Humidifier atomization module reliability evaluation method and system

Abstract

The invention relates to a reliability evaluation method and system for an atomization module of a humidifier in the technical field of new-generation information, and the method comprises the steps: processing an interaction relation among parameters through employing a preset stress coupling model according to an obtained environment parameter set, and obtaining an accelerated aging path of the atomization module under a simulated multi-stress condition; if the preliminary life prediction value is lower than the target reliability evaluation standard, the influence of the interaction is recalculated by adjusting thermal cycle parameters in simulation, and an optimized acceleration test condition is determined; according to the determined acceleration test protocol, processing the fusion of historical data and newly collected data to obtain quantitative indexes for product quality control, and judging the stability of the indexes under the interaction effect; and if the quantitative index shows that the stability is insufficient, protocol parameters are optimized by adopting a gradient boosting decision tree algorithm, enhanced prediction accuracy is obtained, and rapid prediction of the real failure mode is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of new generation information technology, and in particular to a reliability evaluation method and system of a humidifier atomization module. BACKGROUND

[0002] As the core component of modern home environment regulation equipment, the reliability of the humidifier atomization module directly affects the market competitiveness and user experience of the product. The module mainly relies on piezoelectric ceramic atomization sheets to oscillate and atomize water molecules, and faces challenges in complex working environments during long-term use.

[0003] The current life test method generally has the problem of disconnection between the test environment and the actual use scene. Although the traditional standard condition test can provide basic life data, it cannot fully consider the interaction of various stress factors in actual use, resulting in significant differences between the test results and the real failure mode. Although the existing electrical stress acceleration test method can shorten the test period, it often ignores the influence of water quality changes, temperature fluctuations and other key environmental factors, so that the failure mechanism exposed by the test is not comprehensive. The stress borne by the atomization module in actual work has obvious coupling characteristics. Electrical stress can accelerate the fatigue aging of piezoelectric ceramics, and at the same time, water evaporation during work can leave mineral deposits on the surface of the atomization sheet. These deposits will form a hard scale layer under the action of temperature changes, and then change the vibration characteristics and electrical parameters of the atomization sheet. The accumulation of scale not only blocks the micro-pore structure of the atomization sheet, but also accelerates the electrode corrosion under the action of the electric field, forming a mutually promoting failure cycle process. The existing test method is difficult to accurately simulate this multi-stress coupling environment, especially it cannot effectively control the speed and distribution characteristics of scale formation. For example, in actual use, users may use hard tap water, and after hundreds of hours of work, uneven calcium and magnesium deposits will form on the surface of the atomization sheet. These deposits gradually harden in dry-wet cycles, eventually leading to a sharp decline in atomization efficiency. Traditional test methods are often carried out under single stress conditions, and cannot reproduce this gradual composite failure process.

[0004] Therefore, how to establish an acceleration test method that can accurately simulate the synergistic effect of electrical stress, scale deposition and thermal cycling and other multiple factors, and realize the rapid prediction of the real failure mode of the atomization module, has become a key problem to improve the product quality control efficiency and the accuracy of reliability evaluation. SUMMARY

[0005] The present application provides a reliability evaluation method of a humidifier atomization module, mainly comprising:

[0006] The initial distribution characteristics of the parameters under the interaction of multiple stresses are determined by collecting historical running data of the atomization module in actual use scenarios, obtaining an environmental parameter set including an electrical stress level, a scale deposition rate, and a thermal cycle frequency; according to the obtained environmental parameter set, the interaction relationship among the parameters is processed by using a preset stress coupling model to obtain an accelerated aging path of the atomization module under simulated multiple stress conditions; key failure indicators such as the influence degree of scale deposition on the piezoelectric ceramic vibration characteristics are extracted from the obtained accelerated aging path, and it is judged whether these indicators exceed a preset failure threshold to obtain a preliminary life prediction value; if the preliminary life prediction value is lower than a target reliability evaluation standard, the influence of the interaction is recalculated by adjusting the thermal cycle parameters in the simulation to determine optimized accelerated test conditions; the atomization module sample is run in a laboratory environment by using the determined accelerated test conditions, scale deposition distribution and electrical stress change data monitored in real time are obtained, and a dynamic trend of the compound failure process is obtained; according to the obtained dynamic trend, a support vector machine algorithm is applied to analyze the matching degree of the trend and the actual failure mode, and it is judged whether the matching degree reaches a preset threshold to obtain a refined failure prediction model; by using the obtained refined failure prediction model, the influence of thermal cycles under multiple water quality change scenarios is simulated to determine a final accelerated test protocol for batch reliability evaluation of the atomization module; according to the determined accelerated test protocol, the fusion of historical data and newly collected data is processed to obtain quantitative indicators for product quality control, and the stability of these indicators under the interaction is judged; if the quantitative indicators show insufficient stability, a gradient boosting decision tree algorithm is used to optimize the protocol parameters to obtain enhanced prediction accuracy, and rapid prediction of the real failure mode is realized.

[0007] The application provides a reliability evaluation system of a humidifier atomization module, mainly comprising: an environmental parameter acquisition module, configured to acquire a set of environmental parameters including electrical stress level, scale deposition rate and thermal cycle frequency by acquiring historical operation data of the atomization module in an actual use scenario, and determine initial distribution characteristics of the parameters under multi-stress interaction; a stress coupling analysis module, configured to process the interaction relationship among the parameters by using a preset stress coupling model according to the acquired set of environmental parameters, and obtain an accelerated aging path of the atomization module under simulated multi-stress conditions; a failure index evaluation module, configured to extract key failure indexes such as the influence degree of scale deposition on piezoelectric ceramic vibration characteristics from the obtained accelerated aging path, judge whether the indexes exceed a preset failure threshold, and obtain a preliminary life prediction value; a test condition optimization module, configured to, if the preliminary life prediction value is lower than a target reliability evaluation standard, re-calculate the influence of interaction by adjusting the thermal cycle parameters in simulation, and determine optimized accelerated test conditions; a dynamic trend monitoring module, configured to run the atomization module sample in a laboratory environment by using the determined accelerated test conditions, acquire real-time monitored scale deposition distribution and electrical stress change data, and obtain a dynamic trend of a composite failure process; a failure model refining module, configured to, for the obtained dynamic trend, analyze the matching degree of the trend and an actual failure mode by using a support vector machine algorithm, judge whether the matching degree reaches a preset threshold, and obtain a refined failure prediction model; an accelerated test protocol generation module, configured to simulate the influence of thermal cycle under multiple water quality change scenarios by using the obtained refined failure prediction model, determine a final accelerated test protocol, and use the final accelerated test protocol for batch reliability evaluation of the atomization module; a quality control evaluation module, configured to process fusion of historical data and newly acquired data according to the determined accelerated test protocol, obtain quantitative indexes for product quality control, and judge the stability of the indexes under interaction; and a prediction accuracy enhancement module, configured to, if the quantitative indexes show insufficient stability, optimize protocol parameters by using a gradient boosting decision tree algorithm, obtain enhanced prediction accuracy, and realize rapid prediction of a real failure mode.

[0008] The application discloses a reliability evaluation method and system of a humidifier atomization module, and aims at the core business problem of predicting composite failure caused by multiple stress interactions such as electric stress, scale deposition and thermal cycle in the actual use of the atomization module. BRIEF DESCRIPTION OF DRAWINGS

[0009] Fig. 1 A flowchart of the reliability evaluation method of the humidifier atomization module.

[0010] Fig. 2 A schematic diagram of the reliability evaluation method of the humidifier atomization module.

[0011] Fig. 3 A framework schematic diagram of the reliability evaluation system of the humidifier atomization module. DETAILED DESCRIPTION

[0012] For a further understanding of the present application, reference will be made to the following detailed description taken in conjunction with the accompanying drawings. The following detailed description is provided to better understand the present application, and is not intended to limit the scope of the application. In addition, it should be pointed out that only the parts related to the present application are shown in the drawings for the convenience of description.

[0013] As Figs. 1-3 , the reliability evaluation method of the humidifier atomization module can specifically include the following steps:

[0014] In step S101, the historical operation data of the atomization module in the actual use scene is collected to obtain an environmental parameter set containing electric stress level, scale deposition rate and thermal cycle frequency, and the initial distribution characteristics of these parameters under multiple stress interactions are determined.

[0015] The historical operation data and real-time maintenance records are collected by the atomization module in the extended use scenario to obtain an environmental parameter set containing the electrical stress level, the scale deposition rate, the thermal cycle frequency, and the corrosion factor influence, thereby obtaining a parameter expansion set. According to the parameter expansion set, a Gaussian mixture model is adopted, wherein the input is a numerical set of the electrical stress level, the scale deposition rate, the thermal cycle frequency, and the corrosion factor influence, and the joint distribution range of the thermal cycle frequency and the corrosion factor influence is determined by fitting multiple Gaussian distribution components to calculate the interaction influence. For the joint distribution range, a statistical regression method is used to obtain the parameter variation trend and the life prediction index under the multi-stress interaction, wherein the life prediction index is calculated based on the cumulative effect of the variation trend to determine the initial distribution characteristics. If the initial distribution characteristics do not match the life prediction index, the parameter expansion set is adjusted by integrating the maintenance records, and the distribution characteristics containing the cleaning cycle optimization are obtained by updating the weight of the corrosion factor influence.

[0016] For example, in one embodiment, the historical operation data is collected by a sensor network deployed on the atomization module. These sensors include voltage monitors, deposition detectors, and temperature recorders installed on the core components of the module, such as heating elements and water channel. In actual use scenarios, such as daily operation of a household atomization humidifier, sensors record data in real time, and collect data every fixed interval, such as every hour, to form a historical data set. This process ensures that the data covers the complete cycle of the module from startup to shutdown, including normal operation and intermittent periods, thereby providing a reliable basis for subsequent parameter extraction.

[0017] It should be noted that this collection method can adapt to different environments, such as indoor constant temperature or variable temperature conditions, but is always limited within the field of atomization equipment to ensure the pertinence of the data.

[0018] Specifically, when obtaining the environmental parameter set, the electrical stress level is first extracted from the historical data. This refers to the electrical load intensity that the module bears during operation, which is quantified, for example, by calculating the product of the average voltage and current.

[0019] For example,

[0020] In a possible implementation, if the module is used for a medical atomizer, the data analysis will filter out the peak value of the electrical stress during the peak period, and combine the operation time to obtain the level indicator. Next, the scale deposition rate is obtained by monitoring the accumulation speed of calcium and magnesium ions in the water channel. The process involves comparing the difference between the initial cleaning state and the deposition amount after use, and dividing the operation time to obtain the rate value. This extraction ensures that the parameter reflects the actual deposition dynamics rather than static measurement. Further, the thermal cycle frequency is obtained based on the statistics of temperature changes. Thermal cycle refers to the repeated process of heating and cooling of the module, and the frequency is calculated as the number of cycles per unit time, such as the number of temperature fluctuations per minute.

[0021] In one embodiment, for the industrial atomization nozzle scenario, the data records the temperature profile of the heating element, identifies the complete cycle from heating up to cooling down, and accumulates the frequency value. The extraction of this parameter emphasizes the correlation with the actual usage scenario, such as rapid cycles caused by high-frequency use, thereby laying the foundation for multi-stress analysis.

[0022] Preferably, to determine the initial distribution characteristics of these parameters under multi-stress interaction, it is necessary to understand the concept of multi-stress interaction. This refers to the mechanism of mutual influence between the electrical stress level, the scale deposition rate, and the thermal cycle frequency, for example, high electrical stress can accelerate scale deposition, while frequent thermal cycles can amplify the instability of electrical stress. The process begins with constructing a joint dataset of the parameter set, and then applying statistical methods such as correlation analysis to identify interaction patterns. The specific analysis process includes first calculating the univariate distribution of each parameter, such as the normal distribution characteristics of the electrical stress level; then, through cross-validation, the interaction effect is investigated, for example, observing how the scale deposition rate changes when the thermal cycle frequency increases, and quantifying the interaction strength such as the correlation coefficient.

[0023] In one embodiment, in the laboratory simulation scenario of the atomization module, the data is processed in groups: one group simulates a high electrical stress environment, and the other group combines variable frequency thermal cycles to compare the distribution shift, thereby obtaining initial distribution characteristics such as mean, variance, and interaction peak distribution. This detailed analysis reveals the coupling relationship between parameters, such as the tendency of distribution to skewness under high interaction, helping to predict potential module failures.

[0024] For example, in another implementation, for the outdoor use scenario of a portable atomizer, the determination of initial distribution characteristics will incorporate real-time data correction. The process involves filtering out noisy data first, then building a multi-dimensional distribution model to examine the joint probability distribution of parameters under interaction. Such an approach not only covers a single scenario, but also demonstrates versatility under different humidity conditions in the same field, ensuring the robustness of the distribution characteristics.

[0025] It can be understood that the initial distribution characteristics obtained through the above steps can support the performance optimization of the atomization module, such as adjusting the materials during the design phase to reduce the impact of interaction stress. This technical effect is reflected in improving the durability of the module, rather than exaggeration.

[0026] In one possible implementation, if the scenario is extended to commercial atomization equipment, the acquisition of the parameter set can combine cloud storage of historical data to further enhance the accuracy of distribution determination, but always remain within the boundaries of the atomization field.

[0027] Step S102, according to the obtained environmental parameter set, the interaction relationship between the parameters is processed by using a preset stress coupling model, to obtain an accelerated aging path of the atomization module under simulated multi-stress conditions.

[0028] The environment parameter set is obtained, a temperature parameter and a humidity parameter are extracted from the environment parameter set, an interaction relationship between the temperature parameter and the humidity parameter is processed by using a preset stress coupling model through a product interaction formula, the product interaction formula is that a temperature value is multiplied by a humidity value and then divided by a standardization factor, and a temperature humidity interaction matrix is obtained. According to the temperature humidity interaction matrix, a thermal humidity stress distribution of the atomization module under a multi-stress condition is simulated, a thermal humidity stress distribution map is determined through a grid division method. An aging acceleration coefficient is extracted from the thermal humidity stress distribution map, the aging acceleration coefficient is obtained by dividing a stress peak value in the cumulative distribution map by a time interval, and if the aging acceleration coefficient is higher than a preset threshold value, a vibration parameter is integrated into the environment parameter set to obtain an extended parameter set. For the extended parameter set, a comprehensive stress simulation scene is constructed, a cumulative damage value of the atomization module is calculated through a finite element analysis by using an element strain calculation formula, the element strain calculation formula is that a stress value is divided by a modulus value, and a cumulative damage path is obtained. According to the cumulative damage path, a path optimization calculation is fused to determine an accelerated aging path of the atomization module under the multi-stress condition.

[0029] For example, in an embodiment, first, an environment parameter set is obtained, which includes temperature, humidity, vibration intensity, chemical corrosion factor and the like, and is derived from the actual working environment of the atomization module.

[0030] For example, real-time data collection through a sensor network deployed in the electronic atomization device ensures that the parameter set covers multiple stress factors, thereby providing basic input for subsequent simulation. This collection process can use wireless transmission to aggregate data into a central processing unit for efficient preliminary screening and normalization processing. In this way, the parameter set can reflect the real exposure conditions of the atomization module in daily use, supporting accurate input for the model. Further, a preset stress coupling model is used to process the interaction between these parameters.

[0031] It should be noted that the stress coupling model is a simulation framework based on multi-variable interaction, which considers the mutual influence between different parameters, for example, temperature rise may amplify the corrosion effect of humidity.

[0032] Specifically, the model first converts the parameter set into a vector form, and then calculates the interaction intensity through a weighted coupling function, for example, the coupling of the temperature parameter and the humidity parameter can be represented as a function form of their product, thereby quantifying the combined stress. This processing process ensures the comprehensiveness of the simulation and avoids isolated analysis of a single parameter.

[0033] In a possible implementation, the model can also incorporate time series analysis to capture the dynamic interaction of parameters over time. Based on the above processing, an accelerated aging path of the atomization module under the multi-stress condition is obtained.

[0034] Exemplarily, the path is a curve generated by iterative calculation, describing the trajectory of the change of the atomization module from the initial state to the aging threshold, for example, in a high temperature and high humidity environment, the path shows that the aging speed is accelerated. The specific process includes inputting the coupling result into the acceleration factor equation, estimating the aging rate, and then constructing the path graph. This path not only predicts the service life of the module, but also guides the design optimization.

[0035] In an embodiment, for a medical atomizer, the path simulation shows that the coupling of vibration and chemical factors accelerates material fatigue, thereby achieving service life evaluation of a specific function.

[0036] Preferably, in another embodiment, the acquisition of the set of environmental parameters can be extended to include air pressure and radiation factors, enhancing the applicability of the model.

[0037] For example, these additional parameters are collected through laboratory simulation environment, and combined with field data, to further refine the interaction relationship processing. This extension does not change the core model, but improves the robustness of the simulation.

[0038] Specifically, the interaction processing of the stress coupling model can adopt a hierarchical structure, first separating independent stresses, and then layering and coupling effects.

[0039] For example, the basic layer calculates the influence of a single parameter, and the advanced layer integrates interactions such as the synergistic effect of vibration intensity and temperature leading to increased thermal stress. This hierarchical method facilitates debugging and ensures the accuracy of the generated path. Under multiple stress conditions, the model can simulate the accelerated aging path, which is represented as a sequence of state points connected in series, reflecting the process of performance degradation of the module.

[0040] It can be understood that the generation of the accelerated aging path also involves threshold setting, for example, setting a performance indicator such as a decrease in atomization efficiency to 80% for module failure, and then backtracking the path length based on the coupling model. This setting makes the path more practical and can bring more reliable service life prediction results in the design of electronic atomization devices.

[0041] In an embodiment, for a portable atomization module, the parameter set focuses on vibration and temperature fluctuations in a mobile environment, and the model generates a path after processing the interaction, showing that aging is mainly caused by repeated thermal cycles, thereby guiding material selection. Further, the model can integrate a feedback mechanism, for example, adjusting parameter weights based on preliminary path results to optimize the simulation accuracy of the interaction relationship. This mechanism improves the adaptability of the path.

[0042] For example, in an industrial atomization system, path simulation reveals that the coupling of humidity and chemical factors accelerates the corrosion path, thereby supporting the development of maintenance strategies.

[0043] Step S103, extracting the key failure indicators from the obtained accelerated aging path, such as the degree of influence of scale deposition on the vibration characteristics of piezoelectric ceramics, judging whether these indicators exceed the preset failure threshold, and obtaining a preliminary life prediction value.

[0044] By the accelerated aging path, the key failure indicators are extracted, and the degree of influence of scale deposition on the vibration characteristics of the ceramic is obtained. According to the degree of influence, it is judged whether the key failure indicators exceed the preset failure threshold, and the exceeding state is obtained. If the exceeding state exists, a linear regression algorithm is used to determine the preliminary life prediction value, wherein the linear regression algorithm fits the data points by the least squares method to obtain the regression coefficient and the intercept. For the preliminary life prediction value, multi-index fusion data is obtained, and an extended aging simulation path is determined, wherein the multi-index fusion data is obtained from the deposition influence analysis and the vibration characteristic change. Through the extended aging simulation path, the comprehensive influence of the fused indicators on the vibration characteristics is judged, and a refined life prediction value is obtained. By the accelerated aging path, the scale deposition amount and the piezoelectric ceramic vibration data are obtained, and the key failure indicators are obtained. According to the key failure indicators, an index quantization analysis is used, wherein the index quantization analysis determines the scale deposition influence calculation of the scale deposition on the piezoelectric ceramic vibration by measuring the ratio of the scale deposition amount to the vibration frequency deviation. For the deposition influence calculation, the vibration characteristic change is obtained, wherein the vibration characteristic change is obtained by comparing the initial vibration amplitude with the current vibration amplitude, and if the vibration characteristic change exceeds the preset failure threshold, the degree of influence is judged to be abnormal. From the degree of influence evaluation, the path data extraction result is extracted, wherein the path data extraction result is obtained by integrating the time series data in the accelerated aging path to obtain the life prediction value. Through the life prediction value, the prediction value is adjusted, wherein the prediction value is adjusted by calibrating the historical failure data to obtain the preliminary life prediction value.

[0045] For example, in one embodiment, the process of extracting key failure indicators from the obtained accelerated aging path first needs to understand the concept of accelerated aging path. Accelerated aging path refers to accelerating the degradation process of materials or components by simulating extreme environmental conditions such as high temperature, high humidity or continuous vibration, so as to quickly obtain failure data under long-term use.

[0046] Specifically, in the field of hot water equipment, such as piezoelectric ceramic components used in electric water heaters, the path can simulate the speed of scale accumulation through laboratory equipment, and the path data includes time series of degradation curves, such as changes in vibration frequency or intensity attenuation values. By analyzing these paths, extracting key failure indicators becomes a core step. Further, the extraction of key failure indicators focuses on the degree of influence of scale deposition on the vibration characteristics of piezoelectric ceramics. Piezoelectric ceramics are materials that use the piezoelectric effect to generate vibrations and are commonly used in hot water systems for descaling or cleaning functions, and their vibration characteristics include resonance frequency and amplitude. Scale deposition refers to the deposits formed by calcium and magnesium ions in water during heating, which adhere to the ceramic surface, causing increased vibration damping and thus reduced vibration efficiency. The quantification of the degree of influence can be achieved by comparing the vibration parameters before and after aging, such as measuring the ratio of deposit thickness to frequency shift, the specific process involves using vibration sensors to record data points in the path, and then calculating the percentage of shift, such as the initial frequency f0 and the aged frequency f1, and the degree of influence is (f0 - f1) / f0 * 100%. This extraction method ensures the objectivity and repeatability of the indicators, and is applicable in different hot water equipment scenarios, such as household water heaters or industrial boiler systems.

[0047] In one possible implementation, assuming that the accelerated aging path is simulated by a constant temperature water bath, digital signal processing techniques can be used to filter noise when extracting indicators to ensure accuracy.

[0048] It should be noted that this process is not limited to a single indicator, but can also be extended to multiple dimensions, such as combining the deposition rate under the influence of temperature. Based on the extracted indicators, determining whether these indicators exceed the pre-set failure threshold is a subsequent logical step. The pre-set failure threshold is set based on historical data or standard specifications, such as a vibration frequency shift of more than 15% being considered a failure threshold. The judgment process is achieved by comparing the extracted value with the threshold value, and if the degree of influence is greater than the threshold value, it is marked as a potential failure point. This judgment helps to identify risks early and enables preventive intervention in hot water equipment maintenance.

[0049] Preferably, when obtaining the preliminary life prediction value, a linear regression model can be used to fit the data of the accelerated aging path.

[0050] Specifically, the extracted indicators are used as input, and the model calculates the remaining time from the current state to the failure threshold, such as using the formula T = (threshold - current value) / degradation rate, where T is the predicted life value. This method is applicable in different scenarios in the same field, such as water heater tests under different water quality conditions.

[0051] In an embodiment, consider the actual business process: for a piezoelectric ceramic of a household water heater, accelerate the aging path simulation for 1000 hours of use, extract the vibration attenuation index caused by scale deposition as 12%, the threshold value as 15%, judge that it does not exceed, but predict the remaining life as 800 hours. This embodiment shows the practicability of the technical solution, which can realize the improvement of the reliability of the equipment. Further, the prediction value can be combined with other indexes to form a comprehensive evaluation.

[0052] For example, by integrating scale thickness measurement, the prediction accuracy is further refined.

[0053] It can be understood that through the above steps, the whole process forms a closed chain from extraction to prediction, ensuring general application in the field of hot water equipment.

[0054] In an embodiment, when applied in an industrial hot water system, the path data can be obtained through online monitoring, the threshold value is judged in real time after the index is extracted, and the prediction value is used for scheduling maintenance plan. Through these implementation manners, the technical solution covers the complete process from data extraction to decision output.

[0055] In step S104, if the preliminary life prediction value is lower than the target reliability evaluation standard, the heat cycle parameters in the simulation are adjusted to recalculate the influence of interaction and determine the optimized accelerated test conditions.

[0056] By obtaining the preliminary life prediction value and comparing it with the reliability standard, a preliminary evaluation result is obtained by judging whether it is lower than the threshold value. If the preliminary evaluation result is lower than the threshold value, the heat cycle parameters are adjusted to determine the adjusted simulation scenario. From the adjusted simulation scenario, a finite element analysis method is used, which calculates the stress distribution and thermal strain interaction of each element by meshing the simulation scenario into finite elements, to obtain the recalculated influence value. According to the recalculated influence value, an alternative scheme of the accelerated test conditions is obtained, the applicability of the alternative scheme is judged to obtain an optimized option. Through the optimized option, the parameter optimization cycle and the final test scheme are fused to determine the optimized accelerated test conditions.

[0057] For example, in an embodiment, when the preliminary life prediction value is lower than the target reliability evaluation standard, the simulation environment needs to be adjusted to improve the prediction accuracy.

[0058] In particular, the preliminary lifetime prediction is a value obtained from initial simulation tests, for example in electronic component reliability assessment, based on expected lifetime calculated under standard thermal cycling conditions. If this value does not meet a predetermined criterion, such as a reliability threshold of 95% or more, an optimization process is initiated. Further, adjusting thermal cycling parameters in the simulation is a key step. Thermal cycling parameters include temperature variation amplitude, cycle frequency and duration, which simulate thermal stress in real-life usage environments. For example.

[0059] In one possible implementation, if the initial parameters are temperature cycling from -40°C to 85°C, 1 hour per cycle, the test intensity can be strengthened by increasing the cycle frequency or extending the temperature range. This adjustment is based on historical data analysis to ensure that the parameter changes better reflect real-life interactions.

[0060] It should be noted that recalculating the impact of interactions involves a multi-factor analysis. Interactions refer to the mutual influence between different environmental factors such as heat, humidity and mechanical stress. In the calculation process, the adjusted parameter data is first collected, and then a simulation model is used to evaluate the superimposed effects of these factors.

[0061] For example.

[0062] It can be understood that through finite element analysis, the cumulative impact of thermal cycling changes on material fatigue is quantified, avoiding the bias of single-factor evaluation. This process ensures that the prediction value is closer to the actual lifetime.

[0063] Preferably, determining the optimized accelerated test conditions is achieved through iterative calculations.

[0064] In particular, in the testing scenario of electronic devices such as circuit boards, parameters are repeatedly adjusted and recalculated until the lifetime prediction value meets the standard.

[0065] For example, if the initial prediction is 5000 hours, and it is below the target of 8000 hours, the thermal cycling intensity is gradually increased, and the interaction such as the risk of crack propagation caused by the interaction of thermal stress and vibration is re-evaluated, and finally the optimized conditions such as temperature cycling from -50°C to 100°C, frequency every 30 minutes are determined.

[0066] In one embodiment, the method is applied to the reliability testing of semiconductor chips. After the preliminary prediction falls below the standard, the thermal cycling parameters are adjusted, such as extending the high-temperature holding time, the interaction between thermal expansion and electromigration is recalculated, and the test conditions are optimized to accelerate the appearance of failure modes, thereby shortening the overall test period.

[0067] For example, in similar fields such as battery assembly testing, the process is similar: if the predicted lifespan is insufficient for the target, adjust the parameters to recalculate the interaction of humidity and heat, and ensure that the optimized conditions improve the accuracy of the evaluation. Further, this optimization method can achieve more reliable life evaluation, reducing the risk of product failure in actual application. In another embodiment, for display screen modules, adjust the calculation of the impact of interactions such as light-heat effects, and determine the conditions to support batch production verification.

[0068] Step S105, using the determined accelerated test conditions, running the atomization module sample in a laboratory environment, obtaining real-time monitored scale deposition distribution and electrical stress change data, and obtaining the dynamic trend of the composite failure process.

[0069] Using the preset accelerated test conditions, running the atomization module sample in a laboratory environment, obtaining real-time monitored scale deposition distribution data from the sample running, obtaining the scale deposition distribution change. Through the scale deposition distribution change, combined with the simultaneously monitored electrical stress change data, determine the preliminary correlation of the composite failure process, and obtain the composite failure process correlation data. According to the composite failure process correlation data, obtain extended thermal stress distribution data from sample running monitoring as a new unique service, and collect the thermal stress distribution data through a thermal stress sensor, and determine whether the thermal stress distribution data exceeds a preset threshold value, then fuse the correlation data to obtain the fused failure data. For the fused failure data, obtain the time series change curve to obtain the dynamic trend of the composite failure process.

[0070] For example, in one embodiment, the accelerated test conditions are determined by simulating high-intensity use environments.

[0071] For example, the laboratory temperature is controlled at 40 to 60 degrees Celsius, the humidity is set to 80% or more, and hard water is introduced as the atomization medium to accelerate the scale formation process. The design of such conditions is based on the extreme situations that the atomization module may encounter in actual application, ensuring that the test can reproduce long-term failure phenomena in a relatively short period of time.

[0072] Specifically, the atomization module sample is placed in a sealed test chamber, and the continuous running time is set to 72 hours or more. The interaction of scale deposition and electrical stress is intensified by controlling variables such as water hardness and operating voltage, thereby obtaining a reliable data basis. Further, when running the atomization module sample in a laboratory environment, a precise control system is used to maintain stable test parameters.

[0073] For example, the sample is installed on a dedicated support, connected to a power supply and water supply device, and the voltage is adjusted in real time during operation to simulate different load conditions. This running mode helps to observe how scale gradually accumulates on the surface of the atomization element and affects the electric field distribution.

[0074] It is noted that the atomization module generally refers to the core component used in household humidifiers or industrial spraying equipment, and its failure often results from the combined effects of scale blockage and electrical stress overload. Through laboratory operation, the changing process can be systematically recorded.

[0075] Exemplarily, the acquisition of real-time monitored scale deposition distribution data can be achieved through an integrated high-resolution imaging system. The specific process includes capturing images of the surface of the atomization element every 5 minutes using an optical sensor, and then analyzing the proportion and thickness distribution of the scale covered area through image processing algorithms. This monitoring method ensures the continuity of the data.

[0076] For example, at the beginning of the test, scale is mainly deposited on the edge of the element, and over time, the distribution expands to the central area, forming an uneven layered structure. This data helps to reveal the dynamic pattern of deposition and provides quantitative basis for subsequent failure analysis.

[0077] In one possible implementation, the acquisition of electrical stress change data relies on embedded voltage and current sensors that measure the electrical field strength and resistance changes inside the atomization module in real time.

[0078] For example, the sensor array is arranged at key positions of the module, and data is collected once every second to record the rise in electrical stress peaks caused by scale accumulation. This change data reflects the process of electrical stability decline of the module, such as the gradual transition from initial uniform distribution of electrical stress to local high stress area, thereby accelerating the fatigue failure of the material. Through this monitoring, detailed time series data curves can be obtained to show the evolution of electrical stress from stability to fluctuation.

[0079] Preferably, based on the above-mentioned acquired data, the dynamic trend of the combined failure process involves data fusion and trend analysis steps.

[0080] Specifically, the scale deposition distribution data and the electrical stress change data are synchronously correlated, and a combined curve is drawn using time series analysis methods.

[0081] For example, a trend of sharp increase in electrical stress when the scale thickness exceeds a certain threshold is observed. This analysis reveals the dynamic mechanism of failure, such as the scale acting as an insulating layer to change the electric field distribution, leading to local overheating and module malfunction. The acquisition of this trend can provide a reference for optimizing the module design, for example, adjusting the scale resistance of the material to delay the occurrence of failure. Further, in another embodiment, the accelerated test conditions can be adjusted to introduce periodic voltage fluctuations to simulate the unstable environment of the actual power grid. When running the sample of the atomization module in this variant, the monitoring data shows that the scale deposition distribution tends to be more irregular, and the electrical stress change presents a pulsed peak. Through this scenario, the dynamic trend analysis shows that the composite failure process is accelerated by about 30%, highlighting the influence of environmental factors.

[0082] It can be understood that the above-mentioned embodiments cover various scenarios in the field of household atomization equipment, such as tests under different water quality conditions, to ensure the universality of the technical solutions.

[0083] For example, when running the sample in a soft water environment, the scale deposition is slower, but the electrical stress change can still be monitored by extending the test time, obtaining a similar dynamic trend curve.

[0084] In an embodiment, the dynamic trend of the composite failure process is presented through a visualization tool, such as generating a three-dimensional chart to show the evolution of scale distribution and electrical stress over time. This presentation helps engineers identify key failure nodes and improve the maintenance strategy of the module accordingly, achieving a longer service life.

[0085] Step S106, for the obtained dynamic trend, a support vector machine algorithm is applied to analyze the matching degree of the trend and the actual failure mode, to determine whether the matching degree reaches a preset threshold, and to obtain a refined failure prediction model.

[0086] A dynamic trend sequence is obtained from historical failure data, and a classification hyperplane is constructed through a support vector machine algorithm. The dynamic trend sequence is mapped to a high-dimensional space as input data, and a key feature vector in the dynamic trend sequence is extracted. For the key feature vector, an actual failure mode is identified, and a cosine similarity is calculated to obtain the vector angle between the key feature vector and the actual failure mode, to obtain a matching degree score of the key feature vector and the actual failure mode. It is determined whether the matching degree score reaches a preset threshold. If the preset threshold is reached, a preliminary failure prediction model is determined. A validation data set of the preliminary failure prediction model is obtained, and the weight parameters and bias parameters of the preliminary failure prediction model are adjusted through gradient descent iteration optimization. According to the adjusted parameters, a refined failure prediction model is obtained.

[0087] Exemplarily, in one embodiment, for the obtained dynamic trend, it is first necessary to understand the source and meaning of the dynamic trend. The dynamic trend is usually extracted from equipment monitoring data, for example, in the field of mechanical equipment failure prediction, the trend curve represented by time series data is formed by the vibration signal or temperature change sequence collected by the sensor. These trends reflect the evolution of the equipment operating state, for example, the gradual increase in vibration amplitude may indicate bearing wear. Through data smoothing processing, such as the moving average method, a more stable dynamic trend can be obtained, providing a reliable basis for subsequent analysis. Such processing helps to filter noise and ensure that the trend accurately captures potential failure signals of the equipment. Further, a support vector machine algorithm is applied to analyze the matching degree of the trend and the actual failure mode. The support vector machine is a supervised learning algorithm used for classification tasks, and its principle is to find a hyperplane to maximize the separation of data points of different categories. In this embodiment, the dynamic trend is taken as the input feature vector, and the actual failure mode is taken as the known label, such as predefined modes such as bearing fracture or gear loosening. The algorithm first constructs a training data set including historical trend data and corresponding failure labels, and then calculates the similarity between the trend vector and the failure mode vector by optimizing the support vector.

[0088] Specifically, the matching degree can be quantified by cosine similarity or Euclidean distance, for example, the angle between the direction of the trend vector and the typical vector of the failure mode is calculated, and if the angle is less than a certain value, it indicates a higher match. Through a kernel function such as a radial basis function, nonlinear trend data can be processed to improve the accuracy of the analysis. This method is particularly suitable for mechanical equipment monitoring and can effectively distinguish between normal trends and abnormal failure signs.

[0089] Preferably, the process of judging whether the matching degree reaches the preset threshold value involves the setting and comparison of the threshold value. The preset threshold value can be determined according to experience or cross-validation, for example, set to 0.8, indicating that the matching degree needs to exceed 80% to be considered valid.

[0090] Exemplarily, if the calculated matching degree is 0.85, it is judged that the threshold value is reached; otherwise, if it is 0.6, it does not meet the standard and needs to be further adjusted.

[0091] In one possible implementation, this judgment can be iteratively performed, and if the threshold value is not reached, the extraction parameters of the dynamic trend can be fed back to optimize, such as increasing the data sampling frequency, so as to improve the matching accuracy. Such threshold value judgment ensures the robustness of the model, and in actual mechanical fault prediction, it can reduce false positives and improve system reliability.

[0092] It should be noted that through the above judgment, a refined failure prediction model is obtained.

[0093] Specifically, if the matching degree reaches the threshold value, the current support vector machine model is considered as a refined version, and the new trend data is integrated for parameter updating, such as adjusting the weight of the support vector to better fit the actual failure mode. Otherwise, go back to the previous step and re-analyze.

[0094] In an embodiment, for blade failure prediction of a wind power generation device, the dynamic trend may include stress distribution under wind speed change, and the SVM analyzes its matching with the known crack pattern. If the threshold value 0.9 is reached, the model is refined into a more accurate prediction tool, which can early warn potential failure and prolong the service life of the device.

[0095] For example, in another embodiment, for failure prediction of a pump type device, the dynamic trend is obtained from flow and pressure data, and the support vector machine algorithm analyzes its matching degree with the jamming pattern. The judgment threshold is set to 0.75, and if the matching is reached, the refined model further improves the generalization ability by increasing more historical failure samples. This diversified scenario demonstrates the universality of the technical solution in the same mechanical prediction field. Further, the technical effect of the refining process is to improve the accuracy of the prediction model, for example, in actual application, the misjudgment rate can be reduced from 15% to 5%, thereby providing more reliable guidance for equipment maintenance.

[0096] In an embodiment, considering the dynamic trend under noise interference, a denoising step such as wavelet transform can be added before applying the support vector machine to purify the data. Then analyze the matching degree to ensure the stability of the threshold judgment.

[0097] Specifically, the refined model can be deployed in a real-time monitoring system, continuously inputting new trend data for dynamic updating.

[0098] It can be understood that this method is applicable to various mechanical subsystems, but is always limited within the field of equipment failure prediction.

[0099] Exemplarily, the threshold value can be adjusted according to the type of the device, for example, the threshold value of high-voltage equipment is higher to ensure safety.

[0100] Step S107, by obtaining the refined failure prediction model, simulating the influence of thermal cycling under various water quality change scenarios to determine the final accelerated test protocol for batch reliability evaluation of the atomization module.

[0101] By obtaining the refined failure prediction model, the model input is historical failure data and water quality parameters, and the output is predicted failure probability. From the water quality change scenario, thermal cycle parameters such as cycle number and temperature range are obtained to determine the thermal cycle impact trend, which is obtained by comparing the correlation between parameter changes and failure probability. According to the thermal cycle impact trend, the module response data under multiple water quality change scenarios is simulated, the simulation uses the Monte Carlo method to generate random water quality variation, and the response data set is obtained, which contains module durability and response time indicators. Using the response data set, factors such as temperature fluctuations and water quality components are integrated, the integration is calculated by weighted average to determine the contribution of each factor to the response, and the acceleration test condition is determined. If the total contribution is higher than the threshold, it is considered as high risk, and the condition threshold is obtained. If the condition threshold exceeds the preset threshold, adjust the test protocol parameters, the adjustment is based on the threshold deviation ratio to modify the test period and intensity, determine the final acceleration test protocol, which includes specific thermal cycle sequence. For the final acceleration test protocol, it is applied in the batch reliability evaluation of the atomization module, which is applied by executing the protocol through batch samples, verifying the batch verification result, which reflects the overall reliability of the module in the simulated scenario.

[0102] Exemplarily, in an embodiment, the reliability of the atomization module is evaluated by obtaining the refined failure prediction model. First, it is necessary to understand the refinement process of the failure prediction model. The model is built based on historical test data and machine learning algorithms, and after refinement, it can more accurately predict the failure probability of the atomization module under different environments.

[0103] Specifically, the refinement process includes data cleaning, feature selection and model iteration, for example, by removing noise data and optimizing parameters, the sensitivity of the model to thermal cycles and water quality variables is improved. In this way, the model can simulate the failure risk in actual use scenarios. Further, simulating the impact of thermal cycles under multiple water quality change scenarios is a key step. Water quality change scenarios refer to the composition of different water solutions, such as pH value ranging from 5 to 9, and mineral content gradient from low to high, these variables will affect the material corrosion and heat conduction efficiency of the atomization module. Thermal cycle impact involves the process of repeated temperature rise and fall, for example, from room temperature to 80 degrees Celsius high temperature cycle, simulating the thermal stress of the atomization module in continuous work.

[0104] Exemplarily, in the simulation, the model inputs water quality parameters and thermal cycle periods, such as setting 10 cycles, each cycle lasts for 30 minutes of heating and cooling stages, and calculates the material fatigue value. Through this simulation, the model outputs the failure probability distribution, which helps to identify high-risk scenarios.

[0105] Preferably, the final accelerated test protocol is determined based on the simulation results. The accelerated test protocol refers to a scheme that shortens the test time but is equivalent to long-term use, for example, increasing the thermal cycle frequency to 2 times the normal use, while adjusting the water quality variables to accelerate the failure to appear.

[0106] Specifically, the protocol development process includes analyzing simulation data, selecting the most representative scenarios, such as rapid thermal cycling under high mineral water quality, and then defining test parameters: test duration of 100 hours, temperature fluctuation amplitude of 50 degrees Celsius, and sample quantity of 50 atomization modules. This protocol is used for batch reliability evaluation to ensure the durability of the modules in actual atomization applications.

[0107] In one possible implementation, different implementation scenarios under different atomization module types are considered to demonstrate the generality of the technical solution.

[0108] For example, in a ceramic core atomization module, the simulation focuses on the impact of water quality on the clogging of the core pores, and the model calculates the expansion coefficient change caused by thermal cycling, and then adjusts the water quality injection amount in the protocol to 5 milliliters per cycle. In a metal core module, the simulation emphasizes the corrosion rate, and the protocol accordingly increases the anti-corrosion coating test step. These scenarios are limited to the field of electronic atomization equipment to ensure consistency.

[0109] It should be noted that the refinement details of the failure prediction model are further expanded. The model uses a supervised learning framework, such as a random forest algorithm, with input features including water quality ion concentration and thermal cycle number, and output as failure time prediction. During refinement, the model hyperparameters are optimized through cross-validation, such as adjusting the tree depth to 10 to reduce overfitting. In this way, the model can handle multivariate interactions in the simulation, such as accelerated degradation of module seals by thermal cycling when pH increases, and the calculation process involves a cumulative damage model, where the damage accumulation value is obtained by iterative addition, and a threshold value of 0.8 is considered high risk.

[0110] For example, in batch reliability evaluation, the protocol can effectively identify potential defects. In specific implementation, a batch of atomization modules are first subjected to accelerated testing, and failure data are recorded and fed back to the model for further refinement. Through this cycle, the evaluation accuracy is improved to ensure stable operation of the modules in a variable water quality environment.

[0111] It can be understood that the technical effect of this method is to provide a reliable prediction and testing framework to avoid unexpected failures in actual deployment.

[0112] In one embodiment, for high humidity water quality scenarios, the simulation adds a humidity variable, and the protocol accordingly extends the cooling phase to simulate the real condensation effect.

[0113] Specifically, the whole process logic is coherent from model refinement to simulation to protocol determination, forming a closed-loop evaluation system.

[0114] Step S108, according to the determined accelerated test protocol, the fusion of historical data and newly collected data is processed to obtain quantitative indicators for product quality control, and the stability of these indicators under interaction is judged.

[0115] Through the accelerated test protocol, the initial set of historical data and newly collected data is obtained from the historical data fusion to obtain the fusion data set. According to the fusion data set, a linear regression algorithm is used, where the input is a pair of variables in the fusion data set, and the output is a fitting coefficient. The quantitative indicator is calculated by the fitting coefficient to determine the quantitative indicator value of the product quality control. Through the quantitative indicator value, the interaction is analyzed, where the interaction analysis evaluates the influence between variables through the product term to obtain the indicator change under the interaction, and the indicator interaction stability is judged. If the indicator interaction stability is lower than the preset threshold, the defect detection process is obtained from the reliability prediction mechanism, where the defect detection process obtains the defect distribution characteristics by scanning the abnormal points of the fusion data set. According to the defect distribution characteristics, the indicator threshold setting and stability verification standard are used, where the stability verification standard evaluates the fluctuation through variance test to judge the stability of the defect distribution characteristics under the interaction.

[0116] Step S109, if the quantitative indicator shows insufficient stability, the gradient boosting decision tree algorithm is used to optimize the protocol parameters to obtain enhanced prediction accuracy, and fast prediction of the real failure mode is realized.

[0117] Obtain the quantitative indicator, and judge the insufficient stability from the quantitative indicator. Use the gradient boosting decision tree algorithm to optimize the protocol parameters from the insufficient stability. Through the protocol parameters, enhanced prediction accuracy is obtained. From the prediction accuracy, fast prediction of the real failure mode is realized.

[0118] For example, in one embodiment, when the quantitative indicator shows insufficient stability, the system first evaluates the performance of the current model. The quantitative indicator can be understood as a series of numerical measures, such as variance or bias indicators of model output, for judging the robustness of the prediction result.

[0119] In particular, the lack of stability can be manifested by a prediction error rate exceeding a pre-set threshold, such as triggering the optimization process when the error rate is higher than five percent in the industrial equipment maintenance scenario. By collecting historical data, such as temperature, vibration, and load parameters from the equipment operation logs, the mean and standard deviation of these indicators are calculated to determine whether further intervention is needed. This evaluation step ensures the relevance of the subsequent optimization, which is applicable in various scenarios in the same field, for example, for different types of mechanical equipment such as pumps or compressors, similar indicator calculations can be used to monitor stability. Further, a gradient boosting decision tree algorithm is employed to optimize the protocol parameters. Here, the gradient boosting decision tree algorithm is an ensemble learning method that builds decision trees through multiple iterations, with each tree fitted to the residuals of the previous tree to minimize the overall loss function. In the industrial predictive maintenance field, this algorithm is particularly suitable for handling data with non-linear relationships, such as time-series data collected by equipment sensors. The protocol parameters refer to adjustable variables in the model, such as learning rate, tree depth, and subsampling ratio, which directly affect the accuracy and generalization ability of the prediction. The optimization process begins with initializing a base decision tree, then calculating the gradient, which is the difference between the current model's predicted values and the true values, and building the next tree based on this. By repeating this process, usually hundreds of times, until the loss function converges. This method can handle complex data patterns in business, such as integrating multi-source data such as vibration signals and current fluctuations to improve the model's sensitivity to potential failures in equipment failure prediction, for example.

[0120] In one possible implementation, for failure mode prediction of pump equipment, the system collects real-time sensor data, including speed, pressure, and temperature readings. First, calculate quantitative indicators such as root mean square error, if the indicator shows insufficient stability, i.e. the error exceeds the threshold, start the gradient boosting decision tree algorithm. The specific steps of the algorithm include dividing the data into training and validation sets, then setting initial parameters such as learning rate to zero point one and maximum tree depth to five. In each iteration, calculate the pseudo-residuals of the current ensemble model, and fit a new regression tree to these residuals. By adjusting the protocol parameters, such as reducing the learning rate to avoid overfitting, gradually enhance the prediction accuracy of the model. In this scenario, the optimized model can quickly identify early signs of pump bearing wear, enabling prediction of the true failure mode, with business effects reflected in reduced equipment downtime and maintenance costs.

[0121] Preferably, during the optimization of the protocol parameters, regularization techniques can be introduced to further enhance stability. The gradient boosting decision tree algorithm allows the addition of L1 or L2 regularization terms to control the complexity of the tree and avoid overfitting in industrial environments with more noisy data.

[0122] For example, in a compressor maintenance scenario, when a quantitative indicator such as accuracy falls below ninety percent, the algorithm optimizes parameters by adjusting subsampling ratios, such as eighty percent of data for training per tree. This adjustment is based on gradient calculations, ensuring that each step moves towards the direction of minimum loss. The business process involves data preprocessing, such as normalizing sensor readings, followed by iterative tree model construction, and finally outputs an optimized parameter set for subsequent prediction tasks.

[0123] It is important to note that the enhanced prediction accuracy is achieved by comparing model performance before and after optimization. After the algorithm iterations end, the system evaluates accuracy using cross-validation methods, such as calculating F1 score or AUC value, which are commonly used in the industry to measure the reliability of failure predictions. Enhanced accuracy means that the model's recognition rate of true failure patterns has improved, for example, from seventy-five percent to ninety-five percent. This improvement supports quick decision-making in business, such as in a device monitoring system, issuing timely alerts to prevent sudden failures.

[0124] In one embodiment, for failure prediction of a generator set, when the quantitative indicator evaluation shows insufficient stability, the gradient boosting decision tree algorithm is used. The specific process includes preparing the data set, such as historical failure records and operating parameters, and then initializing the algorithm parameters. The algorithm optimizes the loss function through gradient descent, adding a tree at each step to correct errors. In this embodiment, the protocol parameter optimization focuses on adjusting the number of trees, for example, setting it to five hundred trees, and monitoring performance on the validation set. The results show that the optimized model can quickly predict true failure patterns such as turbine blade fatigue, and the business effects include improving system reliability and extending equipment life without introducing additional hardware costs.

[0125] Specifically, the principle of the gradient boosting decision tree algorithm lies in its boosting mechanism, that is, weak learners are gradually combined into strong learners. In industrial predictive maintenance, this is suitable for handling high-dimensional data, such as multi-sensor fusion scenarios. When optimizing protocol parameters, the system calculates the gradient vector of each iteration and splits the decision tree nodes accordingly, selecting features such as maximum information gain for splitting. This detailed process ensures the robustness of the algorithm, providing a reliable enhancement path in the case of insufficient stability. Further, the fast prediction of true failure patterns is achieved by deploying the optimized model. The model output includes failure probability and type, such as mechanical fracture or electrical short circuit, which is integrated into the monitoring platform in the business process. Fast prediction means that the response time is shortened to seconds, suitable for real-time industrial environments.

[0126] For example, in another embodiment, for the maintenance of a conveyor belt system, when the quantified indicators such as recall rate are insufficient, the algorithm optimizes parameters by adjusting the maximum number of leaf nodes. The process involves data cleaning, feature selection such as extracting time domain features, and then iterative training. After optimization, the model accurately predicts the failure of belt wear, enhancing the overall production efficiency. In a preferred embodiment, multiple sets of data sources are combined for optimization, and the algorithm outputs enhanced accuracy after processing, supporting a general prediction framework for various industrial equipment.

[0127] The application provides a reliability evaluation system of a humidifier atomization module, mainly comprising: an environmental parameter acquisition module, used for acquiring an environmental parameter set containing an electrical stress level, a scale deposition rate and a thermal cycle frequency by acquiring historical operation data of the atomization module in an actual use scenario, and determining initial distribution characteristics of the parameters under multi-stress interaction; a stress coupling analysis module, used for processing the interaction relationship among the parameters by using a preset stress coupling model according to the acquired environmental parameter set, and obtaining an accelerated aging path of the atomization module under simulated multi-stress conditions; a failure index evaluation module, used for extracting key failure indexes such as the influence degree of scale deposition on piezoelectric ceramic vibration characteristics from the obtained accelerated aging path, judging whether the indexes exceed a preset failure threshold, and obtaining a preliminary life prediction value; a test condition optimization module, used for re-calculating the influence of interaction and determining optimized accelerated test conditions by adjusting the thermal cycle parameters in simulation if the preliminary life prediction value is lower than a target reliability evaluation standard; a dynamic trend monitoring module, used for running the atomization module sample in a laboratory environment by using the determined accelerated test conditions, acquiring real-time monitored scale deposition distribution and electrical stress change data, and obtaining a dynamic trend of a composite failure process; a failure model refining module, used for applying a support vector machine algorithm to analyze the matching degree of the trend and an actual failure mode according to the obtained dynamic trend, judging whether the matching degree reaches a preset threshold, and obtaining a refined failure prediction model; an accelerated test protocol generation module, used for simulating the influence of thermal cycles under multiple water quality change scenarios by using the obtained refined failure prediction model, determining a final accelerated test protocol, and using the final accelerated test protocol for batch reliability evaluation of the atomization module; a quality control evaluation module, used for processing the fusion of historical data and newly acquired data according to the determined accelerated test protocol, obtaining quantitative indexes for product quality control, and judging the stability of the indexes under interaction; and a prediction accuracy enhancement module, used for optimizing protocol parameters by using a gradient boosting decision tree algorithm if the quantitative indexes show insufficient stability, obtaining enhanced prediction accuracy, and realizing rapid prediction of a real failure mode. The above embodiment is only one of the preferred embodiments of the application and should not be used to limit the protection scope of the application, but any modification or polishing without substantial meaning made on the basis of the main design idea and spirit of the application, which still solves the same technical problem as the application, should be included in the protection scope of the application.

Claims

1. A method of reliability assessment of a humidifier atomizing module, characterized in that, The method comprises the following steps: Step S101, by collecting historical running data of the atomization module in the actual use scene, obtaining an environment parameter set containing electric stress level, scale deposition rate and thermal cycle frequency, and determining the initial distribution characteristics of these parameters under the interaction of multiple stresses; Step S102, according to the obtained environment parameter set, using a preset stress coupling model to process the interaction relationship among these parameters, obtaining an accelerated aging path of the atomization module under the simulated multiple stress conditions; Step S103, extracting key failure indicators from the obtained accelerated aging path, such as the influence degree of scale deposition on the piezoelectric ceramic vibration characteristics, judging whether these indicators exceed the preset failure threshold, and obtaining a preliminary life prediction value; Step S104, if the preliminary life prediction value is lower than the target reliability evaluation standard, adjusting the thermal cycle parameters in the simulation, recalculating the influence of the interaction, and determining the optimized accelerated test conditions; Step S105, using the determined accelerated test conditions, running the atomization module sample in the laboratory environment, obtaining the real-time monitored scale deposition distribution and electric stress change data, and obtaining the dynamic trend of the composite failure process; Step S106, for the obtained dynamic trend, applying a support vector machine algorithm to analyze the matching degree of the trend and the actual failure mode, judging whether the matching degree reaches a preset threshold, and obtaining a refined failure prediction model; Step S107, by using the obtained refined failure prediction model, simulating the influence of thermal cycle under multiple water quality change scenes, determining the final accelerated test protocol, and using it for batch reliability evaluation of the atomization module; Step S108, according to the determined accelerated test protocol, processing the fusion of historical data and newly collected data, obtaining quantitative indicators for product quality control, and judging the stability of these indicators under the interaction; Step S109, if the quantitative indicators show insufficient stability, using a gradient boosting decision tree algorithm to optimize the protocol parameters, obtaining enhanced prediction accuracy, and realizing rapid prediction of the real failure mode.

2. The reliability evaluation method of a humidifier atomizing module according to claim 1, wherein, The step S101 comprises: By collecting historical running data and real-time maintenance records of the atomization module in the extended use scene, obtaining an environment parameter set containing electric stress level, scale deposition rate, thermal cycle frequency and corrosion factor influence, and obtaining a parameter expansion set; According to the parameter expansion set, a Gaussian mixture model is used, wherein the input is a numerical set of electric stress level, scale deposition rate, thermal cycle frequency and corrosion factor influence, the interaction influence is calculated by fitting multiple Gaussian distribution components, and the joint distribution range of thermal cycle frequency and corrosion factor influence is determined; For the joint distribution range, the parameter change trend and life prediction index under the interaction of multiple stresses are obtained by a statistical regression method, wherein the life prediction index is calculated based on the cumulative effect of the change trend, and the initial distribution characteristics are judged; If the initial distribution characteristics and the life prediction index do not match, the parameter expansion set is integrated and adjusted by updating the weight of the corrosion factor influence to obtain a distribution characteristic containing cleaning cycle optimization.

3. The reliability evaluation method of a humidifier atomizing module according to claim 1, wherein, The step S102 comprises: Obtaining an environment parameter set, extracting a temperature parameter and a humidity parameter from the environment parameter set, and processing an interaction relationship between the temperature parameter and the humidity parameter by using a preset stress coupling model through a product interaction formula, wherein the product interaction formula is a temperature value multiplied by a humidity value and then divided by a standardization factor, to obtain a temperature-humidity interaction matrix; According to the temperature-humidity interaction matrix, simulating a thermal-humidity stress distribution of the atomization module under a multi-stress condition, and determining a thermal-humidity stress distribution map by using a grid division method; Extracting an aging acceleration coefficient from the thermal-humidity stress distribution map, obtaining the aging acceleration coefficient by dividing a stress peak value in the cumulative distribution map by a time interval, and if the aging acceleration coefficient is higher than a preset threshold value, integrating a vibration parameter into the environment parameter set to obtain an extended parameter set; For the extended parameter set, constructing a comprehensive stress simulation scene, and calculating a cumulative damage value of the atomization module by using a unit strain calculation formula through finite element analysis, wherein the unit strain calculation formula is a stress value divided by a modulus value, to obtain a cumulative damage path; According to the cumulative damage path, fusing path optimization calculation to determine an accelerated aging path of the atomization module under the multi-stress condition.

4. The reliability evaluation method of a humidifier atomizing module according to claim 1, wherein, The step S103 comprises: Through the accelerated aging path, extracting a key failure indicator to obtain an influence degree of the scale deposition amount on the ceramic vibration characteristics; According to the influence degree, judging whether the key failure indicator exceeds a preset failure threshold value to obtain an exceeding state; If the exceeding state exists, determining a preliminary life prediction value by using a linear regression algorithm, wherein the linear regression algorithm fits data points by using a least square method to obtain a regression coefficient and an intercept; For the preliminary life prediction value, obtaining multi-index fusion data to determine an extended aging simulation path, wherein the multi-index fusion data is obtained from a deposition influence analysis and a vibration characteristic change; Through the extended aging simulation path, judging a comprehensive influence of the fused indicators on the vibration characteristics to obtain a refined life prediction value; Through the accelerated aging path, obtaining the scale deposition amount and the piezoelectric ceramic vibration data to obtain the key failure indicator; According to the key failure indicator, using index quantization analysis, wherein the index quantization analysis determines a deposition influence calculation of the scale deposition amount on the piezoelectric ceramic vibration by measuring a ratio of the scale deposition amount and a vibration frequency deviation; For the deposition influence calculation, obtaining a vibration characteristic change, wherein the vibration characteristic change is obtained by comparing an initial vibration amplitude and a current vibration amplitude, and if the vibration characteristic change exceeds a preset failure threshold value, judging that the influence degree evaluation is abnormal; From the influence degree evaluation, extracting path data extraction results, wherein the path data extraction results are obtained by integrating time sequence data in the accelerated aging path to obtain a life prediction value; Through the life prediction value, combining a prediction value adjustment, wherein the prediction value adjustment is obtained by calibrating historical failure data to obtain a preliminary life prediction value.

5. The reliability evaluation method of a humidifier atomizing module according to claim 1, wherein, The step S104 comprises: By obtaining the preliminary life prediction value, comparing with a reliability standard to judge whether it is lower than a threshold value to obtain a preliminary evaluation result; If the preliminary evaluation result is lower than the threshold value, adjusting a thermal cycle parameter to determine an adjusted simulation scene; From the adjusted simulation scene, a finite element analysis method is adopted, which calculates the stress distribution and thermal strain interaction of each element by meshing the simulation scene into finite elements, to obtain the recalculated influence value; According to the recalculated influence value, an alternative scheme of accelerated test conditions is obtained, and the applicability of the alternative scheme is judged to obtain an optimized option; Through the optimized option, the parameter optimization cycle and the final test scheme are fused to determine the optimized accelerated test conditions.

6. The reliability evaluation method of a humidifier atomizing module according to claim 1, wherein, The step S105 includes: Using the preset accelerated test conditions, the atomization module sample is run in a laboratory environment, and the real-time monitored scale deposition distribution data is obtained from the sample running to obtain the scale deposition distribution change; Through the scale deposition distribution change, combined with the simultaneously monitored electrical stress change data, the preliminary correlation of the composite failure process is determined to obtain the composite failure process correlation data; According to the composite failure process correlation data, the extended thermal stress distribution data is obtained from the sample running monitoring as a new unique service, and the thermal stress distribution data is collected through a thermal stress sensor, and if the thermal stress distribution data exceeds a preset threshold, the correlation data is fused to obtain the fused failure data; For the fused failure data, a time series change curve is obtained to obtain the dynamic trend of the composite failure process.

7. The reliability evaluation method of a humidifier atomizing module according to claim 1, wherein, The step S106 includes: From the historical failure data, a dynamic trend sequence is obtained, and a support vector machine algorithm is used to construct a classification hyperplane, the dynamic trend sequence is mapped to a high-dimensional space as input data, and the key feature vector in the dynamic trend sequence is extracted; For the key feature vector, the actual failure mode is identified, the cosine similarity is used to calculate the included angle between the key feature vector and the actual failure mode vector to obtain the matching degree score of the key feature vector and the actual failure mode; It is judged whether the matching degree score reaches a preset threshold, and if the preset threshold is reached, a preliminary failure prediction model is determined; A verification data set of the preliminary failure prediction model is obtained, and the weight parameters and bias parameters of the preliminary failure prediction model are adjusted through gradient descent iteration optimization; According to the adjusted parameters, a refined failure prediction model is obtained.

8. The reliability evaluation method of a humidifier atomizing module according to claim 1, wherein, The step S107 includes: Through the obtained refined failure prediction model, the model input is historical failure data and water quality parameters, and the output is a predicted failure probability, the thermal cycle parameters such as cycle number and temperature range are obtained from the water quality change scene to determine the thermal cycle influence trend, which is obtained by comparing the correlation between parameter change and failure probability; According to the thermal cycle influence trend, the module response data under multiple water quality change scenes is simulated, the Monte Carlo method is used to generate random water quality variation to obtain a response data set, which contains module durability and response time indicators; Using the response data set, the influencing factors such as temperature fluctuation and water quality composition are integrated, the contribution of each factor to the response is calculated by weighted average, the accelerated test conditions are judged, and if the total contribution is higher than the threshold, it is considered as high risk, and the condition threshold is obtained; If the condition threshold exceeds a preset threshold, adjusting the test protocol parameters, the adjustment is based on the threshold deviation ratio to modify the test cycle and intensity, determining the final accelerated test protocol, which includes a specific thermal cycle sequence; For the final accelerated test protocol, apply in the batch reliability evaluation of the atomization module, the application is executed by a batch sample, verify the batch verification result, which reflects the overall reliability of the module under the simulated scenario.

9. The reliability evaluation method of a humidifier atomizing module according to claim 1, wherein, The step S108 comprises: Through the accelerated test protocol, obtain the initial set of historical data and newly collected data from the historical data fusion, and obtain the fusion data set; According to the fusion data set, a linear regression algorithm is used, wherein the input is a pair of variables in the fusion data set, and the output is a fitting coefficient. The quantitative index is calculated by the fitting coefficient to determine the quantitative index value of the product quality control; Through the quantitative index value, the interaction analysis is carried out, wherein the interaction analysis evaluates the influence between variables through the product term to obtain the index change under the interaction and judge the index interaction stability; If the index interaction stability is lower than the preset threshold, obtain the defect detection process from the reliability prediction mechanism, wherein the defect detection process obtains the defect distribution characteristics by scanning the abnormal points of the fusion data set; According to the defect distribution characteristics, the index threshold setting and stability verification standard are adopted, wherein the stability verification standard evaluates the fluctuation by variance test to judge the stability of the defect distribution characteristics under the interaction.

10. A reliability evaluation system for a humidifier atomization module, characterized in that, The system is used for implementing the reliability evaluation method of the humidifier atomization module according to any one of claims 1-9, and the system comprises: an environmental parameter acquisition module, which is used for acquiring a set of environmental parameters including an electrical stress level, a scale deposition rate and a thermal cycle frequency by acquiring historical operation data of the atomization module in an actual use scenario, and determining initial distribution characteristics of the parameters under multi-stress interaction; a stress coupling analysis module, which is used for obtaining an accelerated aging path of the atomization module under simulated multi-stress conditions by processing the interaction relationship among the parameters according to the obtained set of environmental parameters and using a preset stress coupling model; an failure index evaluation module, which is used for extracting key failure indexes such as the influence degree of scale deposition on piezoelectric ceramic vibration characteristics from the obtained accelerated aging path, judging whether the indexes exceed a preset failure threshold, and obtaining a preliminary life prediction value; a test condition optimization module, which is used for re-calculating the influence of interaction and determining optimized accelerated test conditions by adjusting thermal cycle parameters in simulation if the preliminary life prediction value is lower than a target reliability evaluation standard; a dynamic trend monitoring module, which is used for obtaining a dynamic trend of a compound failure process by running an atomization module sample in a laboratory environment, acquiring real-time monitored scale deposition distribution and electrical stress change data, and determining the dynamic trend; a failure model refining module, which is used for analyzing the matching degree of the trend and an actual failure mode by applying a support vector machine algorithm to the obtained dynamic trend, judging whether the matching degree reaches a preset threshold, and obtaining a refined failure prediction model; an accelerated test protocol generation module, which is used for simulating the influence of thermal cycles under multiple water quality change scenarios by using the obtained refined failure prediction model, determining a final accelerated test protocol, and using the final accelerated test protocol for batch reliability evaluation of the atomization module; a quality control evaluation module, which is used for processing fusion of historical data and newly acquired data according to the determined accelerated test protocol, obtaining quantitative indexes for product quality control, and judging the stability of the indexes under interaction; and a prediction accuracy enhancement module, which is used for optimizing protocol parameters by using a gradient boosting decision tree algorithm to obtain enhanced prediction accuracy and realize rapid prediction of a real failure mode if the quantitative indexes show insufficient stability.

Images