A dynamic prediction method for the residual life of a cement-based composite material
Patent Information
- Application Number
- CN202611211056.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-11
- Publication Date
- 2026-11-03
AI Technical Summary
现有水泥基复合材料寿命预测方法中，数据时空基准不统一、路径识别与权重耦合、损伤不可逆、模型约束不足及环境因素未引入等问题，导致预测结果不一致且无法自适应调整。
By acquiring multi-source state data, standardizing the data, independently calculating the basic damage amount, identifying the dominant damage path, dynamically updating the weights, applying non-decreasing constraints, constructing a damage evolution model in conjunction with environmental correction coefficients, calibrating the failure threshold using physical failure standards, and determining the remaining lifetime using the inversion model.
It achieves unique identification of damage paths and continuous adjustment of dynamic weights, ensuring monotonically increasing damage state quantities, and effectively introduces environmental factors, avoiding non-physical prediction results and improving the self-consistency and accuracy of lifetime prediction.
Smart Images
Figure CN123030874A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of structural monitoring and life assessment technology, and in particular to a dynamic prediction method for the remaining life of cement-based composite materials. Background Technology
[0002] High-toughness cement-based composite materials possess superior crack resistance, ductility, and energy dissipation capacity, making them widely used in bridges, tunnels, building reinforcement, and other infrastructure. However, during long-term service, these materials are typically affected by factors such as freeze-thaw cycles, chloride corrosion, carbonation, wet-dry cycles, fatigue loads, and traffic loads, leading to gradual deterioration of pore structure, microcrack initiation, crack propagation, and a decline in macroscopic mechanical properties. Existing life prediction methods suffer from the following shortcomings: First, single-index or fixed-weight methods are insufficient to reflect the dynamic changes in the dominant damage mechanisms of materials at different service stages. While multi-source data fusion methods collect information from multiple dimensions such as porosity, cracks, mechanics, and environment, they generally suffer from problems such as the lack of a unified reference system due to time discreteness, outliers skewing the normalization boundary, and inconsistent physical meanings of comprehensive damage quantities due to the failure to distinguish the mapping direction between positive and negative indices. More critically, the mixed calculation of damage contribution values used in path identification and dynamic weights after path switching easily leads to a circular definition between path identification and weight updates. Furthermore, weight switching may cause a local decrease in comprehensive damage quantities, violating the law of irreversible damage accumulation. In addition, the lack of parameter constraints in life evolution models may result in no finite solution for inversion. If the failure threshold is given by the model itself, it falls into a self-verifying circular verification. Although environmental loads are collected, they are not incorporated into the service process correction, ultimately leading to inherent inconsistencies in the prediction results, such as inconsistent time references, logical contradictions in damage accumulation, and untraceable sources of failure criteria. Moreover, they cannot adaptively adjust the prediction reference according to changes in the service environment. Summary of the Invention
[0003] To overcome the above shortcomings, this invention provides a dynamic prediction method for the remaining life of cement-based composite materials, aiming to improve the problems of inconsistent spatiotemporal data benchmarks, path identification and weight coupling, irreversible damage, insufficient model constraints, and lack of introduction of environmental factors in existing methods.
[0004] To achieve the above objectives, the present invention provides the following technical solution:
[0005] A method for dynamically predicting the remaining life of cement-based composite materials, comprising:
[0006] S1. Obtain multi-source state data of cement-based composite materials. The multi-source state data includes at least indicators reflecting pore characteristics, crack characteristics, mechanical property degradation, and environmental load. The multi-source state data is then processed for data normalization to obtain standardized state data.
[0007] S2. Based on the standardized state data, independently calculate the contribution of pore damage, crack damage, and performance degradation, respectively, as the basic damage amount independent of the dynamic hierarchical weight.
[0008] S3. Based on the basic damage quantities and their time-varying characteristics of each dimension, the dominant damage path of the current service stage is determined by path identification rules, and the dynamic hierarchical weights of each dimension are smoothly updated according to the determined path.
[0009] S4. The basic damage quantity is weighted and fused according to the updated dynamic hierarchical weights, and a non-decreasing constraint is applied to the fusion result to obtain a monotonically increasing dynamic damage state quantity.
[0010] S5. Calculate the environmental correction factor based on the environmental load index in the multi-source state data, and convert the calendar time into equivalent service time based on the environmental correction factor.
[0011] S6. Based on the dynamic damage state quantity and equivalent service time, construct a damage evolution model with parameter boundary constraints. Use a physical failure standard independent of the damage evolution model parameter fitting to calibrate the failure threshold. Use the calibrated failure threshold as the termination criterion for damage evolution. Determine the remaining service life through the inversion model and output the warning level.
[0012] Furthermore, in S1, the data normalization process includes:
[0013] Data with different sampling frequencies are converted to a unified sampling time for time alignment.
[0014] Outlier detection and correction are performed on data points within a sliding window;
[0015] The corrected data is normalized according to the positive and negative directions of the indicators to eliminate the dimensional differences between the indicators.
[0016] Furthermore, in S3, the path identification rules specifically include:
[0017] The basic damage amount of each dimension is compared with the corresponding threshold to form a set of candidate damage paths;
[0018] The instantaneous damage path is determined based on the number of candidate paths and the corresponding path selection strategy.
[0019] A time window consistency check is performed on the instantaneous damage path at multiple consecutive sampling times to determine the dominant damage path in the current service phase.
[0020] Furthermore, the path selection strategy specifically includes:
[0021] When there are multiple candidate damage paths, the dominance score of each candidate path is calculated by comprehensively considering the ratio of the basic damage amount to the corresponding threshold, the evolution rate of the basic damage amount in that dimension, and the failure sensitivity of that path, and the path with the highest score is selected as the instantaneous path.
[0022] When the candidate damage path set is empty, the instantaneous path of the previous time step is retained. When there is no historical path at the initial time step, the path with the largest ratio of basic damage amount to the corresponding threshold is selected as the instantaneous path.
[0023] The failure sensitivity is predetermined by the degree of influence of the basic damage perturbation in the calibration sample on the measured failure time, and is independent of the dynamic hierarchical weight.
[0024] Furthermore, the step of smoothly updating the dynamic hierarchical weights of each dimension according to the determined path includes:
[0025] Based on the determined current dominant damage path, the corresponding target weight vector is retrieved from the preset path-weight mapping relationship;
[0026] A smooth transition method is used to adjust the current dynamic level weights toward the target weight vector, and the rate of the smooth transition is controlled by a preset smoothing coefficient.
[0027] Furthermore, the preset smoothing coefficient is predetermined through parameter optimization with the goal of minimizing the weighted sum of lifetime prediction error, path switching oscillation number, and path switching response delay.
[0028] Furthermore, in S4, the step of applying a non-decreasing constraint to the fusion result includes:
[0029] The historical maximum dynamic damage state quantity is used as the lower limit benchmark of the current dynamic damage state quantity. When the original dynamic damage state quantity obtained by weighted fusion is lower than the lower limit benchmark, the lower limit benchmark is maintained as the current dynamic damage state quantity.
[0030] Furthermore, in S5, the step of converting calendar time into equivalent service time based on the environmental correction factor includes:
[0031] Based on the environmental load indices in the multi-source state data, the environmental load indices are weighted and summed and subject to base constraints to obtain the environmental correction coefficients.
[0032] The equivalent service time is obtained by weighting and accumulating the time steps of each sampling interval according to the environmental correction coefficient.
[0033] Furthermore, in S6, the step of adopting a physical failure criterion independent of the damage evolution model parameter fitting includes:
[0034] Using equivalent service time as the independent variable and dynamic damage state quantity as the dependent variable, a monotonically increasing function is established as a damage evolution model, and constraints are imposed on the model parameters of the monotonically increasing function, including at least the constraint that makes the function have a finite inversion solution at the failure threshold.
[0035] Failure criteria are defined based on at least one physical characteristic of critical bearing capacity loss, macroscopic crack penetration, or pore structure collapse, and mapped to a fixed cutoff point in the damage evolution model.
[0036] The present invention has the following beneficial effects:
[0037] 1. In this invention, by decoupling the calculation of basic damage quantity from the dynamic hierarchical weight, the damage contribution of the three dimensions of porosity, cracks and performance is used as the basic input for path identification, and its value is not affected by the adaptive adjustment of weight. At the same time, through a multi-level judgment logic consisting of threshold screening, dominance scoring, fallback judgment and time window consistency verification, the unique identification of damage path under any input state is ensured, and the exponential smoothing method is combined to keep the dynamic hierarchical weight continuously changing when switching paths.
[0038] 2. In this invention, by applying a non-decreasing constraint to the weighted fusion result and using the historical maximum dynamic damage state quantity as the lower limit benchmark, the damage state quantity is always monotonically increasing, which effectively maintains the physical and logical self-consistency of the damage accumulation process and eliminates the interference of the non-physical decline of damage quantity caused by weight switching on the fitting of the subsequent lifetime prediction model.
[0039] 3. In this invention, by applying boundary constraints to the model parameters, the damage evolution curve can be effectively inverted at the failure threshold, avoiding non-physical prediction results of infinite life; the failure threshold is calibrated using a physical failure standard independent of the model parameter fitting, avoiding cyclical verification between the threshold value and the model prediction; at the same time, the equivalent service time is calculated by environmental load, so that the actual deterioration driving effect of environmental factors is incorporated into the life prediction model. Attached Figure Description
[0040] Figure 1 This is a flowchart of a dynamic prediction method for the remaining life of cement-based composite materials proposed in this invention.
[0041] Figure 2 This is a schematic diagram of the basic damage calculation and damage path identification proposed in this invention;
[0042] Figure 3 This is a schematic diagram of the path-weight mapping, exponential smoothing update, and dynamic damage state quantity construction proposed in this invention.
[0043] Figure 4This is a schematic diagram of the environmental load conversion to equivalent service time and exponential damage evolution model proposed in this invention.
[0044] Figure 5 This is a schematic diagram of the failure threshold calibration, failure time inversion, and early warning output proposed in this invention. Detailed Implementation
[0045] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0046] Example 1
[0047] In a first embodiment of the invention, lifetime prediction is involved under freeze-thaw-chloride coupling conditions. The invention provides a dynamic prediction method for the remaining lifetime of cement-based composite materials, such as... Figure 1 As shown, it includes the following steps:
[0048] S1. Obtain multi-source state data of cement-based composite materials. The multi-source state data includes at least indicators reflecting pore characteristics, crack characteristics, mechanical property degradation, and environmental load. The multi-source state data is then processed for data normalization to obtain standardized state data.
[0049] Furthermore, in S1, the data normalization process includes:
[0050] Data with different sampling frequencies are converted to a unified sampling time for time alignment.
[0051] Outlier detection and correction are performed on data points within a sliding window;
[0052] The corrected data is normalized according to the positive and negative directions of the indicators to eliminate the dimensional differences between the indicators.
[0053] Specifically, the entire life cycle of high-toughness cement-based composite materials is monitored, and the acquired data covers microstructure, macroscopic properties, and external factors. In this embodiment, during the laboratory calibration phase, pore structure data is acquired periodically via X-ray computed tomography (CT) scans, supplemented by mercury intrusion porosimetry (MIP) for calibration verification. During the field monitoring phase, ultrasonic wave velocity is used as a proxy indicator, and a pre-established correlation model is used to invert porosity changes. In specific engineering applications, other equivalent methods such as nuclear magnetic resonance (NMR), ultrasonic testing, or calibrated non-destructive testing proxy models can also be used to obtain pore structure data.
[0054] Crack evolution data were used to identify surface cracks using digital image correlation techniques and machine vision algorithms. Indicators included crack density, average crack width, maximum crack width, and crack propagation rate. Simultaneously, acoustic emission sensors were deployed to collect the number of acoustic emission events during the damage process.
[0055] During the laboratory calibration phase, mechanical performance data are obtained by conducting destructive tests at different time intervals on parallel specimens of the same batch to obtain the retention rate of flexural strength and elastic modulus. In field applications, residual stiffness is monitored by local non-destructive testing or by strain gauges installed on the structure.
[0056] Environmental load data is obtained by using temperature and humidity sensors to record real-time environmental conditions, combined with chloride concentration monitors and fatigue loading cycle counters to acquire environmental impact factors.
[0057] Because the sampling frequencies of various sensors are inconsistent—for example, environmental sensors sample once per hour, while crack monitoring samples once every 20 freeze-thaw cycles—this embodiment sets a unified sampling time sequence as follows: For continuous mechanical indices, linear interpolation is used to align the data to a unified time point; for step-type or cumulative environmental records, the previous value retention method is used to fill in the data, ensuring that at each evaluation node, the data of each dimension have a one-to-one correspondence on the time axis.
[0058] To avoid prediction fluctuations caused by measurement equipment malfunctions or pulse interference, this embodiment performs outlier detection within a sliding window. Define the first... Each indicator at time The measured value is Calculate the local mean of this indicator within the current sliding window. and local standard deviation The criteria for identifying outliers are:
[0059] ;
[0060] or:
[0061] ;
[0062] in, The mutation threshold is the theoretical maximum rate of change of a material's physical quantity within adjacent sampling intervals. For example, for the elastic modulus, the mutation threshold within a single cycle is set to 5% of the initial value.
[0063] After an outlier is detected, if the outlier is located in the middle of the valid data sequence, it is replaced by the linear mean of two adjacent valid points; if it is located at the end of the sequence, the calibrated mean or state estimate under the same working conditions is used for smoothing.
[0064] The corrected data undergoes dimensional transformation. Since different indicators have different effects on damage, directional normalization is required. For positively correlated indicators where damage increases with the indicator value, such as porosity and crack width, the calculation formula is:
[0065] ;
[0066] in, and Indicators The normalized lower and upper bounds; This is the dimensionless damage characterization value. The closer its value is to 1, the closer the degree of degradation of the index is to the upper limit of the normalization boundary.
[0067] For inverse indices where the degree of damage increases as the index value decreases, such as strength retention rate and stiffness retention rate, the calculation formula is:
[0068] ;
[0069] This formula ensures the normalized index. All showed a trend of increasing with the severity of the damage.
[0070] In this embodiment, the normalization boundary of each monitoring indicator is predetermined by calibration samples, and the specific values are shown in Table 1.
[0071] Table 1. Indicators and Normalization Boundaries
[0072]
[0073] The normalization boundaries mentioned above are all predetermined by the calibration samples and remain unchanged during the prediction phase. The normalized values of each index are strictly limited to... Within the range, the influence of different physical dimensions on subsequent damage assessment is eliminated.
[0074] By performing normalization after outlier correction, the contamination of the normalization boundary by outlier extreme values is avoided. At the same time, the corresponding normalization direction is adopted for positive and negative indices, so that indices with different physical meanings are comparable in a unified damage evolution direction, providing a reliable data foundation for subsequent calculation of basic damage and identification of damage paths.
[0075] S2. Based on the standardized state data, independently calculate the contribution of pore damage, crack damage, and performance degradation, respectively, as the basic damage amount independent of the dynamic hierarchical weight.
[0076] Specifically, the calculation of the basic damage quantity uses pre-calibrated index layer weights, which are fixed values and do not change due to any external factors. The basic damage quantity is generated only from the standardized state data at the current moment and the aforementioned fixed weights, and its calculation process does not involve any form of dynamic adjustment.
[0077] Before calculating the basic damage amount, fixed index layer weights need to be set for the internal indicators of each damage dimension. These weights are determined by sensitivity analysis of the calibration samples and reflect the contribution of different observed indicators to the evolution of the same damage path. Each group of weights must satisfy the non-negativity constraint and the sum of the weights within the group must equal 1.
[0078] Contribution of pore damage It is used to characterize the degree of degradation of the internal microstructure of a material. The calculation formula is:
[0079] ;
[0080] in, for The amount of pore damage at any given time; For the normalized first Pore structure indicators; This is a fixed weight for the corresponding indicator. This indicator integrates changes in porosity, average pore size, and connectivity, reflecting the degree of internal porosity of the material caused by the loss of hydration products or physical erosion.
[0081] Contribution of crack damage Used to characterize the geometric evolution of macroscopic cracks in materials. Its calculation formula is:
[0082] ;
[0083] in, for The amount of crack damage at any given moment; For the normalized first One crack evolution index; This represents the fixed weight of the corresponding indicator. This indicator quantifies the cumulative effect of cracks in length, width, and spatial distribution density, and is a core parameter for evaluating the degradation of a material's crack resistance.
[0084] Contribution to performance degradation It is used to characterize the degree of loss of macroscopic mechanical properties of a material. The calculation formula is:
[0085] ;
[0086] in, for Performance degradation at any given moment; For the normalized first One mechanical performance index; This represents a fixed weight for the corresponding indicator. This indicator is directly related to the material's load-bearing capacity and deformation modulus, reflecting the final manifestation of damage at the macroscopic performance level.
[0087] In this embodiment, for high-toughness cement-based composite materials under freeze-thaw-chloride coupling environment, the calculation of each basic damage quantity adopts the following fixed weight allocation scheme verified by calibrated samples:
[0088] Pore damage ,in , , These are the normalized porosity increment, average pore size growth rate, and pore connectivity index, respectively.
[0089] Crack damage ,in , , These are the normalized crack density, average crack width, and cumulative acoustic emission energy, respectively.
[0090] Performance degradation ,in , , These are the normalized flexural strength retention rate, elastic modulus retention rate, and residual bearing capacity retention rate, respectively.
[0091] Based on the above calculations, at each sampling time... You will get a set of , , The basic damage vector is calculated from the standardized state data at the current moment and the pre-calibrated fixed weights, and its value does not involve any dynamic adjustment.
[0092] By using pre-calibrated fixed weights to calculate the three damage dimensions independently, the basic damage quantity is decoupled from the dynamic weights at the generation source, ensuring its objectivity and stability as an input for damage path identification.
[0093] S3. Based on the basic damage quantities and their time-varying characteristics of each dimension, the dominant damage path of the current service stage is determined by path identification rules, and the dynamic hierarchical weights of each dimension are smoothly updated according to the determined path.
[0094] Furthermore, such as Figure 2 As shown, in S3, the path identification rules specifically include:
[0095] The basic damage amount of each dimension is compared with the corresponding threshold to form a set of candidate damage paths;
[0096] The instantaneous damage path is determined based on the number of candidate paths and the corresponding path selection strategy.
[0097] A time window consistency check is performed on the instantaneous damage path at multiple consecutive sampling times to determine the dominant damage path in the current service phase.
[0098] Specifically, through multi-level logical judgment, the system ensures accurate identification of the switching of material damage evolution mechanisms under complex service environments. The obtained pore damage quantity... Crack damage amount and performance degradation Each path is compared with a preset judgment threshold. Candidate damage path set. The definition of is:
[0099] ;
[0100] in, For the first Basic damage level; This is the threshold for determining the corresponding path.
[0101] That is, the candidate conditions for pore damage paths are The candidate conditions for crack damage paths are: The candidate conditions for performance degradation paths are In this embodiment, a pore path threshold is set. Crack path threshold Performance path threshold .
[0102] Furthermore, the path selection strategy specifically includes:
[0103] When there are multiple candidate damage paths, the dominance score of each candidate path is calculated by comprehensively considering the ratio of the basic damage amount to the corresponding threshold, the evolution rate of the basic damage amount in that dimension, and the failure sensitivity of that path, and the path with the highest score is selected as the instantaneous path.
[0104] When the candidate damage path set is empty, the instantaneous path of the previous time step is retained. When there is no historical path at the initial time step, the path with the largest ratio of basic damage amount to the corresponding threshold is selected as the instantaneous path.
[0105] The failure sensitivity is predetermined by the degree of influence of the basic damage perturbation in the calibration sample on the measured failure time, and is independent of the dynamic hierarchical weight.
[0106] Specifically, when the candidate set When multiple elements are involved, a dominant score is required. Determine a unique instantaneous path:
[0107] ;
[0108] in, This indicates the percentage of base damage, reflecting the current damage level relative to the trigger threshold. This represents the evolution rate, a normalized slope of the change in basic damage amount within the current time window, reflecting the severity of damage development. The failure sensitivity is represented by physical constants predetermined from the calibration samples, reflecting the contribution weight of this type of damage to the final failure. In this embodiment, the pore path value is taken as... The crack path value is The performance path value is ; , , The values represent the scoring weights, which are 0.25, 0.35, and 0.40 in this embodiment.
[0109] After calculating the scores of all candidate paths, select... The path with the largest value is selected as the instantaneous damage path. If the set is empty, the path from the previous sampling time is used; if the initial time is empty, the path with the largest ratio of basic damage to threshold is selected.
[0110] Furthermore, the step of smoothly updating the dynamic hierarchical weights of each dimension according to the determined path includes:
[0111] Based on the determined current dominant damage path, the corresponding target weight vector is retrieved from the preset path-weight mapping relationship;
[0112] A smooth transition method is used to adjust the current dynamic level weights toward the target weight vector, and the rate of the smooth transition is controlled by a preset smoothing coefficient.
[0113] Furthermore, the preset smoothing coefficient is predetermined through parameter optimization with the goal of minimizing the weighted sum of lifetime prediction error, path switching oscillation number, and path switching response delay.
[0114] Specifically, to filter out frequent path jumps caused by measurement noise, this embodiment uses a sliding window with a length of 5 sampling moments for consistency verification. The frequency of each instantaneous path is counted within the window, and the path with the most occurrences is determined as the dominant damage path in the current service phase; when the occurrence frequencies are the same, the current damage path from the previous sampling moment is retained.
[0115] After determining the dominant damage path, the corresponding target weight vector is retrieved from the preset mapping table, as shown in Table 2.
[0116] Table 2 Path-Weight Mapping Table
[0117]
[0118] To avoid discontinuities in the damage state due to abrupt weight changes during path switching, an exponential smoothing method is used to update the weights of the current dynamic level. :
[0119] ;
[0120] in, This represents the current dynamic level weight component; The dynamic hierarchical weight components are those from the previous sampling time. The target weight component corresponding to the current damage path; The smoothing coefficient is determined through parameter optimization with the objective of minimizing the weighted sum of lifetime prediction error, path switching oscillation count, and path switching response delay, and its value range is [value range missing]. In this embodiment, parameter optimization is used to set... .
[0121] The multi-level judgment logic, which forms a candidate set through threshold comparison, determines the instantaneous path through dominant scoring, and filters noise interference through time window consistency verification, ensures the uniqueness and noise resistance of path recognition under any input state. At the same time, based on the recognition result, the corresponding target weight is called from the preset mapping relationship and transitioned in an exponential smoothing manner, so that the dynamic hierarchical weights maintain a continuous transition when switching paths, avoiding oscillations in the damage state quantity caused by sudden weight changes.
[0122] S4. The basic damage quantity is weighted and fused according to the updated dynamic hierarchical weights, and a non-decreasing constraint is applied to the fusion result to obtain a monotonically increasing dynamic damage state quantity.
[0123] Furthermore, such as Figure 3 As shown, in S4, the step of applying a non-decreasing constraint to the fusion result includes:
[0124] The historical maximum dynamic damage state quantity is used as the lower limit benchmark of the current dynamic damage state quantity. When the original dynamic damage state quantity obtained by weighted fusion is lower than the lower limit benchmark, the lower limit benchmark is maintained as the current dynamic damage state quantity.
[0125] Specifically, multi-dimensional damage indicators are integrated into a unified scalar state variable, ensuring that its evolution process conforms to the physical laws governing irreversible material damage. This is achieved by utilizing the current dynamic hierarchical weights. , , Pore damage Crack damage amount and performance degradation Perform linear weighted fusion. The calculation formula is:
[0126] ;
[0127] in, for The initial dynamic damage state at time t. This variable represents the total material damage level under the influence of the current dominant damage path, considering micropores, macroscopic cracks, and mechanical degradation. Due to dynamic weighting... It will be adjusted in real time according to the path identification results, and this variable can sensitively capture the changes in the main contradictions at different service stages.
[0128] Weight reallocation due to sensor noise, measurement errors, or dominant path switching may lead to changes in the fused weights. A localized decrease in values was observed. The cumulative damage to cement-based composite materials under environmental and load conditions is irreversible. In this embodiment, a non-decreasing constraint is introduced, using the historical maximum value as the current lower limit:
[0129] ;
[0130] in, The initial time is taken as a monotonically increasing dynamic damage state quantity used for lifetime prediction. This constraint ensures the physical and logical consistency of the damage state quantity, meaning that the degree of damage to the material will only be maintained or increased as service progresses, and will not decrease on its own without repair intervention.
[0131] In this embodiment, when the dominant path switches from pore expansion to crack expansion, the actual effect of the non-decreasing constraint is as follows: the original dynamic damage state quantity at the 110th sampling time is 0.452, which remains unchanged after constraint; at the 120th sampling time, due to the path switching, the original dynamic damage state quantity decreases to 0.448, and the non-decreasing constraint maintains it at its historical maximum value of 0.452, activating the constraint mechanism; at the 130th sampling time, the original dynamic damage state quantity recovers to 0.461, and the constrained state quantity subsequently increases to 0.461. The above process shows that the non-decreasing constraint can effectively eliminate the non-physical decline of the damage state quantity caused by the path switching, ensuring that the damage evolution sequence maintains a monotonically increasing physical consistency.
[0132] By embedding non-decreasing constraints into the processing stage of the weighted fusion result, and using the historical maximum dynamic damage state quantity as the lower limit benchmark at the current moment, the fused damage state quantity can maintain the historical maximum level when the original weighted value undergoes a non-physical decrease due to path switching. This ensures that the damage state quantity output to subsequent stages always remains monotonically increasing, effectively maintaining the physical and logical self-consistency of the damage accumulation process within the framework of dynamic weight adaptive adjustment.
[0133] S5. Calculate the environmental correction factor based on the environmental load index in the multi-source state data, and convert the calendar time into equivalent service time based on the environmental correction factor.
[0134] Furthermore, such as Figure 4 As shown, in S5, the step of converting calendar time into equivalent service time based on the environmental correction factor includes:
[0135] Based on the environmental load indices in the multi-source state data, the environmental load indices are weighted and summed and subject to base constraints to obtain the environmental correction coefficients.
[0136] The equivalent service time is obtained by weighting and accumulating the time steps of each sampling interval according to the environmental correction coefficient.
[0137] Specifically, the accelerated effect of external environmental factors such as temperature and humidity fluctuations, corrosive medium concentrations, and cyclic loading on the material degradation rate is quantified, enabling the life prediction model to dynamically respond to changes in the service environment. The environmental correction factor characterizes the damage acceleration rate of the current environmental conditions relative to standard maintenance or reference conditions. Its calculation process is based on a weighted summation of normalized environmental load indices, with a base constraint applied. The calculation formula is as follows:
[0138] ;
[0139] in, To normalize the environmental load index, the first Environmental factors (such as chloride concentration, carbon dioxide concentration, dynamic load amplitude, etc.) in Standardized values for time; The environmental impact coefficient is determined by calibration tests of the material's degradation rate under different single environments, reflecting the sensitivity of specific factors to damage evolution; To set a lower limit for the environmental correction factor and ensure that the environmental correction factor is not lower than the physically reasonable value, in this embodiment, a setting is provided. .
[0140] Because environmental loads fluctuate dynamically over time, the damage evolution of materials is not a linear function of calendar time. This embodiment obtains the equivalent service time by weighted accumulation of the actual step size of each sampling interval. The calculation formula is as follows:
[0141] ;
[0142] in, The sampling interval step size is the time interval between two adjacent monitoring sessions, which can be expressed as calendar days, number of freeze-thaw cycles, or number of fatigue loading cycles. This is the equivalent service time, reflecting the actual service progress of the material after experiencing accumulated environmental stress. If... This indicates that the material is aging faster than expected under the current environment, and the equivalent time will be faster than the calendar time.
[0143] In this embodiment, a specimen under a freeze-thaw-chloride coupling environment is used as an example, and the freeze-thaw cycle influence coefficient is set. Chloride corrosion influence coefficient Humidity fluctuation influence coefficient At a calendar step of 20 cycles, the environmental correction factors for mild (low salt concentration), harsh (high salt concentration), and extreme (low temperature superimposed strong corrosion) environments were 1.15, 1.46, and 1.82, respectively, corresponding to equivalent time increments of 23.0, 29.2, and 36.4 cycles, and cumulative equivalent service times of 23.0, 52.2, and 88.6 cycles, respectively. These results indicate that a larger environmental correction factor leads to a faster progression of the equivalent service time at the same calendar step, reflecting the actual accelerating effect of environmental factors on material damage evolution.
[0144] By constructing environmental correction coefficients and embedding them into the conversion process from calendar time to equivalent service time, the real-time changes of various types of environmental loads are uniformly mapped to scaling factors on the time axis. This enables the time step in the life prediction model to be adaptively adjusted according to the actual degradation driving effect of the service environment, thus establishing a standardized time reference that does not depend on a fixed calendar scale for describing the damage evolution law.
[0145] S6. Based on the dynamic damage state quantity and equivalent service time, construct a damage evolution model with parameter boundary constraints. Use a physical failure standard independent of the damage evolution model parameter fitting to calibrate the failure threshold. Use the calibrated failure threshold as the termination criterion for damage evolution. Determine the remaining service life through the inversion model and output the warning level.
[0146] Furthermore, such as Figure 5 As shown, in S6, the step of adopting a physical failure criterion independent of the damage evolution model parameter fitting includes:
[0147] Using equivalent service time as the independent variable and dynamic damage state quantity as the dependent variable, a monotonically increasing function is established as a damage evolution model, and constraints are imposed on the model parameters of the monotonically increasing function, including at least the constraint that makes the function have a finite inversion solution at the failure threshold.
[0148] Failure criteria are defined based on at least one physical characteristic of critical bearing capacity loss, macroscopic crack penetration, or pore structure collapse, and mapped to a fixed cutoff point in the damage evolution model.
[0149] Specifically, by establishing a mathematical mapping between damage amount and equivalent service time, a quantitative assessment of the future failure time is achieved. To avoid self-verifying cyclical verification of model prediction results, this embodiment adopts a physical failure criterion independent of the damage evolution model.
[0150] During the laboratory calibration phase, a specimen is considered to have failed if it meets any of the following conditions: the residual bearing capacity of the four-point bending decreases to 70% of the initial value, the maximum crack width reaches 0.50 mm, or the residual stiffness decreases to 65% of the initial value.
[0151] The measured failure time when the calibrated sample reaches the above physical failure standard is recorded, and the corresponding dynamic damage state quantity at that time is extracted. Through statistical analysis of multiple samples, the failure threshold of this embodiment is determined. This threshold, serving as the termination criterion for damage evolution model calculations, has clear physical support.
[0152] Based on equivalent service time The independent variable is the dynamic damage state quantity. With [variable name] as the dependent variable, the following exponential damage evolution model is established:
[0153] ;
[0154] in, The starting time for fitting The corresponding dynamic damage state quantity benchmark value; For damage evolution amplitude parameter, representing the theoretical upper limit of the asymptotic increment of material damage evolution; This is a damage evolution rate parameter, reflecting the severity of damage as it grows over time; is the base of the natural logarithm.
[0155] To ensure that the model is physically valid and has a finite inversion solution, the following constraints are applied when using the nonlinear least squares method for parameter fitting: , This ensures that damage increases monotonically as the service life progresses; This ensures that the damage evolution curve eventually crosses the failure threshold, thereby avoiding non-physical predictions of infinite lifespan.
[0156] By setting a failure threshold Substituting the fitted damage evolution model, the equivalent failure time is obtained by inversion. Therefore, the remaining equivalent service life is calculated. If you need to output the remaining lifespan of the calendar time... Then, combined with the estimated value of the future environmental correction coefficient Conversion In this embodiment The average of the environmental correction coefficients over the most recent three sampling periods was used.
[0157] Based on the current dynamic damage state quantity With failure threshold The proportional relationship is used to output the corresponding warning level and handling suggestions, as shown in Table 3.
[0158] Table 3. Correspondence between Warning Levels and Response Recommendations
[0159]
[0160] By calibrating the failure threshold using a physical failure standard independent of the damage evolution model parameter fitting, the determination of the failure criterion is separated from the model fitting process, avoiding the cyclical verification between the threshold value and the model prediction; at the same time, boundary constraints are applied to the parameters of the damage evolution model, so that the model has a finite inversion solution in mathematics, ensuring the computability of the lifetime prediction.
[0161] Example 2
[0162] In a second embodiment of the present invention, lifetime prediction under a carbonization-fatigue coupled environment is involved. This embodiment uses the same prediction method steps S1 to S6 as Embodiment 1, but the differences lie in the specimen conditions, environmental load parameters, composition of environmental correction factors, and failure judgment criteria, as detailed below:
[0163] Beam-type specimens were prepared using high-toughness cement-based composite materials. The carbonization-fatigue coupled test conditions were as follows:
[0164] The carbon dioxide volume fraction was 20%, the relative humidity was 70±5%, and the temperature was 20±2℃; the fatigue loading frequency was 5Hz, the maximum stress was 60% of the static bending strength, and the stress ratio was 0.1; each Condition data is collected once per fatigue cycle. The independent physical failure criterion is: the specimen develops a through crack and the residual stiffness decreases to below 65% of the initial stiffness.
[0165] The environmental correction factor is calculated as follows:
[0166] ;
[0167] in, , and These represent the normalized carbon dioxide concentration, humidity, and fatigue load indices, respectively. The average environmental correction factor in this embodiment is calculated to be 1.38.
[0168] Taking specimen CF-11 as an example, in the first... Prediction was performed during the fatigue cycle, with the dynamic damage state variable D=0.571, and the identified path was the dominant crack propagation path. By substituting the failure threshold into the damage evolution model for inversion, the predicted failure cycle number was obtained. The actual number of failure cycles was [number]. The prediction relative error was 3.56%.
[0169] Example 3
[0170] In the third embodiment of the present invention, on-site monitoring of bridge deck reinforcement layers is involved. This embodiment uses the same prediction method steps S1 to S6 as Embodiment 1, but the differences lie in the monitoring object and working conditions, data acquisition method, environmental load parameters, and failure judgment criteria, as detailed below:
[0171] A high-toughness cement-based composite reinforcement layer for a bridge deck was selected for on-site monitoring. The reinforcement layer was 35 mm thick, and the monitoring area was divided into four sub-segments. Monitoring included: temperature and humidity, traffic load, crack images, acoustic emission events, core drilling pore detection, and elastic modulus estimated using the ultrasonic rebound method. Data acquisition methods differed from laboratory parallel specimen tests; a combination of on-site core drilling and non-destructive testing was used to obtain pore structure and mechanical property data.
[0172] In the 18th month, the pore damage amount was obtained by monitoring sub-segment B-2. Crack damage amount Performance degradation The current path is identified as the dominant crack propagation path. After dynamic weight smoothing and monotonic constraints, the dynamic damage state quantity... .
[0173] By substituting the failure threshold into the damage evolution model for inversion, the model predicted that the failure threshold would be reached in 24.3 months. Subsequent field testing showed that the maximum crack width in this section reached 0.51 mm in 24 months, while the local stiffness retention rate decreased to 64.7%, meeting the independent physical failure criteria. The prediction error was 1.25%, and the actual early warning lead time was 6 months.
[0174] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.
Claims
1. A method for dynamically predicting the remaining life of cement-based composite materials, characterized in that, include: S1. Obtain multi-source state data of cement-based composite materials. The multi-source state data includes at least indicators reflecting pore characteristics, crack characteristics, mechanical property degradation, and environmental load. The multi-source state data is then processed for data normalization to obtain standardized state data. S2. Based on the standardized state data, independently calculate the contribution of pore damage, crack damage, and performance degradation, respectively, as the basic damage amount independent of the dynamic hierarchical weight. S3. Based on the basic damage quantities and their time-varying characteristics of each dimension, the dominant damage path of the current service stage is determined by path identification rules, and the dynamic hierarchical weights of each dimension are smoothly updated according to the determined path. S4. The basic damage quantity is weighted and fused according to the updated dynamic hierarchical weights, and a non-decreasing constraint is applied to the fusion result to obtain a monotonically increasing dynamic damage state quantity. S5. Calculate the environmental correction factor based on the environmental load index in the multi-source state data, and convert the calendar time into equivalent service time based on the environmental correction factor. S6. Based on the dynamic damage state quantity and equivalent service time, construct a damage evolution model with parameter boundary constraints. Use a physical failure standard independent of the damage evolution model parameter fitting to calibrate the failure threshold. Use the calibrated failure threshold as the termination criterion for damage evolution. Determine the remaining service life through the inversion model and output the warning level.
2. The method for dynamically predicting the remaining life of a cement-based composite material according to claim 1, characterized in that, In S1, the data normalization process includes the following steps: Data with different sampling frequencies are converted to a unified sampling time for time alignment. Outlier detection and correction are performed on data points within a sliding window; The corrected data is normalized according to the positive and negative directions of the indicators to eliminate the dimensional differences between the indicators.
3. The method for dynamically predicting the remaining life of a cement-based composite material according to claim 1, characterized in that, In S3, the path identification rules specifically include: The basic damage amount of each dimension is compared with the corresponding threshold to form a set of candidate damage paths; The instantaneous damage path is determined based on the number of candidate paths and the corresponding path selection strategy. A time window consistency check is performed on the instantaneous damage path at multiple consecutive sampling times to determine the dominant damage path in the current service phase.
4. The method for dynamically predicting the remaining life of a cement-based composite material according to claim 3, characterized in that, The path selection strategy specifically includes: When there are multiple candidate damage paths, the dominance score of each candidate path is calculated by comprehensively considering the ratio of the basic damage amount to the corresponding threshold, the evolution rate of the basic damage amount in that dimension, and the failure sensitivity of that path, and the path with the highest score is selected as the instantaneous path. When the candidate damage path set is empty, the instantaneous path of the previous time step is retained. When there is no historical path at the initial time step, the path with the largest ratio of basic damage amount to the corresponding threshold is selected as the instantaneous path. The failure sensitivity is predetermined by the degree of influence of the basic damage perturbation in the calibration sample on the measured failure time, and is independent of the dynamic hierarchical weight.
5. The method for dynamically predicting the remaining life of a cement-based composite material according to claim 3, characterized in that, The step of smoothly updating the dynamic hierarchical weights of each dimension according to the determined path includes: Based on the determined current dominant damage path, the corresponding target weight vector is retrieved from the preset path-weight mapping relationship; A smooth transition method is used to adjust the current dynamic level weights toward the target weight vector, and the rate of the smooth transition is controlled by a preset smoothing coefficient.
6. The method for dynamically predicting the remaining life of a cement-based composite material according to claim 5, characterized in that, The preset smoothing coefficient is predetermined by parameter optimization with the goal of minimizing the weighted sum of lifetime prediction error, path switching oscillation number, and path switching response delay.
7. The method for dynamically predicting the remaining life of a cement-based composite material according to claim 1, characterized in that, In S4, the step of applying a non-decreasing constraint to the fusion result includes: The historical maximum dynamic damage state quantity is used as the lower limit benchmark of the current dynamic damage state quantity. When the original dynamic damage state quantity obtained by weighted fusion is lower than the lower limit benchmark, the lower limit benchmark is maintained as the current dynamic damage state quantity.
8. The method for dynamically predicting the remaining life of a cement-based composite material according to claim 1, characterized in that, In S5, the step of converting calendar time into equivalent service time based on the environmental correction factor includes: Based on the environmental load indices in the multi-source state data, the environmental load indices are weighted and summed and subject to base constraints to obtain the environmental correction coefficients. The equivalent service time is obtained by weighting and accumulating the time steps of each sampling interval according to the environmental correction coefficient.
9. The method for dynamically predicting the remaining life of a cement-based composite material according to claim 1, characterized in that, In S6, the step of adopting a physical failure criterion independent of the damage evolution model parameter fitting includes: Using equivalent service time as the independent variable and dynamic damage state quantity as the dependent variable, a monotonically increasing function is established as a damage evolution model, and constraints are imposed on the model parameters of the monotonically increasing function, including at least the constraint that makes the function have a finite inversion solution at the failure threshold. Failure criteria are defined based on at least one physical characteristic of critical bearing capacity loss, macroscopic crack penetration, or pore structure collapse, and mapped to a fixed cutoff point in the damage evolution model.