining whether the specimen has experienced mechanical property failure. It is calculated by converting the initial tensile strength and the strength retention rate failure coefficient and is the direct basis for strength dimension failure node retrieval. The failure sampling node is the first discrete measurement time point in the denoised degradation sequence whose value reaches or exceeds the absolute failure threshold. It marks the time position where the failure state is first observed and is the upper boundary node for interpolation calculation. The failure crossing node pair is a node combination consisting of the failure sampling node and its previous adjacent measurement node. The two nodes correspond to the two sides of the failure threshold, respectively, defining the time interval where the failure time occurs.
[0119] In this example, when calculating the absolute failure threshold, it can be performed independently for each sample. The initial outer diameter and initial tensile strength measurements for each sample are read separately, and a unique absolute failure threshold is calculated for each sample using a preset failure coefficient. This eliminates the interference of initial sample dispersion on failure determination. Alternatively, it can be calculated using a unified benchmark under the same working condition. The average initial performance of all samples under the same working condition is taken as the benchmark value, and the absolute failure threshold for that condition is calculated uniformly, thus simplifying the calculation process.
[0120] After calculating the absolute failure threshold, the failure sampling node retrieval process is initiated. For the outer diameter and intensity data subsequences in the denoised degradation sequence, the performance values are compared with the corresponding failure thresholds point by point in ascending time order. The first sampling node whose value is greater than or equal to the corresponding absolute failure threshold is identified as the failure sampling node. The time and performance parameters of this failure sampling node and its preceding adjacent nodes are extracted to construct failure-crossing node pairs. Finally, based on the assumption that the performance between nodes changes linearly with time, a proportional conversion relationship between the performance difference and the time difference is established. The time coordinate corresponding to when the performance value is exactly equal to the absolute failure threshold is solved inversely. The failure time of a single sample is obtained through linear interpolation. In this way, the lifetime accuracy of discrete measurements is improved to a continuous time scale through interval linear interpolation, avoiding lifetime assessment errors caused by rounding down to the nearest whole period.
[0121] For example, there are two methods for calculating the failure time from the denoised degradation sequence. The first is a time-series serial point-by-point search interpolation. Starting from the beginning time of the denoised degradation sequence, the performance values are compared and verified against the failure threshold one by one in chronological order. After each node comparison is completed, if the value does not reach the threshold, the search continues until the first failure sampling node that meets the threshold requirement is located. Then, its preceding adjacent nodes are locked to construct a failure crossing node pair, which is substituted into the linear interpolation formula to calculate the accurate failure time. This method uses the time-series locked point-by-point serial search and interval interpolation calculation logic. By verifying the degradation time sequence sequentially, it ensures that the first failure node is located. The calculation process is stable and controllable, and the results are traceable.
[0122] The second method is multi-sequence batch parallel retrieval interpolation. First, all samples' outer diameter and intensity denoised subsequences are grouped in batches. Simultaneously, parallel failure node retrieval is initiated for multiple groups of sequences. A binary search algorithm is used to locate the failure sampling nodes in each sequence, generating failure-crossing node pairs for all samples in batches. Then, parallel interpolation operations are performed synchronously on all node pairs to solve for the failure time of all samples in one batch, while truncating samples that did not reach the failure threshold within the test period are marked. This method employs the computational logic of batch parallel retrieval and fast location algorithms, improving the efficiency of multi-sample batch processing through parallel computation.
[0123] Furthermore, in the scenario of wet-heat accelerated aging life assessment of nylon 66 precision assembly components, precise failure time calculation is performed for a single test sample under a certain temperature and humidity condition: First, based on the initial outer diameter and initial tensile strength benchmark value of the nylon sample, combined with the preset relative expansion failure coefficient and strength retention rate failure coefficient, the absolute thresholds for dimensional failure and strength failure corresponding to the initial state of the nylon sample are calculated and generated. Then, the denoised degradation sequence obtained by the Pool Adjacent Violators Algorithm (PAVA) is retrieved. According to the outer diameter data subsequence and strength data subsequence in the denoised degradation sequence, the first failure sampling node with a value greater than or equal to the corresponding absolute threshold for dimensional failure or absolute threshold for strength failure is determined point by point in ascending time order. Then, based on the time parameters and performance parameters of the failure sampling node and its preceding adjacent nodes, a failure crossing node pair covering the failure threshold is constructed. Finally, based on the assumption that the performance of nodes between the failure cross-node pairs changes linearly with time, a proportional conversion relationship is constructed. The time coordinates corresponding to when the performance value is equal to the absolute threshold of the size failure or the absolute threshold of the strength failure are solved in reverse. The failure time corresponding to the single sample is obtained by linear interpolation, thereby improving the lifetime assessment accuracy of discrete measurements to a continuous time scale.
[0124] Sixth Embodiment This embodiment provides an exemplary scheme for constructing a dual-path lifetime model through phased decoupling and fitting. In this example, the median characteristic lifetime and the denoised degradation sequence are first matched by grouping according to temperature and humidity dimensions to obtain a lifetime statistical subset and a moisture absorption degradation sequence subset. Then, an empirical statistical lifetime model is constructed step by step through humidity index solution and temperature regression. Simultaneously, mass transfer parameters are fitted for each operating condition and coupled with the performance-moisture absorption relationship to construct a mass transfer mechanism lifetime model. Step S30 includes steps E11~E15: Step E11: Group and match the median characteristic lifetime and the denoised degradation sequence according to the temperature and humidity dimension to determine the lifetime statistical subset with multiple humidity levels at the same temperature and the moisture absorption degradation sequence subset across the entire temperature range.
[0125] Step E12: Calculate the humidity acceleration index based on the lifetime statistics subset, and correct the humidity coupling effect of the median characteristic lifetime using the humidity acceleration index to obtain the logarithmic lifetime.
[0126] Step E13: Perform least squares linear regression on the logarithmic lifetime and the reciprocal of the absolute temperature to obtain the activation energy and the precondition constant, so as to construct the empirical statistical lifetime model.
[0127] Step E14: Based on the aforementioned moisture absorption degradation sequence subset, the moisture diffusion coefficient and equilibrium moisture absorption under each operating condition are obtained by fitting each condition, and the physical parameters of the mass transfer process are determined.
[0128] Step E15: Based on the linear conversion relationship between size and strength and moisture absorption, and coupled with the physical parameters of the mass transfer process, the mass transfer mechanism lifetime model is constructed.
[0129] The lifetime statistics subset is a set of median characteristic lifetime data corresponding to different relative humidity conditions under the same temperature gradient. It is the fundamental data source for calculating the humidity acceleration index and is used to separate the effect of humidity on material lifetime. For example, the median characteristic lifetime data of three conditions at 60 degrees Celsius with relative humidity of 75%, 85%, and 95% constitute a single-temperature lifetime statistics subset. The hygroscopic degradation sequence subset is a set of hygroscopic mass degradation sequences covering the entire temperature gradient. It is the fundamental data source for fitting moisture diffusion kinetics and isothermal adsorption characteristics and is used to extract physical parameters of the mass transfer process. For example, the hygroscopic degradation sequences of water bath conditions and isothermal and constant-humidity conditions at temperatures from 30 to 70 degrees Celsius constitute a full-temperature-range hygroscopic degradation sequence subset. The humidity acceleration index is a power-law exponent characterizing the degree of influence of relative humidity on the material degradation rate. It is a core parameter describing the humidity effect in empirical acceleration models; a larger value indicates a more significant impact of humidity changes on lifetime. Logarithmic lifetime is the median characteristic lifetime logarithmic value after humidity effect correction. It is a standardized dependent variable used in linear regression with the reciprocal of temperature, transforming the exponential relationship of the Arrhenius temperature relation into a linear one for easier parameter solving. Activation energy is a physical parameter characterizing the temperature sensitivity of a material's degradation reaction. It represents the minimum energy required to initiate a degradation reaction, typically measured in kJ / mol, and is a core parameter in empirical statistical models describing the temperature effect.
[0130] The pre-exponential constant is the pre-exponential factor in the empirical accelerated life model, related to the material's intrinsic properties and failure criteria, and is the physical parameter corresponding to the intercept term obtained from linear regression. The empirical statistical lifetime model is a quantitative temperature-humidity-life relationship model based on accelerated test data and obtained through statistical fitting. It is characterized by its simplicity and ease of engineering use, and adopts the Peck Accelerated Life Model form. The moisture diffusion coefficient is a kinetic parameter characterizing the diffusion rate of water molecules within the material, conforming to the Arrhenius temperature relation, and is a core physical quantity describing the rate of mass transfer. The equilibrium moisture absorption is the water content of the material when it reaches saturation steady state under specific temperature and humidity conditions. It is described by the Guggenheim-Anderson-de Boer Isothermal Adsorption Model (GAB) and has a quantitative relationship with temperature and humidity, serving as a boundary condition parameter for the mass transfer process. The physical parameters of the mass transfer process encompass all physical parameters describing the moisture mass transfer process, including the moisture diffusion coefficient, equilibrium moisture absorption, diffusion activation energy, and pre-exponential factor, forming the core parameter basis for constructing the mechanistic model. The performance-moisture absorption linear conversion relationship is a quantitative conversion model that describes the linear correspondence between the outer diameter expansion, tensile strength reduction and moisture absorption of a sample. It serves as a bridge connecting the microscopic mass transfer process and the macroscopic performance degradation.
[0131] In this example, when grouping and matching the original data to obtain two subsets, it can be done by directly indexing and grouping according to the operating condition labels. Based on the temperature and humidity labels of each operating condition, all humidity operating condition lifetime data at the same temperature are filtered and summarized into a lifetime statistics subset. The mass moisture absorption degradation sequences of all operating conditions are extracted and summarized into a moisture absorption degradation sequence subset, thus completing the standardized split of the dataset. Alternatively, it can be done by hierarchical aggregation according to the temperature gradient, aggregating layer by layer from low to high temperature, generating lifetime statistics subsets for corresponding temperatures layer by layer, and simultaneously integrating moisture absorption degradation data according to the temperature dimension to generate a full-temperature-range sequence subset, thus adapting to the step-by-step calling requirements of staged fitting.
[0132] After constructing the two subsets, a phased fitting process using a dual-model approach was initiated. For the empirical statistical model, firstly, based on the same temperature and humidity data from the lifetime statistical subset, a power-law fitting was used to calculate the humidity acceleration index. Then, this index was used to correct for the humidity effect on the median characteristic lifetime for all operating conditions, resulting in a standardized logarithmic lifetime. Subsequently, a least-squares linear regression was performed between the logarithmic lifetime and the reciprocal of the corresponding absolute temperature to obtain the activation energy and precondition constant, thus completing the construction of the empirical statistical lifetime model. For the mass transfer mechanism model, firstly, based on the moisture absorption degradation sequence subset, the moisture diffusion coefficient and equilibrium moisture absorption for each operating condition were fitted using the Fick diffusion model. Then, the Arrhenius relation of the diffusion coefficient and the GAB isothermal adsorption relation of the equilibrium moisture absorption were fitted separately to determine the physical parameters of the entire mass transfer process. Finally, the linear transformation relationship between size, strength and moisture absorption is embedded, and the mass transfer mechanism lifetime model is coupled and constructed by combining the failure threshold. In this way, the phased fitting strategy of temperature and humidity decoupling avoids the ambiguity of the physical meaning of parameters caused by multi-parameter coupling fitting, and improves the parameter accuracy and physical consistency of the two types of models.
[0133] For example, the phased fitting construction method of the dual model includes two types. The first is a serial step-by-step decoupled fitting, which executes the fitting process serially in the order of empirical first, then mechanism, and humidity first, then temperature. First, the lifetime statistical subset of a single temperature is extracted to solve the humidity acceleration index. After fixing the humidity parameters, linear regression across the entire temperature domain is performed to obtain the activation energy and precondition constant, thus completing the identification of the empirical statistical lifetime model. Then, single-condition diffusion fitting, full-temperature-domain kinetic fitting, isothermal adsorption fitting, and performance conversion coupling are performed sequentially to gradually complete the construction of the mass transfer mechanism lifetime model. This method adopts a step-by-step serial calculation and single-factor decoupling logic. The parameter solution of each step is independent and controllable, the physical meaning is clear, and the fitting process is traceable.
[0134] The second approach is a global multi-constraint collaborative fitting method. This method integrates the parameter solving of both the empirical and mechanistic models into a unified optimization framework, aiming to minimize the prediction error of the entire operating condition lifetime. Simultaneously, it applies physical boundary constraints on diffusion activation energy and lifetime activation energy, as well as humidity monotonicity constraints on equilibrium moisture absorption. Through iterative optimization algorithms, the core parameters of both models are adjusted synchronously, achieving optimal prediction consistency and global fitting accuracy while preserving their respective physical meanings. Finally, it synchronously outputs the complete set of parameters for both the empirical statistical lifetime model and the mass transfer mechanism lifetime model. This method employs a global optimization and multi-constraint collaborative computational logic, fully utilizing full-condition data information, avoiding the accumulation of errors from step-by-step fitting, and achieving higher fitting accuracy and model consistency.
[0135] Furthermore, using nylon 66 as the research object, a full-process life prediction was implemented. First, material pretreatment was completed: circular samples and dumbbell-shaped standard tensile specimens were prepared by injection molding. The samples were then placed in an 80-degree vacuum oven for vacuum drying for more than 24 hours. After drying, initial outer diameter and initial mass measurements were performed to standardize the initial moisture content and dimensional benchmarks of the samples. Subsequently, accelerated aging tests were initiated, constructing a multi-gradient temperature and humidity environment including multiple temperature gradients with different relative humidity conditions, as well as a water bath immersion condition with corresponding temperature gradients. Samples were taken out in batches according to a preset time period, and data acquisition for outer diameter measurement, mass measurement, and tensile strength testing was completed sequentially to determine the original degradation time sequence data of the nylon samples. Next, data preprocessing is performed, using the Pool Adjacent Violators Algorithm (PAVA) for order-preserving denoising. This algorithm processes the original degradation time-series data to obtain a denoised degradation sequence, and failure criteria are simultaneously established. The median characteristic lifetime for each operating condition is calculated using linear interpolation based on these criteria. Then, model parameter fitting is performed, using the median characteristic lifetime and the denoised degradation sequence as the fitting basis. An empirical statistical lifetime model and a mass transfer mechanism lifetime model are simultaneously constructed using a staged fitting strategy. The empirical statistical lifetime model adopts the Peck Accelerated Life Model, which describes the quantitative relationship between temperature, humidity, and product characteristic lifetime, as shown in the formula: The physical meanings of each parameter are as follows: L is the median time to reach the failure criterion, i.e., the characteristic lifetime, in hours (h); A is the proportionality coefficient related to the intrinsic properties of the material and the failure criterion, i.e., the precondition constant, without units; RH is the ambient relative humidity, substituted in as a decimal from 0 to 1, without units; n is the humidity acceleration index characterizing the intensity of the influence of humidity on lifetime, the larger n is, the stronger the influence of humidity, without units. The activation energy, which characterizes the temperature sensitivity of the failure reaction and corresponds to the diffusion barrier of water molecules, is expressed in electron volts (eV). Boltzmann constant, a physical constant, has the following values. =8.617×10 -5eV / K, the unit is eV / K; T is absolute temperature, i.e., thermodynamic temperature, converted from Celsius to Kelvin, the unit is Kelvin (K). To avoid the distortion of the physical meaning of parameters due to multicollinearity of temperature and humidity, an empirical statistical lifetime model is constructed using a staged constraint fitting strategy. First, the humidity effect is separated by operating under the same temperature conditions, and then the temperature effect is fitted using full-condition data. Stage one is the derivation of the humidity acceleration index n. Using two sets of operating conditions with the same temperature but different humidity at 60 degrees, the temperature terms completely cancel each other out, and the humidity index n can be directly solved. Let the lifetime at 60 degrees - 85%RH be L1 = 186.07 h, humidity RH1 = 0.85; and the lifetime at 60 degrees - 95%RH be L2 = 143.67 h, humidity RH2 = 0.95. After substituting into the PECK formula and dividing the two equations, A and the temperature terms completely cancel each other out: L2 / L1 = (RH1 / RH2) n Taking the natural logarithm of both sides and rearranging, we get: n = ln(RH2 / RH1) / ln(L1 / L2). Substituting the values, we get: n = ln(0.95 / 0.85) / ln(186.07 / 143.67) = 0.1112 / 0.2586 = 2.3251. Stage two is the activation energy. Following the derivation of the precondition constant A, and fixing n=2.3251, the PECK formula is transformed into a univariate linear regression form. By moving the humidity term to the left side, we obtain... Let: dependent variable y = (Logarithmic lifetime after humidity correction), independent variable x = 1 / T (reciprocal of absolute temperature), intercept C0 = Slope K=E a / The formula simplifies to a standard univariate linear equation: y = C0 + Kx. Least-squares linear fitting is performed on the (x, y) points of the four working conditions to obtain the intercept C0 and slope K. Then, the model parameters are derived: A = exp(C0), E a =K The empirical statistical lifetime model was constructed. The mass transfer mechanism lifetime model is the Fick-GAB-Arrhenius model, which, starting from the microscopic mechanism of water molecule mass transfer and material expansion, constructs a three-layer mechanism framework of "isothermal adsorption-molecular diffusion-volume expansion." Through layer-by-layer transfer, it achieves a quantitative mapping from environmental parameters to size lifetime, possessing clear physical connotations and stronger extrapolation reliability. The specific implementation steps are: based on the 0.5% relative expansion failure criterion, through the size-hygroscopic linear relationship... The water absorption rate was converted to failure rate, and then substituted into the Fick's Law of Diffusion diffusion model to solve for the precise failure time of a single sample. A staged fitting strategy was adopted: first, the influence of humidity on equilibrium moisture absorption was separated using the Guggenheim-Anderson-de Boer Isothermal Adsorption Model (GAB); then, the influence of temperature on diffusion rate was separated using the Arrhenius Temperature Relation to ensure the clear physical meaning of the parameters. The combined effect of temperature and humidity was integrated to establish a lifetime prediction model, calculate the acceleration factor and the characteristic lifetime at room temperature for each operating condition, and back-calculate the water absorption rate at any time point based on the Fick diffusion law, which was then converted into a predicted value for the outer diameter. For thin, circular samples, since the thickness is much smaller than the diameter, moisture mainly diffuses along the thickness direction of the upper and lower surfaces, and the contribution of lateral diffusion is negligible. Therefore, a two-sided one-dimensional Fick diffusion model was used to describe the change of water absorption rate over time, and the formula is as follows: In the formula: To balance the moisture absorption rate, The water diffusion coefficient is... The thickness of the disc, For time. = Known , , By substituting into the equation and finding the root numerically, the precise failure time of a single sample can be obtained. The GAB isotherm adsorption model describes the quantitative relationship between equilibrium moisture absorption rate and relative humidity. This model, based on the BET model, introduces a multilayer adsorption correction coefficient and is applicable to a wide humidity range of 0–95% RH. It can accurately characterize the S-type isotherm adsorption curve of strongly polar polymers such as nylon. GAB isotherm adsorption model (humidity effect separation): Balanced moisture absorption rate The relationship with relative humidity is described by the GAB (Guggenheim-Anderson-de Boer) model, where the monolayer saturation moisture absorption is... Linear variation with temperature: The Arrhenius diffusion coefficient model: The diffusion of water molecules in the amorphous region of polymers is a typical thermally activated process. The diffusion rate changes with temperature according to the Arrhenius equation. The higher the temperature, the more intense the molecular thermal motion, and the easier it is for water molecules to break through the gaps between polymer chain segments, resulting in an exponential increase in the diffusion coefficient. The water diffusion rate is dominated by temperature, and the diffusion coefficient D satisfies the Arrhenius equation: In the formula: Pre-exponential factor, For diffusion activation energy, The gas constant is 8.314 J / (mol). K)), Here, is the absolute temperature. A logarithmic linear regression was performed on the diffusion coefficients fitted at the four temperatures to obtain the temperature effect parameters: Finally, based on the empirical statistical lifetime model and the mass transfer mechanism lifetime model, the temperature and humidity parameters of the target service environment, namely the temperature T and relative humidity RH of the target operating condition, are substituted into the equation. The failure threshold is then solved with respect to time t. Through dual-path collaborative calculation, the dimensional lifetime, strength lifetime, and full-cycle dynamic moisture absorption evolution parameters of the nylon sample are obtained, which is the complete target prediction result.
[0136] Seventh Embodiment This embodiment provides an exemplary scheme for dual-path collaborative nylon lifetime prediction. In this example, the temperature and humidity parameters of the target service environment are first normalized in dimensions and their value ranges are verified to generate a standardized environmental parameter set adapted to both models. Then, the parameter set is input into the empirical statistical lifetime model and the mass transfer mechanism lifetime model respectively for independent solution to obtain the initial value of the mechanism path lifetime and the dynamic moisture absorption evolution parameters. Subsequently, based on the deviation of the operating conditions and the prediction cycle span, the confidence weights are dynamically allocated and weighted to obtain the baseline predicted lifetime. Finally, after correction by physical constraints and applicable boundary verification, the target predicted lifetime is output. Step S40 includes steps F11~F14: Step F11: Perform dimensional normalization and value range verification on the temperature and humidity parameters of the target service environment to generate an environmental parameter set that adapts to the empirical statistical lifetime model and the mass transfer mechanism lifetime model.
[0137] Step F12: Substitute the environmental parameter set into the empirical statistical lifetime model and the mass transfer mechanism lifetime model respectively to calculate independently, and obtain the initial value of the mechanism path lifetime and the dynamic moisture absorption evolution parameters.
[0138] Step F13: Based on the deviation between the target operating condition and the test operating condition, and the time span of the prediction period, assign confidence weights to the initial value of the mechanism path lifetime and the dynamic moisture absorption evolution parameters, and calculate the baseline predicted lifetime using weighted average.
[0139] Step F14: By applying the physical constraints and applicable boundaries of the empirical statistical lifetime model and the mass transfer mechanism lifetime model, the rationality verification and boundary correction of the benchmark predicted lifetime are performed to obtain the target predicted lifetime of the nylon sample.
[0140] The temperature and humidity parameters of the target service environment are the temperature and relative humidity input indicators of the actual application scenario for which the service life is to be predicted, and are the core input conditions for initiating the service life prediction process. Examples include 25 degrees Celsius and 85% RH for a normal temperature service scenario, 70 degrees Celsius and 95% RH for a high-temperature and high-humidity accelerated service scenario, and saturated humidity conditions for a water bath immersion scenario. Dimensional normalization is a preprocessing operation that converts temperature and humidity parameters of different units and magnitudes into a unified standard input format for the model, eliminating differences in parameter dimensions and ensuring the compatibility between the input and the model interface. Value range verification is a verification operation that checks whether the input temperature and humidity parameters are within the experimental condition range fitted by the model and marks the interpolation or extrapolation attributes; it is a preliminary step for determining the prediction confidence level. The environmental parameter set is an input set containing standardized absolute temperature and relative humidity values after normalization and range verification, which can be directly passed to two types of service life models for calculation. The empirical statistical life model is a temperature and humidity-life empirical relationship model based on the Peck Accelerated Life Model and obtained by statistical fitting of experimental data. It has high prediction accuracy and high computational efficiency within the interpolation interval.
[0141] The mass transfer mechanism lifetime model is a physically driven lifetime model constructed by coupling Fick's Law of Diffusion, the Guggenheim-Anderson-de Boer Isothermal Adsorption Model (GAB), and the Arrhenius Temperature Relation. This model offers stronger physical plausibility under extrapolation scenarios. The initial lifetime value of the mechanistic path is the predicted initial lifetime value obtained independently from the mass transfer mechanism lifetime model and is the core input for the dual-path fusion calculation. The dynamic moisture absorption evolution parameters are a set of dynamic predicted data showing the changes in sample moisture absorption rate, dimensional expansion, and strength retention over time throughout the entire cycle, used to comprehensively characterize the evolution of the degradation process. The deviation between the target operating condition and the experimental operating condition is the relative deviation between the target service temperature and humidity parameters and the boundary of the experimental operating condition matrix. This is a quantitative indicator measuring the degree of prediction extrapolation; a larger deviation indicates a higher risk of model extrapolation. The time span of the prediction period is the length of time corresponding to the predicted target failure lifetime, which is another core dimension affecting the confidence of the model. Long-term predictions rely more on the physical consistency of the mechanism model.
[0142] Confidence weights are weighted coefficients assigned to different prediction paths. Higher weights indicate a larger proportion of the corresponding path's result in the fusion result, used to dynamically adapt to different prediction scenarios. The baseline predicted lifetime is the preliminary lifetime prediction value obtained by weighting and fusing the two prediction results according to their confidence weights; it is the direct object of rationality verification. Physical constraints and applicable boundaries are the pre-defined physical reasonable ranges and applicable operating conditions for both types of models, such as non-negative lifetime values, monotonous degradation trends, and temperature and humidity not exceeding the material's physical limits, used to correct unreasonable prediction results.
[0143] In this example, when preprocessing the temperature and humidity parameters of the target service environment to generate an environmental parameter set, a single-point static verification method can be used. The target temperature and humidity values are read all at once, and after dimensional normalization (converting Celsius temperature to Kelvin and percentage humidity to decimals), they are compared item by item with the preset test condition intervals. After marking interpolation or extrapolation attributes, a standardized environmental parameter set is output, thus completing the standardized encapsulation of the input parameters. Alternatively, a multi-time-segment spectral batch processing method can be used. A temperature and humidity cyclic spectrum containing multiple time segments is input, and dimensional normalization and value interval verification are performed segment by segment to generate environmental parameter sets corresponding to each time period in batches.
[0144] After constructing the environmental parameter set, a dual-path independent calculation process is initiated. The environmental parameter set is input into both the empirical statistical lifetime model and the mass transfer mechanism lifetime model, and the two independent solutions are executed simultaneously. The initial lifetime value of the mechanism path and the dynamic moisture absorption evolution parameters throughout the entire cycle are output, while the initial lifetime value of the empirical path is output simultaneously as a fusion reference. Subsequently, based on the deviation between the target operating condition and the experimental operating condition, and combined with the time span of the prediction period, the confidence weights corresponding to the two results are dynamically calculated, and the initial lifetime values are weighted and summed to obtain the baseline predicted lifetime. Finally, the physical constraint rules and applicable boundary conditions corresponding to the two types of models are retrieved to perform a rationality check on the baseline predicted lifetime and evolution parameters, correcting prediction results that exceed the physical boundaries, and finally outputting the target predicted lifetime of the nylon sample. In this way, by using a dual-path collaborative fusion approach, the accuracy advantage of the empirical model and the extrapolation reliability of the mechanism model are taken into account, improving the lifetime prediction accuracy and engineering adaptability in all scenarios.
[0145] For example, there are two ways to achieve the target predicted lifetime through dual-path collaboration. The first is a serial step-by-step fusion prediction, which proceeds sequentially from parameter preprocessing. First, the normalization and interval verification of temperature and humidity parameters are completed to generate an environmental parameter set. Then, the empirical statistical lifetime model and the mass transfer mechanism lifetime model are called sequentially to perform calculations, obtaining the initial values of empirical lifetime, the initial values of mechanism path lifetime, and dynamic moisture absorption evolution parameters, respectively. Subsequently, confidence weights are calculated based on the deviation of the target operating condition and the prediction span, and weighted fusion is performed to obtain the baseline predicted lifetime. Finally, physical constraint verification and boundary correction are performed to output the final target predicted lifetime. This method adopts a computational logic of step-by-step serial execution and single-task closed-loop processing. The calculation process is clear and controllable, and intermediate results can be traced segment by segment.
[0146] The second method is multi-condition batch parallel collaborative prediction. First, parallel normalization and interval verification are performed on multiple sets of target service environment parameters input in batches, generating corresponding environmental parameter sets in batches. Then, dual-path parallel computation is simultaneously initiated for all conditions, obtaining the initial values of mechanistic path lifetime, empirical path lifetime, and dynamic moisture absorption evolution parameters for each set of conditions in batches. Subsequently, weight allocation and weighted fusion are performed in batches to generate the baseline predicted lifetime for each condition. Simultaneously, the rationality verification and boundary correction of all results are completed, and finally, the target predicted lifetime for all conditions is output in batches. This method employs batch parallel processing and end-to-end collaborative computational logic, which can fully utilize computing resources and improve the efficiency of multi-condition prediction.
[0147] Furthermore, the dynamic outer diameter prediction method, the PECK-power function prediction path, extends the prediction from lifetime to full-cycle size prediction by introducing a power function degradation model to describe the change in outer diameter over time. In this model, the water absorption and expansion of nylon follows Fick's diffusion law, with the initial size expansion being proportional to the power of time. This is generalized to a universal power function degradation form: D(t) = D0 + a t b In the formula, D(t): the outer diameter of the disc at time t, in mm; D0: the initial outer diameter, in mm; a: the degradation rate coefficient, strongly correlated with temperature and humidity, with a increasing value at higher temperatures and humidity; b: the degradation time exponent, an intrinsic parameter of the material, which is a constant under the same failure mechanism. The degradation exponent b is statistically determined by fitting the b value of a single sample to the denoised outer diameter sequence of each test sample using logarithmic linearization: ln(D(t)) D0)=lna+b lnt, where lnt is the x-coordinate and ln(D) is the y-coordinate. A linear regression is performed with D0) as the ordinate, and the slope is the b-value for a single sample. The mean of all valid samples is then taken as the intrinsic degradation index of the material. Outer diameter prediction is derived; given arbitrary temperature and humidity (T, RH), the corresponding failure life can be calculated using the PECK model. The corresponding expansion amount at this time is .Will Substituting into the power function degradation formula, we can solve for the degradation rate coefficient under this temperature and humidity: ; Get a and Then, the predicted outer diameter at any time t can be calculated: ; The Fick diffusion-size transition prediction path, based on physical mechanisms, is entirely derived from the diffusion equation and does not rely on failure lifetime to deduce parameters, possessing independent predictive capabilities. The prediction process is as follows: input the target temperature and humidity (T, RH), and the GAB model calculates the equilibrium moisture absorption rate. Calculate the diffusion coefficient D at the corresponding temperature using the Arrhenius equation; substitute it into the Fick diffusion formula to calculate the instantaneous water absorption rate at any time t. Through the size-moisture absorption linear relationship Converted to outer diameter expansion rate, the predicted outer diameter is finally obtained: .
[0148] Eighth embodiment This embodiment provides an exemplary scheme for implementing lifetime prediction using a dual-path independent solution. In this example, a standardized set of environmental parameters is first input into an empirical statistical lifetime model to obtain the initial value of the empirical path lifetime and the power-law degradation rate coefficient, forming the empirical side calculation result. Then, the isothermal adsorption calculation unit and diffusion kinetics calculation unit of the mass transfer mechanism lifetime model are sequentially called to calculate the equilibrium moisture absorption and moisture diffusion coefficient under the target operating condition. These are then substituted into the two-dimensional one-dimensional moisture diffusion equation to numerically solve for the initial value of the mechanism path lifetime. Finally, based on the power-law degradation rate coefficient, a full-cycle degradation extrapolation is completed, generating dynamic moisture absorption evolution parameters. Please refer to... Figure 5 , Figure 5 This is a flowchart illustrating the eighth embodiment of the method for predicting the service life of nylon materials according to this application. Step F12 includes steps G11 to G15: Step G11: Substitute the set of environmental parameters into the empirical statistical lifetime model, solve the lifetime using the pre-identified activation energy and humidity acceleration index, output the initial value of the empirical path lifetime and the power function degradation rate coefficient under the corresponding operating conditions, and form the empirical side calculation results.
[0149] Step G12: Input the temperature and humidity mapping parameters of the environmental parameter set into the isothermal adsorption calculation unit of the mass transfer mechanism lifetime model, and calculate the equilibrium moisture absorption under the target operating condition by combining the monolayer saturated moisture absorption parameters.
[0150] Step G13: Input the absolute temperature parameter from the environmental parameter set into the diffusion kinetics calculation unit of the mass transfer mechanism lifetime model, and perform exponential calculation based on the prefitted diffusion activation energy and pre-exponential factor to obtain the moisture diffusion coefficient at the target temperature.
[0151] Step G14: Based on the equilibrium moisture absorption and the moisture diffusion coefficient as boundary conditions for the mass transfer process, and combining the failure critical moisture absorption and the performance moisture absorption conversion relationship, the two-dimensional moisture diffusion equation is substituted into the equation to obtain the initial value of the mechanism path lifetime.
[0152] Step G15: Based on the degradation rate coefficient of the function, combined with the degradation time exponent and the initial performance benchmark value, perform a full-cycle extrapolation to generate the dynamic moisture absorption evolution parameters.
[0153] The empirical calculation results are the complete set of computational outputs of the empirical statistical lifetime model based on the input operating conditions. They include single-point failure lifetime and full-cycle degradation model parameters, serving as a complete output carrier of the empirical path prediction capability. Examples include initial values for empirical path lifetime, power-law degradation rate coefficients, intermediate humidity correction values, and temperature response coefficients.
[0154] The isothermal adsorption calculation unit is the functional module in the mass transfer mechanism lifetime model responsible for solving the equilibrium moisture absorption. It embeds the Guggenheim-Anderson-de Boer Isothermal Adsorption Model (GAB), and inputting temperature and humidity parameters outputs the corresponding steady-state moisture absorption level. Monolayer saturated moisture absorption parameters are core physical parameters describing the monolayer saturated moisture absorption capacity of a material, and are one of the core fitting parameters of the GAB isothermal adsorption model, exhibiting a linear variation with temperature. Examples include monolayer saturated moisture absorption Mm, adsorption equilibrium constant CG, and multilayer adsorption correction coefficient k. Equilibrium moisture absorption is the moisture absorption rate per unit mass when the sample reaches steady state under specific temperature and humidity conditions. It represents the steady-state boundary condition of the moisture diffusion process and determines the final dimensional expansion and performance degradation limits of the material. The diffusion kinetics calculation unit is the functional module in the mass transfer mechanism lifetime model responsible for solving the moisture diffusion coefficient. It embeds the Arrhenius temperature relation, and inputting temperature parameters outputs the diffusion rate parameters at the corresponding temperature.
[0155] The diffusion activation energy is a physical parameter characterizing the energy threshold required for water molecules to overcome the gaps between polymer chain segments. It is a core indicator determining the temperature sensitivity of the diffusion rate, and its unit is usually J / mol. The smaller the diffusion activation energy, the weaker the influence of temperature on the diffusion rate. The pre-exponential factor is a proportionality constant in the Arrhenius diffusion equation, related to the free volume inside the material and the mobility of molecular chain segments. It is an intrinsic parameter describing the intrinsic diffusion capability of the material. The moisture diffusion coefficient is a core kinetic parameter characterizing the rate of water molecule diffusion inside the material, and its unit is usually mm² / h. A larger value indicates a faster moisture absorption rate and a shorter time for the material to reach moisture absorption equilibrium. The critical moisture absorption capacity for failure is the critical moisture absorption rate corresponding to the dimensional or strength failure criterion. It is calculated from the failure threshold and the performance moisture absorption conversion relationship and is the target boundary condition for finding the root of the diffusion equation. The performance moisture absorption conversion relationship is a linear conversion model describing the quantitative correspondence between moisture absorption and dimensional expansion and strength reduction. It is a bridge connecting the microscopic mass transfer process and macroscopic performance degradation. For example, a 1% increase in moisture absorption corresponds to a 0.3% expansion of the outer diameter and an 8% decrease in tensile strength.
[0156] The two-sided one-dimensional moisture diffusion equation, based on Fick's Law of Diffusion, is the governing equation describing the synchronous moisture absorption and diffusion process on both the upper and lower surfaces of a thin circular sample. It is the core mathematical model for solving the mechanistic lifetime problem. The power-law degradation rate coefficient is a parameter of the power-law model describing the rate of material degradation. It is strongly correlated with temperature and humidity conditions; the higher the temperature and humidity, the larger the coefficient and the faster the degradation rate. The degradation time exponent is an intrinsic parameter describing the time nonlinearity of the degradation process. It is a constant under the same material and the same failure mechanism, with a theoretical value of 0.5 for the Fick diffusion-controlled degradation process. The initial performance baseline values are the initial outer diameter, initial mass, and initial tensile strength measurements of the sample before aging, serving as the starting baseline for full-cycle degradation extrapolation.
[0157] In this example, when initiating the dual-path computation process, it can be performed in a sequential, progressive manner. First, the empirical statistical lifetime model is invoked to complete the empirical calculations. Then, the isothermal adsorption computation unit and the diffusion kinetics computation unit are initiated sequentially to solve for the mechanistic parameters. Finally, lifetime root calculation and evolution parameter generation are performed, thus completing the entire sequential computation under a single operating condition. Alternatively, a dual-path parallel scheduling approach can be used, simultaneously initiating the empirical model computation thread and the mechanistic model computation thread. The two paths independently load model parameters and input data, synchronously completing the empirical result output and mechanistic parameter solution. After both paths have completed their computations, the results are aggregated, thereby improving the overall efficiency of the prediction computation.
[0158] After the input of the environmental parameter set and the startup of the calculation thread are completed, the empirical path solution is executed first. The environmental parameters are substituted into the empirical statistical lifetime model, and the initial value of the empirical path lifetime is directly calculated based on the pre-identified activation energy and humidity acceleration index. Simultaneously, the power function degradation rate coefficient under the corresponding operating condition is obtained by inverse solving, forming a complete empirical calculation result. Simultaneously, the mechanistic path solution is executed. The temperature and humidity mapping parameters in the environmental parameter set are extracted and input into the isothermal adsorption calculation unit. Combined with the monolayer saturated hygroscopic parameters, the equilibrium moisture absorption under the target operating condition is calculated. Then, the absolute temperature parameter in the environmental parameter set is extracted and input into the diffusion kinetics calculation unit. Based on the pre-fitted diffusion activation energy and pre-exponential factor, an exponential operation is performed to obtain the moisture diffusion coefficient at the target temperature. Subsequently, using the equilibrium moisture absorption and moisture diffusion coefficient as boundary conditions for the mass transfer process, the relationship between the critical moisture absorption at failure and the performance moisture absorption conversion is embedded. The result is substituted into the two-dimensional moisture diffusion equation to perform numerical root-finding, obtaining the initial value of the mechanism path lifetime. Finally, based on the power function degradation rate coefficient, combined with the intrinsic degradation time exponent of the material and the initial performance benchmark value, the performance degradation process over the entire time cycle is deduced point by point, generating complete dynamic moisture absorption evolution parameters. In this way, through the dual-path independent solution architecture, the calculation processes of the two types of models are guaranteed not to interfere with each other, which facilitates separate verification and weight allocation, while taking into account the dual needs of macroscopic lifetime prediction and full-cycle process deduction.
[0159] For example, the dual-path independent calculation can be implemented in two ways. The first is a single-threaded serial progressive calculation, which executes the empirical path first and the mechanistic path later. First, the parameters of the empirical statistical lifetime model are called and the lifetime is solved, and the empirical calculation results are output. Then, the calculation process of the mechanistic model is started, sequentially completing the calculation of equilibrium moisture absorption, diffusion coefficient, root finding of the diffusion equation, and derivation of evolution parameters. All steps are performed serially and verified one by one, finally outputting all calculation results. This method uses a single-threaded, progressive calculation logic, which provides strong process controllability. The calculation results of each step can be verified in real time, facilitating the investigation of parameter deviations and calculation errors.
[0160] The second approach employs multi-unit distributed parallel computing. This method breaks down the empirical model solution, isothermal adsorption calculation, diffusion kinetics calculation, diffusion equation root finding, and evolution parameter derivation into multiple independent computational units. Based on a unified set of environmental parameters, all units synchronously initiate parallel computation. Each unit independently calls upon its corresponding model parameters to complete the calculation. After all units have completed their calculations, the initial values of the mechanistic path lifetime and dynamic hygroscopic evolution parameters are aggregated, and the initial values of the empirical path lifetime are simultaneously output. This method utilizes multi-unit distributed parallel scheduling, which fully leverages multi-core computing resources to shorten the overall computation time, resulting in higher computational efficiency.
[0161] Further, please refer to Figure 6 , Figure 7 , Figure 6The following are the fitting plots for the four stages of the model in this application. Figure 7For the typical working condition prediction diagram of this application, the same batch of injection-molded nylon 66 thin circular sheet samples and GB / T 1040.2 standard tensile specimens were used as the objects. The samples were first placed in an 80-degree vacuum oven for continuous drying for more than 24 hours until the mass change between two consecutive weighings was less than 0.01%. After drying, the samples were immediately transferred to a desiccator to cool to room temperature. The initial outer diameter, initial mass and initial tensile strength were measured as soon as possible to eliminate the discrete interference of material crystallinity, internal stress and initial moisture content from the source. Subsequently, a water bath immersion condition covering four temperature gradients of 30°C, 50°C, 60°C, and 70°C was constructed, along with constant temperature and humidity condition matrices at 50°C, 60°C, and 70°C, respectively, with 75%RH, 85%RH, and 95%RH. The dried samples were placed in batches into the corresponding environments to initiate the hygrothermal accelerated aging test. Periodic testing was performed using a sampling cycle with denser sampling at the beginning and sparser sampling at the end. After each sampling, the surface condensation was gently wiped off and the samples were allowed to stand at room temperature. The average measurement of the outer diameter at three points, high-precision mass weighing, and tensile strength testing were completed sequentially. The round samples were immediately returned to the aging environment after measurement to achieve continuous tracking of the same samples. When the mass change rate of three consecutive measurements was less than 0.01% / day, it was determined that moisture absorption equilibrium had been reached. The equilibrium parameters were recorded simultaneously, and finally, the original degradation time series data under all conditions were obtained. In the data preprocessing stage, the Pool Adjacent Violators Algorithm (PAVA) is used to perform order-preserving denoising. This eliminates random fluctuations and abnormal drops in measurements while preserving the non-decreasing physical laws of the degradation process, resulting in a denoised degradation sequence. Then, combining the 0.5% relative expansion failure criterion and the strength retention rate failure criterion, adjacent measurement nodes crossed by the failure threshold are located. The precise failure time of a single sample is solved through linear interpolation. After removing discrete outliers using the Grubbs method, the median of the remaining failure times is taken to obtain the median characteristic lifetime for each operating condition. In the model building stage, the median characteristic lifetime and the denoised degradation sequence are used as the data foundation. Two types of lifetime models are simultaneously constructed using a phased fitting strategy: the empirical statistical side uses the Peck Accelerated Life Model. First, the humidity acceleration index is solved separately using data from multiple humidity conditions at the same temperature. Then, the humidity-corrected logarithmic lifetime and the reciprocal of the absolute temperature are subjected to least-squares linear regression to obtain the activation energy and precondition constant, completing the parameter identification for temperature and humidity decoupling.The mass transfer mechanism is coupled with Fick's Law of Diffusion, the Guggenheim-Anderson-de Boer Isothermal Adsorption Model (GAB), and the Arrhenius Temperature Relation. A linear conversion relationship between performance and moisture absorption is first established. The linear coefficient of moisture absorption-expansion in the size dimension is 0.2865%, and the goodness of fit R² = 0.9556. The linear coefficient of moisture absorption-strength decay in the strength dimension is 11.5189%, both of which are highly consistent with the degradation law of nylon materials. The Fick diffusion equation was then fitted under each working condition, and the equilibrium water absorption rates under water bath conditions of 70°C, 60°C, 50°C, and 30°C were obtained as 2.0614%, 1.9659%, 1.8230%, and 1.2446%, respectively. The corresponding water diffusion coefficients were 0.016524 mm² / h, 0.009798 mm² / h, 0.006025 mm² / h, and 0.003507 mm² / h, respectively. The goodness of fit for each working condition was higher than 0.96. Subsequently, the temperature dependence of the diffusion coefficient was fitted using the Arrhenius equation. The size-based diffusion activation energy was 32.70 kJ / mol with a goodness of fit R² = 0.9572, falling within the typical range for water molecule diffusion activation energy in amorphous polymer regions. The intensity-based diffusion activation energy was 50.42 kJ / mol, reflecting a stronger temperature sensitivity to mechanical property degradation. Simultaneously, the quantitative relationship between equilibrium moisture absorption and humidity was fitted using the GAB model, achieving a goodness of fit R² of 0.9785, accurately describing the S-type isothermal adsorption characteristics brought about by the strongly polar groups of nylon. Finally, a complete mass transfer mechanism lifetime model was obtained by embedding a failure criterion. After model construction, the fitting accuracy was confirmed through leave-one-out cross-validation. Comparing the activation energies of the two models revealed that the apparent activation energy of the Peck model was significantly higher than the diffusion activation energy. This is essentially due to the different rate control steps; the diffusion activation energy only corresponds to the water molecule mass transfer energy barrier, while the apparent activation energy covers the entire tandem process of adsorption, diffusion, hydrogen bond reconstruction, and chain segment relaxation. The high energy barrier of chain segment relaxation dominates the overall temperature sensitivity of failure.Taking a typical service condition of 30°C-95%RH and an initial outer diameter of 31.90mm as an example, the life prediction stage uses dual models and dual-path collaborative calculation to output results: In the short to medium term, within the range of 0~1000h, the deviation of the outer diameter prediction values of the two models is small, and the Peck power function result is slightly higher. In the long-term 10000h scale, the Fick mechanism model tends to be dimensionally stable due to the constraint of equilibrium moisture absorption, which is more in line with the physical law of nylon moisture saturation. Under the 0.5% relative expansion failure criterion, the empirical model predicts a life of about 71 days, and the mechanism model predicts a life of about 145 days. Together, they constitute a life range that combines rapid engineering assessment and reliable physical extrapolation. At the same time, in view of the large dispersion of tensile strength data, a strength life prediction path is independently constructed based on the mechanism model, which effectively solves the problem of insufficient adaptability of the single statistical model and comprehensively improves the accuracy, physical interpretability and engineering scenario adaptability of nylon material wet heat life assessment.
[0162] For example, please refer to Figure 8 , Figure 8This is a schematic diagram of the electron microscopy results of this application. To address the need for dimensional degradation and service life assessment of nylon 66 precision structural components under humid and hot conditions, multi-gradient temperature and humidity accelerated aging tests and dual-model life prediction verification were conducted. The rationality of the prediction method is supported from two dimensions: macroscopic performance degradation laws and microscopic morphology evolution mechanisms. In this embodiment, nylon 66 circular samples and standard tensile specimens from the same batch of injection-molded materials were used as test subjects. After being vacuum-dried at 80 degrees Celsius to unify the initial moisture content, they were placed in constant temperature and humidity chambers and constant temperature water baths under multiple temperature and humidity conditions to initiate accelerated aging tests. Samples were taken in batches according to a preset cycle of denser samples at the beginning and sparser samples at the end. The outer diameter was measured, the mass was weighed, and the tensile strength was tested simultaneously to obtain complete original degradation time series data. At the same time, samples were taken at different moisture absorption stages, such as dry state, water absorption rate of 0.34%, water absorption rate of 0.49%, and water absorption rate of 0.99%. The microscopic morphology of the sample cross-section was observed by scanning electron microscopy (SEM). The observation results showed that as the water absorption increased, the internal pores of the material gradually increased until the pores merged, which ultimately led to a significant decrease in the tensile strength of the sample. This provides microscopic evidence for the mechanism of performance degradation caused by moisture absorption aging. In the data preprocessing stage, the Pool Adjacent Violators Algorithm (PAVA) is used to perform order-preserving denoising, which effectively eliminates measurement noise and numerical fallback anomalies introduced by instrument fluctuations and operational differences, ensuring the non-decreasing physical consistency of degraded data and significantly improving the reliability of subsequent lifetime calculations and model fitting results. After denoising, the median characteristic lifetime under each operating condition is obtained by linear interpolation using a failure criterion of 0.5% relative expansion. In the model building stage, a staged fitting strategy is adopted to effectively decouple the coupling effect of temperature and humidity. A power function empirical prediction path based on the Peck Accelerated Life Model and a mass transfer mechanism prediction path coupled with Fick's Law of Diffusion, the Guggenheim-Anderson-de Boer Isothermal Adsorption Model (GAB), and the Arrhenius Temperature Relation are constructed respectively. The fitting parameters of the two types of models have clear physical meanings and their values are within the reasonable range for polymer materials, forming a complementary prediction system that is suitable for different application scenarios. Among them, the Peck-power function prediction path is easy to calculate and suitable for rapid engineering evaluation scenarios, while the Fick diffusion mechanism path has strong physical constraints and is suitable for long-term working condition extrapolation and in-depth mechanism research scenarios.Cross-validation under experimental conditions showed that both prediction paths could accurately predict the dimensional failure life and full-cycle dynamic outer diameter variation of Nylon 66. Under typical service conditions of 30 degrees Celsius and 95% RH, the characteristic life of the Nylon 66 sample corresponding to the 0.5% relative expansion failure criterion was approximately 75 days, which can provide quantitative data support for the service life design, environmental adaptability assessment, and tolerance margin design of precision structural components.
[0163] This application provides a device for predicting the service life of nylon materials. The device includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method for predicting the service life of nylon materials in the above embodiment 1.
[0164] The following is for reference. Figure 9 The diagram illustrates a structural schematic of a device suitable for predicting the service life of nylon materials in the embodiments of this application. The device for predicting the service life of nylon materials in the embodiments of this application may include, but is not limited to, mobile terminals such as xenon lamp aging test chambers and electro-hydraulic servo fatigue testing machines, as well as fixed terminals such as differential scanning calorimeters, dynamic mechanical analyzers, and thermogravimetric analyzers. Figure 9 The device shown for predicting the lifespan of nylon materials is merely an example and should not be construed as limiting the functionality and scope of the embodiments of this application.
[0165] like Figure 9As shown, the nylon material lifespan prediction device may include a processing unit 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 1002 or a program loaded from a storage device 1003 into a random access memory (RAM) 1004. The RAM 1004 also stores various programs and data required for the operation of the nylon material lifespan prediction device. The processing unit 1001, the read-only memory 1002, and the RAM 1004 are interconnected via a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to I / O interface 1006: input devices 1007 including, for example, touchscreens, touchpads, keyboards, mice, image sensors, microphones, accelerometers, gyroscopes, etc.; output devices 1008 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 1003 including, for example, magnetic tapes, hard disks, etc.; and communication devices 1009. Communication device 1009 allows the nylon material lifespan prediction device to communicate wirelessly or wiredly with other devices to exchange data. Although the figure shows a nylon material lifespan prediction device with various systems, it should be understood that implementation or possession of all the systems shown is not required. More or fewer systems may be implemented alternatively.
[0166] Specifically, according to the embodiments disclosed in this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from storage device 1003, or installed from read-only memory 1002. When the computer program is executed by processing device 1001, it performs the functions defined in the methods of the embodiments disclosed in this application.
[0167] The nylon material service life prediction device provided in this application, employing the nylon material service life prediction method described in the above embodiments, can solve the technical problem of the difficulty in predicting the replacement cycle of nylon materials. Compared with the prior art, the beneficial effects of the nylon material service life prediction device provided in this application are the same as those of the nylon material service life prediction method provided in the above embodiments, and other technical features of the nylon material service life prediction device are the same as those disclosed in the previous embodiment method, and will not be repeated here.
[0168] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.
[0169] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.
[0170] This application provides a computer-readable storage medium having computer-readable program instructions (i.e., a computer program) stored thereon, the computer-readable program instructions being used to execute the method for predicting the service life of nylon materials in the above embodiments.
[0171] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, system, or device. The program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, radio frequency (RF), etc., or any suitable combination thereof.
[0172] The aforementioned computer-readable storage medium may be included in a device for predicting the service life of nylon materials; or it may exist independently and not be assembled into a device for predicting the service life of nylon materials.
[0173] The aforementioned computer-readable storage medium carries one or more programs. When these programs are executed by a nylon material service life prediction device, the nylon material service life prediction device performs the following actions: acquires raw degradation time-series data of nylon samples, wherein the raw degradation time-series data includes the outer diameter, mass, and tensile strength of each group of nylon samples arranged according to time periods under each group of temperature and humidity conditions; processes the raw degradation time-series data using a sequence-preserving denoising algorithm to obtain a denoised degradation sequence, and calculates the median characteristic lifetime corresponding to the operating condition by combining linear interpolation with the failure criterion; uses the median characteristic lifetime and the denoised degradation sequence as the fitting basis, and constructs an empirical statistical lifetime model and a mass transfer mechanism lifetime model through a staged fitting strategy; based on the empirical statistical lifetime model and the mass transfer mechanism lifetime model, substitutes the temperature and humidity parameters of the target service environment, and obtains the target predicted lifetime of the nylon sample through dual-path collaborative calculation.
[0174] Computer program code for performing the operations of this application can be written in one or more programming languages or a combination thereof, including object-oriented programming languages such as Java, Smalltalk, and C++, as well as conventional procedural programming languages such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).
[0175] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation that may be implemented in systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.
[0176] The modules described in the embodiments of this application can be implemented in software or hardware. The names of the modules do not necessarily limit the functionality of the unit itself.
[0177] The readable storage medium provided in this application is a computer-readable storage medium that stores computer-readable program instructions (i.e., a computer program) for executing the above-described method for predicting the service life of nylon materials, thereby solving the technical problem of the difficulty in predicting the replacement cycle of nylon materials. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as the beneficial effects of the method for predicting the service life of nylon materials provided in the above embodiments, and will not be repeated here.
[0178] The above description is only a part of the embodiments of this application and does not limit the patent scope of this application. All equivalent structural transformations made under the technical concept of this application and using the contents of the specification and drawings of this application, or direct / indirect applications in other related technical fields, are included in the patent protection scope of this application.
Claims
1. A method for predicting the service life of nylon materials, characterized in that, The method includes: Obtain the original degradation time series data of nylon samples, wherein the original degradation time series data includes the outer diameter, mass and tensile strength of each group of nylon samples arranged according to the time period under each group of temperature and humidity conditions. The original degraded time series data is processed by a sequence-preserving denoising algorithm to obtain a denoised degraded sequence. Combined with linear interpolation based on the failure criterion, the median characteristic lifetime corresponding to the working condition is calculated. Using the median characteristic lifetime and the denoised degradation sequence as the fitting basis, an empirical statistical lifetime model and a mass transfer mechanism lifetime model are constructed through a staged fitting strategy. Based on the empirical statistical lifetime model and the mass transfer mechanism lifetime model, the target predicted lifetime of the nylon sample is obtained by substituting the temperature and humidity parameters of the target service environment through dual-path collaborative calculation.
2. The method for predicting the service life of nylon materials as described in claim 1, characterized in that, The step of obtaining the original degradation time series data of the nylon samples, wherein the original degradation time series data includes the outer diameter, mass, and tensile strength of each group of nylon samples arranged according to the time period under each group of temperature and humidity conditions, includes: The nylon sample and the standard tensile specimen were placed in a vacuum oven for constant temperature drying. When the mass change rate of two consecutive weighings was lower than the preset mass change rate, the dried sample was obtained. The dried samples were placed in batches in a constant temperature and humidity test chamber with multiple temperature and humidity gradients and a constant temperature water bath with multiple temperature gradients to obtain aged samples through an aging process. The aging samples were obtained in batches according to the time period of denser samples at the beginning and sparser samples at the end, and the samples were wiped and left to stand to obtain the test samples. The outer diameter of the test sample was measured at three points, the mass was weighed with high precision, and the tensile strength was tested. The parameter values at each time point were recorded to obtain the original degradation time series data.
3. The method for predicting the service life of nylon materials as described in claim 1, characterized in that, The steps of processing the original degraded time-series data using an order-preserving denoising algorithm to obtain a denoised degradation sequence, and then calculating the median characteristic lifetime corresponding to the operating condition using linear interpolation based on the failure criterion, include: The original degradation time series data is split into groups according to the test conditions and sample numbers to generate multiple independent degradation subsequences; The degraded subsequence is processed according to the order-preserving denoising algorithm, and adjacent data blocks that do not satisfy monotonicity are merged by the adjacent violation point pooling algorithm to obtain the denoised degraded sequence. Based on the relative expansion failure coefficient and the strength retention rate failure coefficient, the data points before and after the failure threshold are located in the denoising degradation sequence, and the failure time is calculated by linear interpolation. Discrete outliers are removed from the failure times of all samples under the same operating condition. The median of the target failure times is taken as the characteristic lifetime of the operating condition, and the median characteristic lifetime is obtained by summing them up.
4. The method for predicting the service life of nylon materials as described in claim 3, characterized in that, The steps of processing the degraded subsequence according to the order-preserving denoising algorithm and merging adjacent data blocks that do not satisfy monotonicity using the adjacent violation point pooling algorithm to obtain the denoised degraded sequence include: The time-series data points in the degenerate subsequence are initialized as independent data blocks, and the sample size and mean value of the independent data blocks are recorded to construct an initial ordered block set arranged in chronological order. The initial ordered block set is traversed sequentially, and the mean values of adjacent independent data blocks are compared to locate the first monotonic violation adjacent block pair where the mean of the later block is less than the mean of the earlier block. The monotonicity violation adjacent block pairs are merged into a single data block to obtain the merged data block; After replacing the monotonicity violation adjacent block pair based on the merged data block, backtrack one bit and re-verify the monotonicity of the preceding block until the entire sequence block set satisfies the non-decreasing constraint, thus obtaining the global order-preserving block set; The mean of each data block in the global order-preserving block set is mapped one by one back to the data points in the corresponding time interval of the degenerate subsequence to obtain the denoised degenerate sequence.
5. The method for predicting the service life of nylon materials as described in claim 3, characterized in that, The step of locating the data points before and after the failure threshold in the denoised degradation sequence based on the relative expansion failure coefficient and the strength retention rate failure coefficient, and calculating the failure time through linear interpolation includes: Based on the initial outer diameter and initial tensile strength reference value of the nylon sample, and combined with the relative expansion failure coefficient and the strength retention rate failure coefficient, the absolute threshold of dimensional failure and the absolute threshold of strength failure corresponding to the initial state of the nylon sample are calculated and generated. Based on the outer diameter data subsequence and intensity data subsequence in the denoised degradation sequence, the first failure sampling node whose value is greater than or equal to the corresponding absolute threshold for size failure or the absolute threshold for intensity failure is retrieved and determined. Based on the time and performance parameters of the failed sampling node and its preceding adjacent node, a failure crossing node pair is constructed. Based on the assumption that the performance of the nodes between the failure cross-node pairs changes linearly with time, a proportional conversion relationship is constructed. The time coordinates corresponding to when the performance value is equal to the absolute threshold of the size failure or the absolute threshold of the strength failure are solved in reverse. The failure time is obtained by linear interpolation.
6. The method for predicting the service life of nylon materials as described in claim 1, characterized in that, The steps of using the median characteristic lifetime and the denoised degradation sequence as the fitting basis to construct an empirical statistical lifetime model and a mass transfer mechanism lifetime model through a staged fitting strategy include: The median characteristic lifetime and the denoised degradation sequence are grouped and matched according to the temperature and humidity dimension to determine the lifetime statistical subset with multiple humidity at the same temperature and the moisture absorption degradation sequence subset across the entire temperature range. The humidity acceleration index is calculated based on the aforementioned lifetime statistical subset, and the humidity coupling effect of the median characteristic lifetime is corrected by the humidity acceleration index to obtain the logarithmic lifetime. The activation energy and the precondition constant are obtained by performing least squares linear regression on the logarithmic lifetime and the reciprocal of the absolute temperature, so as to construct the empirical statistical lifetime model. Based on the aforementioned moisture absorption degradation sequence subset, the moisture diffusion coefficient and equilibrium moisture absorption under each temperature and humidity condition are obtained by fitting each working condition, and the physical parameters of the mass transfer process are determined. Based on the linear conversion relationship between size and strength and moisture absorption, and coupled with the physical parameters of the mass transfer process, a mass transfer mechanism lifetime model is constructed.
7. The method for predicting the service life of nylon materials as described in claim 1, characterized in that, The step of obtaining the target predicted lifetime of the nylon sample through dual-path collaborative calculation, based on the empirical statistical lifetime model and the mass transfer mechanism lifetime model, by substituting the temperature and humidity parameters of the target service environment, includes: The temperature and humidity parameters of the target service environment are normalized in dimensions and their value ranges are verified to generate an environmental parameter set that is compatible with the empirical statistical lifetime model and the mass transfer mechanism lifetime model. The environmental parameter set is substituted into the empirical statistical lifetime model and the mass transfer mechanism lifetime model respectively for independent calculation to obtain the initial value of the mechanism path lifetime and the dynamic moisture absorption evolution parameters. Based on the degree of deviation between the target operating condition and the test operating condition, as well as the time span of the prediction period, the initial value of the mechanism path lifetime and the dynamic moisture absorption evolution parameter are assigned confidence weights, and the baseline predicted lifetime is obtained by weighted calculation. By applying the physical constraints and applicable boundaries of the empirical statistical lifetime model and the mass transfer mechanism lifetime model, the rationality verification and boundary correction of the baseline predicted lifetime are performed to obtain the target predicted lifetime of the nylon sample.
8. The method for predicting the service life of nylon materials as described in claim 7, characterized in that, The step of substituting the environmental parameter set into the empirical statistical lifetime model and the mass transfer mechanism lifetime model respectively for independent calculation to obtain the initial value of the mechanism path lifetime and the dynamic moisture absorption evolution parameters includes: Substitute the set of environmental parameters into the empirical statistical lifetime model, solve the lifetime using the pre-identified activation energy and humidity acceleration index, output the initial value of empirical path lifetime and the power function degradation rate coefficient under the corresponding working condition, and form the empirical side calculation result. The temperature and humidity mapping parameters of the environmental parameter set are input into the isothermal adsorption calculation unit of the mass transfer mechanism lifetime model, and the equilibrium moisture absorption under the target working condition is obtained by combining the monolayer saturated moisture absorption parameters. The absolute temperature parameter in the environmental parameter set is input into the diffusion kinetics calculation unit of the mass transfer mechanism lifetime model. Based on the prefitted diffusion activation energy and the pre-exponential factor, an exponential operation is performed to obtain the moisture diffusion coefficient at the target temperature. Based on the equilibrium moisture absorption and the moisture diffusion coefficient as boundary conditions of the mass transfer process, the failure critical moisture absorption and the performance moisture absorption conversion relationship are substituted into the two-dimensional moisture diffusion equation for numerical root finding, and the initial value of the mechanism path lifetime is obtained. Based on the degradation rate coefficient of the function, combined with the degradation time exponent and the initial performance benchmark value, a full-cycle extrapolation is performed to generate the dynamic moisture absorption evolution parameters.
9. A device for predicting the service life of nylon materials, characterized in that, The device for predicting the service life of nylon materials includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the method for predicting the service life of nylon materials as described in any one of claims 1 to 8.
10. A storage medium, characterized in that, The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, it implements the steps of the method for predicting the service life of nylon materials as described in any one of claims 1 to 8.
