LCM (Liquid Crystal Module) aging test method, device and equipment and storage medium

Abstract

The invention provides an LCM module aging test method, device and equipment and a storage medium, and the method comprises the steps: obtaining the initial state data and initial environment parameter combination of an LCM module through intelligent environment simulation and initialization processing, and carrying out the multi-mode alternate aging test and real-time monitoring processing of the module through an intelligent environment simulation system, and obtaining performance data in a dynamic environment, and carrying out surface periodic scanning and working sound recording to obtain microstructure data and acoustic data. Through intelligent fault detection and deep analysis processing, fault features and reasons of the module are identified, predictive analysis is performed on the fault in combination with a machine learning algorithm, and a predictive fault trend and an optimized environmental parameter combination are generated. And performing comprehensive performance evaluation on the module according to the fault features, the reason analysis, the prediction trend and the optimization parameters. According to the invention, through combination of intelligent environment simulation, real-time monitoring and deep analysis, accurate performance evaluation and fault prediction of the LCM are realized.

Description

Technical Field

[0001] The present invention relates to the field of testing technology, and in particular to an LCM module aging test method, device, equipment and storage medium. Background Art

[0002] LCM modules (liquid crystal display modules) are key components widely used in modern electronic devices. Their quality and reliability directly affect the performance and user experience of terminal products. In practical applications, LCM modules need to undergo aging tests to simulate the working state under long-term use conditions and evaluate their performance stability and potential failures. However, traditional aging test methods usually use a single environmental parameter for static testing and only perform simple performance tests on specific working modes. This method has obvious deficiencies in test accuracy and comprehensiveness, and it is difficult to truly reflect the long-term performance of the module in a complex dynamic environment.

[0003] In addition, existing aging test methods often rely on manual analysis and qualitative judgment, lacking the ability to deeply mine and intelligently analyze test data, making it difficult to effectively identify the potential failure characteristics and causes of the module. This limitation makes it difficult to optimize module performance in a timely manner during the product design and production stages, thereby increasing the risk of product failure and subsequent maintenance costs. Summary of the invention

[0004] The main purpose of the present invention is to solve the technical problems of the existing LCM module aging test method, such as single test environment, insufficient dynamic simulation capability, incomplete fault detection, and lack of intelligent analysis and predictive evaluation capabilities; A first aspect of the present invention provides an LCM module aging test method, the LCM module aging test method comprising: Perform intelligent environment simulation and initialization processing on the target LCM module to obtain the initial state data and initial environment parameter combination of the target LCM module, and initialize the intelligent environment simulation system according to the preset dynamic environment adjustment algorithm; Using the intelligent environment simulation system, a multi-mode alternating aging test and real-time monitoring process are performed on the target LCM module according to the initial state data and the initial environment parameter combination to obtain the performance data of the target LCM module under the dynamically adjusted environment conditions, and the surface of the target LCM module is periodically scanned and the working sound is recorded to obtain microstructure data and acoustic data; Perform intelligent fault detection and in-depth analysis on the performance data, microstructure data and acoustic data to obtain fault characteristics and cause analysis of the target LCM module, and use a machine learning algorithm to perform predictive analysis on the fault characteristics and cause analysis to obtain predictive fault trends and optimized environmental parameter combinations; According to the fault characteristics, cause analysis, predictive fault trends and optimized environmental parameter combinations, a comprehensive performance evaluation is performed on the target LCM module to obtain a quantitative performance evaluation result.

[0005] Optionally, in a first implementation of the first aspect of the present invention, the intelligent environment simulation and initialization processing is performed on the target LCM module to obtain the initial state data and initial environment parameter combination of the target LCM module, and the intelligent environment simulation system is initialized according to a preset dynamic environment adjustment algorithm, including: Performing a pre-aging process on the target LCM module to obtain an initial response characteristic curve of the target LCM module, and generating initial state data according to the initial response characteristic curve; The environmental simulation chamber is parameterized according to the initial state data to obtain an initial environmental parameter combination, and a high-precision optical scanning system is used to perform an initial scan on the surface of the target LCM module to obtain initial microstructure data; The initial response characteristic curve, the initial environmental parameter combination and the initial microstructure data are input into a preset dynamic environment adjustment algorithm, and the intelligent environment simulation system is initialized according to the output initial adjustment strategy.

[0006] Optionally, in a second implementation of the first aspect of the present invention, the intelligent environment simulation system is used to perform a multi-mode alternating aging test and real-time monitoring processing on the target LCM module according to the initial state data and the initial environmental parameter combination, to obtain the performance data of the target LCM module under the dynamically adjusted environmental conditions, and to periodically scan the surface of the target LCM module and record the working sound, to obtain the microstructure data and acoustic data including: According to the preset multi-mode test sequence, the target LCM module is tested for static image display, dynamic video playback and grayscale gradient to obtain display performance data in different modes; Use the intelligent environmental simulation system to dynamically adjust temperature, humidity, light intensity and electromagnetic interference intensity to obtain the response data of the target LCM module under different environmental conditions; A high-precision optical scanning system is used to periodically scan the surface of the target LCM module to obtain microstructure change data, and a high-sensitivity microphone array is used to record the acoustic characteristics of the target LCM module under different working conditions to obtain acoustic data.

[0007] Optionally, in a third implementation of the first aspect of the present invention, the intelligent fault detection and in-depth analysis processing of the performance data, microstructure data and acoustic data are performed to obtain the fault characteristics and cause analysis of the target LCM module, and a machine learning algorithm is used to perform predictive analysis on the fault characteristics and cause analysis to obtain a predictive fault trend and an optimized environmental parameter combination, including: Perform multi-dimensional feature extraction on performance data, microstructure data and acoustic data to obtain the comprehensive feature vector of the target LCM module; Analyzing the comprehensive feature vector using a pre-trained fault diagnosis model to obtain the fault characteristics and fault degree assessment results of the target LCM module; According to the fault characteristics and fault degree assessment results, combined with similar cases in the historical database, the fault cause is traced and analyzed to obtain a fault cause analysis report; A machine learning algorithm is used to perform predictive analysis on the fault characteristics and cause analysis to obtain a predictive fault trend and an optimized environmental parameter combination.

[0008] Optionally, in a fourth implementation of the first aspect of the present invention, the fault cause tracing analysis is performed based on the fault characteristics and the fault degree assessment result in combination with similar cases in the historical database to obtain a fault cause analysis report, which includes: Performing feature vectorization processing on the fault characteristics and the fault degree assessment results to obtain a fault feature vector; Using the fault feature vector to perform similarity matching in the historical database to obtain a preliminary matching case set, and sorting the case set by similarity based on weighted Euclidean distance to obtain N most similar cases, where N is a natural number greater than 1; Perform multi-dimensional cluster analysis on the N most similar cases to obtain the main fault types and key influencing factors; According to the main fault types and key influencing factors, combined with the initial state data and initial environmental parameter combination of the target LCM module, a multi-layer causal relationship graph is constructed, and the multi-layer causal relationship graph is analyzed by a graph traversal algorithm to obtain the fault cause chain; Based on the fault cause chain and historical solutions, the decision tree algorithm is used to generate fault diagnosis and solution recommendations to obtain a fault cause analysis report.

[0009] Optionally, in a fifth implementation of the first aspect of the present invention, the using of a machine learning algorithm to perform predictive analysis on the fault characteristics and cause analysis to obtain a predictive fault trend and an optimized environmental parameter combination includes: Perform text mining and feature extraction on fault characteristics and cause analysis reports to obtain key fault indicators and environmental factors; Using a long short-term memory network to perform time series analysis on historical data of key fault indicators and environmental factors, and obtain future trend prediction results of the key fault indicators and change patterns of the environmental factors; The time-lag correlation between fault indicators and environmental factors is evaluated by time series correlation analysis method to obtain dynamic correlation characteristics; Based on the future trend prediction results of fault indicators, the change patterns and dynamic correlation characteristics of environmental factors, a multi-objective optimization problem is constructed, and the multi-objective optimization problem is solved with the goal of minimizing the risk of faults and maximizing the performance stability, and a preliminary environmental parameter adjustment plan is obtained; Using the reinforcement learning algorithm, based on the preliminary environmental parameter adjustment plan, the optimized environmental parameter combination is obtained through virtual environment simulation and iterative optimization; The future trend prediction result and the optimized environmental parameters are combined and input into a pre-trained fault evolution model, and a plurality of possible fault development paths are generated through Monte Carlo simulation to obtain a probability-weighted predictive fault trend.

[0010] Optionally, in a sixth implementation of the first aspect of the present invention, the comprehensive performance evaluation of the target LCM module is performed according to the fault characteristics, cause analysis, predictive fault trend and optimized environmental parameter combination, and the quantitative performance evaluation result obtained includes: According to the fault characteristics and cause analysis, the key aging indicators of the LCM module to be tested are determined, and the aging test evaluation indicator set is obtained; Using the optimized environmental parameter combination, the LCM module to be tested is subjected to accelerated aging test, and the changes of various aging indicators over time are recorded to obtain aging curve data; Based on the predictive failure trend and aging curve data, the aging rate of the LCM module under normal use conditions is predicted by data extrapolation method to obtain the estimated aging life; Taking into account the aging test evaluation index set, aging curve data and aging life estimation, the comprehensive aging performance score of the LCM module to be tested is calculated by the weighted scoring method to obtain a quantitative performance evaluation result.

[0011] A second aspect of the present invention provides an LCM module aging test device, the LCM module aging test device comprising: The environment simulation module is used to perform intelligent environment simulation and initialization processing on the target LCM module, obtain the initial state data and initial environment parameter combination of the target LCM module, and initialize the intelligent environment simulation system according to the preset dynamic environment adjustment algorithm; An aging test module is used to use the intelligent environment simulation system to perform multi-mode alternating aging tests and real-time monitoring processing on the target LCM module according to the initial state data and the initial environment parameter combination, to obtain the performance data of the target LCM module under the dynamically adjusted environment conditions, and to periodically scan the surface of the target LCM module and record the working sound to obtain microstructure data and acoustic data; A fault analysis module is used to perform intelligent fault detection and in-depth analysis and processing on the performance data, microstructure data and acoustic data to obtain the fault characteristics and cause analysis of the target LCM module, and use a machine learning algorithm to perform predictive analysis on the fault characteristics and cause analysis to obtain predictive fault trends and optimized environmental parameter combinations; The performance evaluation module is used to perform a comprehensive performance evaluation on the target LCM module according to the fault characteristics, cause analysis, predictive fault trends and optimized environmental parameter combinations to obtain a quantitative performance evaluation result.

[0012] The third aspect of the present invention provides an LCM module aging test device, comprising: a memory and at least one processor, wherein instructions are stored in the memory, and the memory and the at least one processor are interconnected via lines; the at least one processor calls the instructions in the memory so that the LCM module aging test device executes the steps of the above-mentioned LCM module aging test method.

[0013] A fourth aspect of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores instructions, which, when executed on a computer, enable the computer to execute the steps of the above-mentioned LCM module aging test method.

[0014] The above-mentioned LCM module aging test method, device, equipment and storage medium obtain the initial state data and initial environmental parameter combination of the LCM module through intelligent environmental simulation and initialization processing, and use the intelligent environmental simulation system to perform multi-mode alternating aging testing and real-time monitoring processing on the module, obtain performance data under dynamic environment, perform periodic surface scanning and working sound recording, and obtain microstructure data and acoustic data. Through intelligent fault detection and in-depth analysis processing, the fault characteristics and causes of the module are identified, and the fault is predictively analyzed in combination with the machine learning algorithm to generate predictive fault trends and optimized environmental parameter combinations. According to the fault characteristics, cause analysis, prediction trends and optimization parameters, the module is comprehensively evaluated for performance. The present invention realizes accurate performance evaluation and fault prediction of the LCM module by combining intelligent environmental simulation, real-time monitoring and in-depth analysis.

[0015] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention are realized and obtained by the structures particularly pointed out in the description, claims and drawings.

[0016] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 This is a schematic diagram of a first embodiment of an LCM module aging test method according to an embodiment of the present invention; Figure 2 A schematic diagram of an embodiment of an LCM module aging test device in an embodiment of the present invention; Figure 3 It is a schematic diagram of an embodiment of an LCM module aging test device in an embodiment of the present invention. DETAILED DESCRIPTION

[0018] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0019] The terms "including" and "having" and any variations thereof mentioned in the embodiments of the present invention are intended to cover non-exclusive inclusions. For example, a process, method, system, product, or device end including a series of steps or units is not limited to the listed steps or units, but may optionally include other steps or units that are not listed, or may optionally include other steps or units that are inherent to these processes, methods, products, or device ends.

[0020] To facilitate understanding of this embodiment, a method for testing LCM module aging disclosed in an embodiment of the present invention is first described in detail. Figure 1 As shown, the method comprises the following steps: 101. Perform intelligent environment simulation and initialization processing on the target LCM module to obtain the initial state data and initial environment parameter combination of the target LCM module, and initialize the intelligent environment simulation system according to a preset dynamic environment adjustment algorithm; In one embodiment of the present invention, the intelligent environmental simulation and initialization processing of the target LCM module to obtain the initial state data and initial environmental parameter combination of the target LCM module, and initialize the intelligent environmental simulation system according to a preset dynamic environmental adjustment algorithm includes: performing pre-aging processing on the target LCM module to obtain the initial response characteristic curve of the target LCM module, and generating initial state data according to the initial response characteristic curve; setting parameters of the environmental simulation cabin according to the initial state data to obtain the initial environmental parameter combination, and using a high-precision optical scanning system to perform an initial scan on the surface of the target LCM module to obtain initial microstructure data; inputting the initial response characteristic curve, the initial environmental parameter combination and the initial microstructure data into a preset dynamic environmental adjustment algorithm, and initializing the intelligent environmental simulation system according to the output initial adjustment strategy.

[0021] Specifically, when the target LCM module is pre-aged, the module is placed under standard environmental conditions (such as temperature 25°C and humidity 50%) for 24 hours. During this period, the key performance parameters of the module are recorded every 10 minutes, including brightness (unit: cd / m²), contrast, response time (unit: ms), color saturation and uniformity. These data form a time series for drawing the initial response characteristic curve. The x-axis of the curve is time (unit: hour), and the y-axis is the normalized value of each parameter. By fitting the curve with a polynomial, the function of the change of each parameter over time is obtained. From these functions, characteristic values ​​are extracted, such as the initial value, stable time (defined as the moment when the parameter change rate is less than 1% / hour), the final value after 24 hours, and the fluctuation range (the difference between the maximum and minimum values). These characteristic values ​​constitute the initial state data, which is stored in JSON format and contains the initial value, stable value, stable time and fluctuation range of each parameter.

[0022] Specifically, when setting the parameters of the environmental simulation chamber according to the initial state data, the test range is first determined. The temperature range is set to -10°C to 60°C, the humidity range is 20% to 90%, the light intensity range is 0 to 1000 lux, and the electromagnetic interference frequency range is 50Hz to 1GHz. The Latin hypercube sampling method is used to generate 50 sets of initial environmental parameter combinations within these ranges. Each set of parameters includes specific temperature values, humidity values, light intensity curves (defined as a function of changes within 24 hours), and electromagnetic interference spectra (defined as intensity values ​​at different frequencies). These 50 sets of parameters are encoded into a 50x4 matrix, where each row represents a set of parameters and the columns correspond to temperature, humidity, light, and electromagnetic interference, respectively. At the same time, the high-precision optical scanning system is started, and the surface of the LCM module is scanned using a three-dimensional profiler with a resolution of 1μm. After filtering and denoising the scanned data, a height map is generated with a resolution of 1024x1024 pixels. This height map is subjected to Fourier transformation and wavelet analysis to extract features such as surface roughness, waviness, and shape deviation. In addition, image recognition algorithms are used to detect surface defects such as scratches, bubbles, or foreign matter, and their location, size, and shape are recorded. These data constitute the initial microstructural data, which are stored in HDF5 format.

[0023] Specifically, the initial response characteristic curve (expressed as polynomial coefficients), the initial environmental parameter combination matrix, and the initial microstructure data (including surface feature parameters and defect information) are input into the preset dynamic environmental adjustment algorithm. The algorithm is based on a deep reinforcement learning model and uses a multi-layer neural network as a policy network and a value network. The input layer of the policy network contains 174 nodes, corresponding to the dimensions of the initial data. The hidden layer uses 3 fully connected layers, each with 512 nodes. The output layer is 50 nodes, corresponding to the adjustment strategies for 50 sets of environmental parameters. The algorithm first standardizes the input data and then generates an initial adjustment strategy through the policy network. This strategy defines how to dynamically adjust 50 sets of environmental parameters within a 24-hour test cycle, including the specific value or rate of change of each set of parameters in each hour. For example, the temperature may be set to start from 25°C and rise by 2°C every hour until it reaches the highest value and then gradually decreases. The strategy also includes the priority and interaction rules of parameter changes, such as triggering a rapid change in humidity when the temperature reaches a certain threshold. The initial adjustment strategy output by the algorithm is encoded as a 50x24x4 three-dimensional tensor, where 50 represents the number of parameter groups, 24 represents a 24-hour cycle, and 4 represents four environmental factors. The intelligent environmental simulation system sets the parameters of each control module based on this tensor, such as the PID parameters of the thermostat, the feedback gain of the humidity controller, the PWM control sequence of the light source, and the frequency scanning program of the electromagnetic interference generator. The system also establishes a real-time data acquisition and feedback mechanism to collect performance data and environmental parameters of the LCM module once a minute to provide input for subsequent dynamic adjustments.

[0024] 102. Using the intelligent environmental simulation system, according to the initial state data and the initial environmental parameter combination, a multi-mode alternating aging test and real-time monitoring process are performed on the target LCM module to obtain the performance data of the target LCM module under the dynamically adjusted environmental conditions, and the surface of the target LCM module is periodically scanned and the working sound is recorded to obtain microstructure data and acoustic data; In one embodiment of the present invention, the intelligent environment simulation system is used to perform a multi-mode alternating aging test and real-time monitoring processing on the target LCM module according to the initial state data and the initial environment parameter combination, to obtain the performance data of the target LCM module under the dynamically adjusted environmental conditions, and to periodically scan the surface of the target LCM module and record the working sound to obtain microstructure data and acoustic data, including: according to a preset multi-mode test sequence, a static image display test, a dynamic video playback test and a grayscale gradient test are performed on the target LCM module to obtain display performance data under different modes; the intelligent environment simulation system is used to dynamically adjust the temperature, humidity, light intensity and electromagnetic interference intensity to obtain the response data of the target LCM module under different environmental conditions; a high-precision optical scanning system is used to periodically scan the surface of the target LCM module to obtain microstructure change data, and a high-sensitivity microphone array is used to record the acoustic characteristics of the target LCM module under different working conditions to obtain acoustic data.

[0025] Specifically, when testing the target LCM module according to the preset multi-mode test sequence, a static image display test is first performed. The test includes a series of standard test images, such as pure color blocks (red, green, blue, white, black), checkerboards, grayscale steps, and color gradients. Each image is displayed for 10 minutes, during which the brightness (unit: cd / m²) and chromaticity coordinates (CIE 1976 Lab*) of the center point and four corner points of the screen are measured every 30 seconds using a calibrated colorimeter. At the same time, a high-speed camera (1000 frames / second) is used to capture the screen refresh process and calculate the response time (the time required for a 10% to 90% change between grayscales). These data are organized into a time series for evaluating brightness uniformity, color accuracy, and response speed. Next, a dynamic video playback test is performed using a standard test video sequence containing fast motion scenes, high-contrast transitions, and complex textures. The video is played at a frame rate of 60Hz for 30 minutes, during which the peak signal-to-noise ratio (PSNR) and structural similarity (SSIM) of each frame are calculated in real time using video analysis software to evaluate dynamic picture quality. At the same time, power consumption data is recorded to evaluate the energy consumption performance under different screen contents. Finally, a grayscale gradient test is performed to display a grayscale gradient image from 0 to 255, with each grayscale lasting 1 second. A high dynamic range camera is used to capture the entire process, and the color banding phenomenon is detected through image analysis algorithms, and the grayscale resolution (the number of distinguishable grayscale levels) and gamma curve are calculated. The data of these three test modes are integrated into a multidimensional array to form a display performance data set, including key indicators such as static display quality, dynamic response characteristics, and color reproduction capabilities.

[0026] Specifically, when the intelligent environment simulation system is used to dynamically adjust the environmental parameters, the temperature, humidity, light intensity and electromagnetic interference intensity are continuously changed within the 24-hour test period according to the initial adjustment strategy. The temperature control adopts the PID algorithm, changing at a rate of 0.1°C / minute in the range of -10°C to 60°C, and staying at each temperature point for 30 minutes. The humidity control uses a variable frequency humidifier and dehumidifier, which is adjusted at a rate of 2% / minute in the range of 20% to 90%, and is coordinated with the temperature change. The light intensity is simulated by the LED array, changing on a logarithmic scale from 0 to 1000 lux, and each intensity level is maintained for 1 hour. Electromagnetic interference is generated using a programmable signal generator, with a frequency sweep from 50Hz to 1GHz and an intensity ranging from 0 to 10V / m. Under each combination of environmental conditions, the aforementioned display performance test is repeated, and the electrical parameters of the LCM module, including driving voltage, current and power consumption, are recorded. The temperature distribution on the screen surface is measured using a thermocouple array (16x16 grid) with a sampling frequency of 1Hz. All environmental parameters and LCM module response data are synchronously recorded into a time series database with a sampling rate of 1 second / time. These data are pre-processed, including denoising, outlier detection and interpolation, to form a multidimensional tensor, which contains the changes in environmental parameters, display performance indicators and electrical characteristics in the time dimension. In this way, the comprehensive response data of the target LCM module under different environmental conditions is obtained, reflecting the sensitivity and stability of the module performance to environmental factors.

[0027] Specifically, in the process of periodically scanning the surface of the target LCM module using a high-precision optical scanning system, white light interferometer technology is used with a resolution of 0.1μm. The scanning frequency is set to once every 4 hours, and each scan covers the entire display area with a resolution of 4096x4096 pixels. The scanned data is image stitched and reconstructed in three dimensions to generate a surface topology map. The topology map is processed using Fourier transform and wavelet analysis to extract surface roughness parameters (such as Ra, Rz), waviness and shape deviation. At the same time, computer vision algorithms are applied to detect surface defects such as scratches, bubbles or foreign objects, and their position, size and shape changes are recorded. These data form a time series that reflects the evolution of the surface microstructure of the LCM module with the aging process. In parallel, a high-sensitivity microphone array (24 omnidirectional microphones, sampling rate 96kHz, 24-bit quantization) is used to record the acoustic characteristics of the LCM module under different working conditions. The microphone array is spherically distributed with a radius of 0.5 meters and the center is aligned with the LCM module. During the recording process, the ambient background noise is controlled below 30dB. The sound data is spectrally analyzed to calculate the power spectral density (PSD) and sound pressure level (SPL). Beamforming technology is used to locate the sound source and identify the source of abnormal sound. The spectrogram is generated by short-time Fourier transform (STFT) to analyze the time-frequency characteristics of the sound. The acoustic data also includes harmonic distortion analysis and calculation of the total harmonic distortion (THD) index. These acoustic feature data are organized into a multidimensional array, with each dimension corresponding to an acoustic parameter and a sampling interval of 1 minute on the time axis. In this way, a comprehensive data set reflecting the microstructural changes and acoustic characteristic evolution of the LCM module is obtained, providing multi-dimensional information input for subsequent fault diagnosis and performance evaluation.

[0028] 103. Perform intelligent fault detection and in-depth analysis on performance data, microstructure data and acoustic data to obtain the fault characteristics and cause analysis of the target LCM module, and use machine learning algorithms to perform predictive analysis on the fault characteristics and cause analysis to obtain predictive fault trends and optimized environmental parameter combinations; In one embodiment of the present invention, the intelligent fault detection and in-depth analysis processing of the performance data, microstructure data and acoustic data to obtain the fault characteristics and cause analysis of the target LCM module, and the predictive analysis of the fault characteristics and cause analysis using a machine learning algorithm to obtain a predictive fault trend and an optimized environmental parameter combination include: multi-dimensional feature extraction of the performance data, microstructure data and acoustic data to obtain a comprehensive feature vector of the target LCM module; using a pre-trained fault diagnosis model to analyze the comprehensive feature vector to obtain the fault characteristics and fault severity assessment results of the target LCM module; based on the fault characteristics and fault severity assessment results, combined with similar cases in a historical database, a fault cause tracing analysis is performed to obtain a fault cause analysis report; using a machine learning algorithm to predictively analyze the fault characteristics and cause analysis to obtain a predictive fault trend and an optimized environmental parameter combination.

[0029] Specifically, when extracting multi-dimensional features from performance data, microstructure data, and acoustic data, the principal component analysis (PCA) algorithm is first applied to the performance data to reduce the high-dimensional data to 50 principal components, retaining 95% of the variance information. These principal components reflect the change patterns of key indicators such as brightness, contrast, and response time. For microstructure data, a convolutional neural network (CNN) is used to extract spatial features. The network structure contains 5 convolutional layers and 3 fully connected layers, outputting a 512-dimensional feature vector to capture surface morphology and defect information. For acoustic data, 20 coefficients are extracted through Mel-frequency cepstral coefficients (MFCC), and then the long short-term memory network (LSTM) is used to process the timing information to obtain a 128-dimensional acoustic feature vector. After normalization, these three sets of feature vectors are feature fused through an autoencoder, which contains 3 layers of encoders and 3 layers of decoders, with an intermediate layer dimension of 256, and finally a 256-dimensional comprehensive feature vector is obtained. This vector contains key information on the performance, structure, and acoustic characteristics of the LCM module, providing a comprehensive data basis for subsequent fault diagnosis.

[0030] Specifically, an integrated model based on transfer learning was used to analyze the comprehensive feature vector using the pre-trained fault diagnosis model. The model consists of three sub-models: support vector machine (SVM), random forest (RF), and deep neural network (DNN). SVM uses the RBF kernel function with a C value of 10 and a gamma of 0.01 to process linearly inseparable feature spaces. The random forest contains 100 decision trees with a maximum depth of 20 to capture complex interactions between features. The DNN has 5 hidden layers, with 512, 256, 128, 64, and 32 nodes per layer, respectively. The ReLU activation function and Dropout (rate of 0.5) are used to prevent overfitting. The three models analyze the comprehensive feature vectors respectively and output the probability distribution of the fault type and the fault severity score (0-100 points). Then the results of the three models are fused by the weighted voting method (weight ratio of 3:2:5) to obtain the final fault diagnosis result. This result includes the top-3 most likely fault types and their probabilities, as well as the corresponding fault severity scores. At the same time, by analyzing the attention weights of the model, the features that contribute most to fault judgment are identified to form a fault feature description. This multi-model fusion method improves the accuracy and robustness of diagnosis and can handle complex fault modes.

[0031] Specifically, when tracing the cause of the fault according to the fault characteristics and fault degree assessment results, the fault feature vector of the current case is first matched with the cases in the historical database for similarity. The cosine similarity algorithm is used to calculate the distance between feature vectors, and the historical cases with the top-50 similarities are selected. The hierarchical clustering algorithm is applied to these cases, and the clustering tree is constructed using Ward's minimum variance method to determine the optimal number of clusters (assessed by the silhouette coefficient). Each cluster represents a class of similar fault modes. Then, the cases in each cluster are mined for association rules, and the Apriori algorithm is used to extract the association rules between fault characteristics, environmental conditions and fault causes. The support threshold is set to 0.3 and the confidence threshold is set to 0.7. These rules form a preliminary causal relationship network. Next, the Bayesian network is applied to probabilistically model the causal relationship, and the network parameters are learned using the maximum likelihood estimation method. Through backward reasoning, the conditional probabilities of various possible causes given the current fault characteristics are calculated. Combined with the expert knowledge base (including the predefined fault-cause mapping table), the reasoning results are verified and refined. Finally, a structured failure cause analysis report is generated, including the most likely failure cause (probability > 0.8), key influencing factors, failure development path, and detailed information of similar historical cases.

[0032] Specifically, an integrated learning framework was used to perform predictive analysis of fault characteristics and cause analysis using machine learning algorithms. First, the sliding window method (window size of 24 hours, step size of 1 hour) was used to segment the historical data into time series and generate labels for each window (whether a fault will occur in the next 48 hours). Then, three base learners were constructed: LSTM network (2 layers, 128 units per layer), 1D-CNN (3 convolutional layers, 2 fully connected layers), and gradient boosted decision tree (GBDT, 500 trees, maximum depth of 5). These models were trained separately to predict the probability of future faults. The outputs of the three models were integrated by the SoftVoting method (weights of 0.4:0.3:0.3) to obtain the final fault probability prediction. For cases where the predicted probability exceeds a threshold (such as 0.7), a regression model (such as support vector regression, SVR) is further used to predict the specific fault occurrence time. At the same time, a genetic algorithm was applied to optimize the environmental parameters. The fitness function was defined as the inverse of the predicted fault probability, the population size was set to 100, and the iteration was 500 generations. Through crossover and mutation operations, new environmental parameter combinations are generated and evaluated. Finally, the parameter combination with the highest fitness is selected as the optimized environmental parameter. The output of this process includes the failure probability curve within the next 48 hours, the expected failure time (if applicable), and the optimized environmental parameter combination that can delay the occurrence of failure to the greatest extent.

[0033] Furthermore, the fault cause tracing analysis is performed based on the fault characteristics and fault degree assessment results in combination with similar cases in the historical database to obtain a fault cause analysis report, including: performing feature vector quantization processing on the fault characteristics and fault degree assessment results to obtain a fault feature vector; performing similarity matching in the historical database using the fault feature vector to obtain a preliminary matching case set, and performing similarity sorting on the case set based on weighted Euclidean distance to obtain N most similar cases, wherein N is a natural number greater than 1; performing multi-dimensional clustering analysis on the N most similar cases to obtain main fault types and key influencing factors; based on the main fault types and key influencing factors, in combination with the initial state data and initial environmental parameter combination of the target LCM module, a multi-layer causal relationship graph is constructed, and the multi-layer causal relationship graph is analyzed through a graph traversal algorithm to obtain a fault cause chain; based on the fault cause chain and historical solutions, a decision tree algorithm is used to generate fault diagnosis and solution recommendations to obtain a fault cause analysis report.

[0034] Specifically, when the fault features and fault severity assessment results are vectorized, the fault type is first encoded as a one-hot vector with a dimension of the predefined number of fault types (assuming 50). The fault severity score (0-100 points) is normalized by min-max. Then, the fault feature description is converted into a word embedding vector through a pre-trained Word2Vec model (the training corpus is a large number of LCM fault reports, and the vector dimension is 300), and the average value is taken as the feature representation. These vectors are spliced ​​together with normalized performance indicators (such as brightness decay rate, response time change rate, etc., a total of 20 indicators) to form a 371-dimensional initial feature vector. Then, principal component analysis (PCA) is used for dimensionality reduction, retaining the principal components required to explain 95% of the variance (assuming it is reduced to 100 dimensions). Finally, the t-SNE algorithm is applied for nonlinear dimensionality reduction to obtain the final 50-dimensional fault feature vector. This vector contains both the category and severity information of the fault, and also encodes the semantic information of the fault description and the changes in key performance indicators, providing a comprehensive and compact representation for subsequent similarity matching.

[0035] When using fault feature vectors for similarity matching in the historical database, the locality sensitive hashing (LSH) algorithm is used to accelerate the search process. The specific implementation uses random projection LSH to build 20 hash tables, and the projection dimension of each hash function is 10. This method can quickly find the approximate nearest neighbor in sublinear time. For the candidate set returned by LSH (assuming 1000 cases), the weighted Euclidean distance between them and the target fault feature vector is calculated. The weight vector is set according to the feature importance (determined by the feature importance score of the random forest), giving higher weights to key features. The distance calculation formula is sqrt(sum(w_i * (x_i - y_i)^2)), where w_i is the weight of the i-th dimension feature, and x_i and y_i are the i-th dimension values ​​of the target vector and the historical case vector, respectively. The candidate cases are sorted according to this weighted distance, and the first N (N is set to 50) are selected as the most similar cases. This process not only considers the overall similarity of the features, but also highlights the impact of key features through weight setting, thereby improving the accuracy and relevance of the match.

[0036] When performing multidimensional cluster analysis on the N most similar cases, a Gaussian mixture model (GMM) is used for soft clustering. First, the Bayesian Information Criterion (BIC) is used to determine the optimal number of clusters (assuming that 5 clusters are obtained). Then, the GMM parameters are initialized, and the expectation maximization (EM) algorithm is used to iteratively optimize the model parameters until the log-likelihood function converges or the maximum number of iterations (set to 100) is reached. Each case is assigned to the cluster with the highest probability while retaining its probability distribution to other clusters. Next, the mean and standard deviation of the features are calculated for each cluster, and the features with the highest discrimination (through analysis of variance ANOVA, p value < 0.01) are identified as the feature descriptors of the cluster. Based on these descriptors, combined with the predefined fault type-feature mapping table in the expert knowledge base, a main fault type label is assigned to each cluster. At the same time, by calculating the mutual information between each feature and the cluster label, the features with the top-10 mutual information values ​​are selected as key influencing factors. This clustering method based on probability model can not only handle complex data distribution, but also provide the membership degree of each case to different fault types, providing richer information for subsequent causal analysis.

[0037] When constructing a multi-layer causal relationship graph, the main fault types are first used as the top-level nodes of the graph, and the key influencing factors are used as the second-level nodes. Then, according to the initial state data and initial environmental parameter combination of the target LCM module, the corresponding nodes are added as the third layer. The causal strength between nodes is estimated using the structured equation model (SEM), and the parameters of the SEM are obtained by the maximum likelihood estimation method. The edge weights of the graph are set as the standardized path coefficients. Next, the PC algorithm (Peter-Clark algorithm) is applied for causal discovery to correct and supplement the initial graph structure constructed by expert knowledge. The PC algorithm first assumes that all nodes are connected, and then gradually removes unnecessary edges through conditional independence tests (using partial correlation coefficients, significance level 0.05). The resulting causal graph is drawn by a graph visualization library (such as NetworkX) for intuitive analysis. In order to obtain the fault cause chain, a depth-first search (DFS) algorithm is implemented. DFS starts from the fault type node and traverses downward along the path with the largest edge weight until it reaches a leaf node or the traversal depth exceeds a preset threshold (such as 5 layers). Each traversal path is recorded as a potential fault cause chain, and finally the top-3 paths with the highest total edge weight are selected as the most likely fault cause chain. This method not only considers direct causal relationships, but also discovers potential indirect impact paths, providing a comprehensive fault cause analysis.

[0038] When generating fault diagnosis and solution recommendations based on the fault cause chain and historical solutions, a decision tree model is first constructed. The features of the decision tree include the attributes of each node in the fault cause chain, and the target variable is the category of the solution. The decision tree is constructed using the C4.5 algorithm, the information gain ratio is used as the feature selection criterion, and pessimistic pruning is used to prevent overfitting. In order to deal with the problem of data imbalance (there are few solution samples for some rare faults), the SMOTE (Synthetic Minority Over-sampling Technique) algorithm is used to oversample the minority class. Each leaf node of the decision tree corresponds to a solution category, and the sample distribution of the node is recorded at the same time. When making recommendations, the fault cause chain data of the target LCM module is input into the decision tree to obtain a solution category. Then, from the historical solution set corresponding to the category, the top-5 specific solutions most similar to the current fault description are calculated using TF-IDF and cosine similarity. Finally, a structured fault cause analysis report is generated, including: 1) description of fault type and severity; 2) the main fault cause chain identified, the confidence of each link (based on the edge weight of the causal graph); 3) key influencing factors and their abnormality (deviation from the normal range); 4) recommended top-5 solutions, each solution includes specific operation steps, expected results and historical success rate; 5) brief description of similar historical cases and solution results. This report is converted into an easy-to-read text format through natural language generation (NLG) technology, and is accompanied by a visual causal relationship diagram, which makes it easier for technicians to quickly understand and execute the fault handling process.

[0039] Furthermore, the use of a machine learning algorithm to perform a predictive analysis on the fault characteristics and cause analysis to obtain a predictive fault trend and an optimized environmental parameter combination includes: performing text mining and feature extraction on the fault characteristics and cause analysis report to obtain key fault indicators and environmental factors; using a long short-term memory network to perform a time series analysis on the historical data of key fault indicators and environmental factors to obtain future trend prediction results of the key fault indicators and the change pattern of the environmental factors; evaluating the time-lag correlation between the fault indicators and environmental factors by a time series correlation analysis method to obtain dynamic correlation characteristics; constructing a multi-objective optimization problem based on the future trend prediction results of the fault indicators, the change pattern of the environmental factors and the dynamic correlation characteristics, and solving the multi-objective optimization problem with the goal of minimizing the risk of failure and maximizing performance stability to obtain a preliminary environmental parameter adjustment plan; using a reinforcement learning algorithm, based on the preliminary environmental parameter adjustment plan, obtaining an optimized environmental parameter combination through virtual environment simulation and iterative optimization; inputting the future trend prediction results and the optimized environmental parameter combination into a pre-trained fault evolution model, generating multiple possible fault development paths through Monte Carlo simulation, and obtaining a probability-weighted predictive fault trend.

[0040] Specifically, when performing text mining and feature extraction on fault characteristics and cause analysis reports, natural language processing technology is first used to preprocess the reports. This includes using the NLTK library for word segmentation, stop word removal, lemmatization, and other operations. Then, the term frequency-inverse document frequency (TF-IDF) algorithm is applied to calculate the importance score of each word, and the words with the top-100 scores are selected as the initial feature set. Next, the latent Dirichlet allocation (LDA) topic model is used, the number of topics is set to 10, and the Gibbs sampling algorithm is used for 1000 iterations to obtain the document-topic distribution and topic-word distribution. Topics with a topic probability greater than 0.1 are extracted as key fault types. For numerical data, such as performance indicators and environmental parameters, the independent component analysis (ICA) algorithm is applied for feature extraction, and independent components that explain 95% of the variance are retained. Finally, the text features, topic distribution, and numerical features are combined into a comprehensive feature vector. The recursive feature elimination (RFE) algorithm is used in combination with the random forest classifier to select the 50 most important features as the final key fault indicators and environmental factors. These metrics and factors are stored in JSON format, with each feature containing its name, type (text / numeric), importance score, and unit (if applicable).

[0041] Specifically, when using the long short-term memory network (LSTM) to perform time series analysis on the historical data of key fault indicators and environmental factors, the data is first time-aligned and normalized. The sliding window method is used with a window size of 48 hours and a step size of 1 hour to generate input sequences and target sequences. The LSTM network structure contains 2 LSTM layers, each with 128 units, followed by a dropout layer (rate 0.5) and a fully connected layer. The Adam optimizer is used with a learning rate of 0.001 and a batch size of 64, and 500 epochs are trained. To handle multivariate prediction problems, an encoder-decoder architecture is used, with the encoder part encoding the historical sequence and the decoder part generating the prediction sequence for the next 48 hours. For each key fault indicator, a separate LSTM model is trained. Environmental factors are used as auxiliary inputs and integrated into the prediction process through the attention mechanism. After the model training is completed, the data of the past 7 days is used as input to generate the fault indicator trend prediction for the next 48 hours. For environmental factors, the autoregressive integrated moving average (ARIMA) model is used to capture its change pattern, and the model order is determined by grid search and AIC criterion. The outputs of LSTM and ARIMA models represent the future trend prediction results of key fault indicators and the change pattern of environmental factors, respectively, and are stored in time series format, including timestamps, predicted values, and 95% confidence intervals.

[0042] Specifically, when evaluating the time-lag correlation between fault indicators and environmental factors by time series correlation analysis, the cross-correlation function (CCF) between each pair of fault indicators and environmental factors is first calculated. The CCF calculation is implemented using the numpy library, and the maximum time lag is set to ±72 hours. For each time lag, the Pearson correlation coefficient and its p-value are calculated. The time lag with the largest absolute value of the correlation coefficient and a p-value less than 0.05 is selected as the best time lag. Then, the dynamic time warping (DTW) algorithm is applied to calculate the similarity between time series. The DTW distance is calculated using the fastdtw library, and the window size is set to 24. Based on the DTW distance, the hierarchical clustering algorithm (using the Ward method) is used to group the time series to determine the set of indicators and factors with similar change patterns. Next, the Granger causality test is implemented using the statsmodels library, and the lag order is selected by the AIC criterion (the maximum lag is set to 24). For the indicator-factor pairs that pass the Granger causality test (p-value < 0.01), their nonlinear correlation is calculated, using the mutual information (MI) and transfer entropy (TE) indicators. MI is calculated using the mutual_info_score function of the sklearn library, and TE is implemented using the JIDT toolkit. Finally, the CCF, DTW distance, Granger causality test results, MI, and TE values ​​are combined into a multidimensional vector to represent the dynamic association characteristics of each pair of indicator-factors. These features are stored in a matrix form, with rows representing fault indicators and columns representing environmental factors, and each cell containing the dynamic association feature vector calculated above.

[0043] Specifically, when constructing a multi-objective optimization problem based on the future trend prediction results of fault indicators, the change pattern of environmental factors and the dynamic correlation characteristics, two objective functions are first defined: the fault risk function and the performance stability function. The fault risk function uses the predicted fault indicator values ​​to calculate the overall risk through weighted summation, and the weight is based on the importance score of each indicator. The performance stability function is defined as the variance of the predicted values ​​of key performance indicators, and the goal is to minimize this variance. The constraints include physical limitations of environmental parameters (such as the reasonable range of temperature and humidity) and equipment operation limitations. The NSGA-II (Non-dominated Sorting Genetic Algorithm II) algorithm is used to solve this multi-objective optimization problem. The population size is set to 100 and the evolutionary generations are 500. The chromosome encoding uses real number encoding to represent the adjusted values ​​of environmental parameters. The crossover operation uses simulated binary crossover (SBX) with a crossover probability of 0.9. The mutation operation uses polynomial mutation with a mutation probability of 1 / n (n is the number of decision variables). The crowding comparison operator is used to maintain population diversity. In each generation of evolution, the previously trained LSTM model and ARIMA model are used to predict the fault indicators and performance indicators under the current environmental parameters and calculate the objective function value. Finally, a Pareto optimal solution set is obtained, which represents the environmental parameter adjustment scheme under different trade-offs. From this solution set, a solution that balances the fault risk and performance stability is selected as the preliminary environmental parameter adjustment scheme, which includes the target value and adjustment step size of each environmental parameter.

[0044] Specifically, the deep deterministic policy gradient (DDPG) algorithm is used to optimize the preliminary environmental parameter adjustment scheme using the reinforcement learning algorithm. First, a virtual environment simulator is built, and the previously trained LSTM model and ARIMA model are used to simulate the response of the LCM module under different environmental parameters. The state space includes the current environmental parameters and the predicted fault indicator values, and the action space is the adjustment amount of the environmental parameters. The reward function is designed as the weighted sum of the negative value of the fault risk and the performance stability. The Actor network and the Critic network of DDPG both adopt a 4-layer fully connected structure, with 400 and 300 hidden layer units respectively, and the ReLU activation function is used. The Ornstein-Uhlenbeck process is used to generate exploration noise, with the initial value ε set to 0.3 and the decay rate set to 0.9999. The experience replay buffer size is set to 100000 and the batch size is 64. The soft update parameter τ is set to 0.001. The training process is carried out for 10,000 episodes, and the maximum number of steps per episode is 1000. In each step, the agent selects an action (environmental parameter adjustment) based on the current strategy, and the environment returns a new state and reward. The agent updates the experience replay buffer and randomly samples from it for learning. The learning process uses the Adam optimizer with a learning rate of 0.0001. After training, the trained Actor network is used to generate the final environmental parameter adjustment strategy. This strategy can adjust environmental parameters in real time according to the current state to balance the risk of failure and performance stability. Finally, the optimized environmental parameter combination is output, including the target value, adjustment range and dynamic adjustment strategy of each parameter.

[0045] Specifically, when the future trend prediction results and the optimized environmental parameters are combined and input into the pre-trained fault evolution model, a fault evolution model based on a graph neural network (GNN) is first constructed. The nodes of the graph represent fault indicators and environmental factors, and the edges represent the dynamic associations between them. The Graph Attention Network (GAT) is used as the basic structure, which contains 3 layers of GAT layers, 8 attention heads in each layer, and the hidden layer dimension is 64. The model pre-training uses historical data and adopts the teacherforcing method. The training goal is to minimize the mean square error between the predicted value and the actual value. After the pre-training is completed, the Monte Carlo method is used to generate multiple possible fault development paths. The specific steps are as follows: First, the LSTM predicted 48-hour fault indicator trend and the optimized environmental parameter combination are used as the initial input. Then, the following process is repeated 1000 times: At each time step, the fault indicator value is sampled from the predicted distribution, and the sampled value and the current environmental parameters are input into the GNN model to predict the state of the next time step. At the same time, the environmental parameters are adjusted according to the strategy obtained by reinforcement learning. This process continues until the preset time horizon (such as 30 days) is reached. Each path records the changes of fault indicators and environmental parameters over time. Finally, all paths are statistically analyzed to calculate the mean, variance, and quantile of the fault indicators at each time point to form a probability-weighted predictive fault trend. The output includes a time series forecast of each fault indicator (including the mean and 95% confidence interval) and the probability distribution of fault occurrence at different time points. This Monte Carlo simulation-based method can capture the uncertainty of the fault evolution process and provide more comprehensive information support for decision-making.

[0046] 104. Based on the fault characteristics, cause analysis, predictive fault trends and optimized environmental parameter combinations, a comprehensive performance evaluation is conducted on the target LCM module to obtain quantitative performance evaluation results.

[0047] In one embodiment of the present invention, the comprehensive performance evaluation of the target LCM module is performed according to the fault characteristics, cause analysis, predictive fault trends and optimized environmental parameter combination to obtain a quantitative performance evaluation result, including: determining the key aging indicators of the LCM module to be tested according to the fault characteristics and cause analysis, and obtaining an aging test evaluation indicator set; using the optimized environmental parameter combination to perform an accelerated aging test on the LCM module to be tested, recording the changes of various aging indicators over time, and obtaining aging curve data; based on the predictive fault trend and aging curve data, predicting the aging rate of the LCM module to be tested under normal use conditions through a data extrapolation method, and obtaining an aging life estimate; comprehensively considering the aging test evaluation indicator set, aging curve data and aging life estimate, calculating the comprehensive aging performance score of the LCM module to be tested through a weighted scoring method, and obtaining a quantitative performance evaluation result.

[0048] Specifically, when determining the key aging indicators of the LCM module to be tested based on the fault characteristics and cause analysis, the fault feature vector is first subjected to principal component analysis (PCA) to retain the principal components that explain 90% of the variance. Then, the mapping relationship between these principal components and the original features is used to identify the original features that contribute most to the principal components. These features are regarded as potential key aging indicators. Next, text mining is performed on the fault cause analysis report, and keywords are extracted using the TF-IDF algorithm. They are matched with the predefined aging indicator list in the expert knowledge base to find out the frequently occurring aging-related indicators. The potential indicators obtained by PCA and the indicators obtained by text mining are merged to form a preliminary candidate set of aging indicators. Then, the random forest algorithm is used to sort the importance of these candidate indicators and select the top 20 indicators with the highest importance scores. In order to ensure the comprehensiveness of the indicators, these 20 indicators are classified according to their physical meanings (such as optical performance, electrical performance, structural stability, etc.), ensuring that each category contains at least 2 indicators. If a certain category of indicators is insufficient, it is supplemented from the less important indicators of that category. Finally, the selected indicator set is compared with industry standards (such as JEDEC standards) to ensure that all required test items are included. The resulting aging test evaluation indicator set usually contains 15-25 indicators, each of which has a clear definition, measurement method and unit, and is stored in a structured JSON format for subsequent processing and analysis.

[0049] When using the optimized environmental parameter combination to perform accelerated aging tests on the LCM module to be tested, the environmental simulation chamber is first set up according to the environmental parameter combination obtained by the optimization algorithm. Environmental parameters include temperature cycles (such as -10°C to 60°C, 4 hours of cycle), humidity changes (such as 20% to 90%, co-varied with temperature), light intensity (such as 0 to 1000 lux, 12 hours of light and dark cycle) and electromagnetic interference intensity (such as 0 to 10V / m, frequency scanning from 50Hz to 1GHz). The test lasts for 1000 hours, during which all aging test evaluation indicators are fully measured every 1 hour. For optical performance indicators (such as brightness, contrast, color coordinates), a calibrated spectrophotometer is used to measure at 5 fixed points (center and four corners). Electrical performance indicators (such as power consumption, response time) are measured by high-precision digital multimeters and high-speed oscilloscopes. Structural stability indicators (such as surface deformation, film peeling degree) are observed and recorded using 3D scanners and microscopes. The data of each measurement is recorded in the time series database together with the timestamp and current environmental parameters. At the same time, an interpolation algorithm is used to fill in missing data to ensure data continuity. For each indicator, its rate of change over time is calculated, and the sliding window method (window size is 24 hours) is used for smoothing to obtain a more stable change trend. Finally, a multidimensional time series data set is obtained, where each dimension corresponds to an aging indicator, and the data point contains a timestamp, a measurement value, and the corresponding environmental parameters. This data set is the aging curve data, which comprehensively records the degradation process of various performance indicators of the LCM module under accelerated aging conditions.

[0050] When predicting the aging rate of the LCM module under normal use conditions based on the predictive failure trend and aging curve data, it is necessary to first establish an acceleration factor model. The Arrhenius equation is used to deal with the temperature acceleration effect, the Peck model is used to deal with the humidity acceleration effect, the inverse square relationship is used to deal with the light intensity acceleration effect, and the linear superposition model is used to deal with the electromagnetic interference acceleration effect. The parameters of these models are obtained by fitting the aging curve data by the least squares method. Then, for each aging indicator, the generalized accelerated life model (GALM) is used to comprehensively consider the influence of multiple stress factors. The parameter estimation of the GALM model adopts the maximum likelihood estimation method and is optimized using the L-BFGS-B algorithm. Next, the predicted value of the environmental factor in the predictive failure trend is substituted into the GALM model to obtain the degradation rate of each indicator under normal use conditions. For the nonlinear degradation process, the Wiener process model is used for modeling, and the model parameters are estimated by the particle filter algorithm. Then, the Monte Carlo simulation method is used to generate 10,000 possible degradation paths, and each path is simulated until the indicator reaches the predefined failure threshold. The 10,000 paths were statistically analyzed to calculate the average failure time and its 95% confidence interval, which is the estimated aging life of the indicator. Finally, considering the different importance of different indicators, the analytic hierarchy process (AHP) was used to determine the weight of each indicator, and the comprehensive aging life estimate of the LCM module was obtained by weighted average calculation. This estimate is in hours and comes with a 95% confidence interval, reflecting the uncertainty of the prediction.

[0051] Considering the aging test evaluation index set, aging curve data and aging life estimation, when calculating the comprehensive aging performance score of the LCM module to be tested by the weighted scoring method, each index needs to be standardized first. For the aging curve data, the percentage change of each index relative to the initial value at the end of the test is calculated, and it is mapped to a score range of 0-100 using min-max standardization, where 0 represents the maximum degree of degradation and 100 represents no degradation. For the aging life estimation, it is compared with the expected life (such as 50,000 hours), and the full score of 100 is obtained if it exceeds the expected life, otherwise it is scored proportionally. Then, the weight of each index is determined using the fuzzy comprehensive evaluation method. Construct a judgment matrix, fill the matrix elements using the expert scoring method, calculate the weight vector using the eigenvalue method, and perform a consistency test (consistency ratio CR<0.1 is valid). Then, the standardized index score is multiplied by the corresponding weight to obtain the weighted score. In order to consider the mutual influence between indicators, the Choquet fuzzy integral is introduced, the fuzzy measure is constructed, and the interaction of each indicator is calculated using the Shapley value method. The final comprehensive aging performance score is calculated by Choquet integral, with a score range of 0-100. In addition, the confidence interval of the score is estimated using the Bootstrap method, and the sampling is repeated 1000 times to calculate the 2.5% and 97.5% quantiles as the 95% confidence interval. Finally, the aging performance of the LCM module is divided into five levels according to the score: excellent (90-100), good (80-89), general (70-79), poor (60-69) and unqualified (<60). The output quantitative performance evaluation results include the total score, the sub-item score of each indicator, the comprehensive grade assessment and the confidence interval, which are presented in a structured JSON format to facilitate subsequent data analysis and report generation.

[0052] In this embodiment, the initial state data and initial environmental parameter combination of the LCM module are obtained through intelligent environmental simulation and initialization processing, and the module is subjected to multi-mode alternating aging test and real-time monitoring processing using an intelligent environmental simulation system to obtain performance data in a dynamic environment, perform periodic surface scanning and work sound recording, and obtain microstructure data and acoustic data. Through intelligent fault detection and in-depth analysis processing, the fault characteristics and causes of the module are identified, and predictive analysis of the fault is performed in combination with a machine learning algorithm to generate predictive fault trends and optimized environmental parameter combinations. A comprehensive performance evaluation of the module is performed based on fault characteristics, cause analysis, predicted trends, and optimized parameters. The present invention achieves accurate performance evaluation and fault prediction of the LCM module by combining intelligent environmental simulation, real-time monitoring, and in-depth analysis.

[0053] The above describes the LCM module aging test method in the embodiment of the present invention. The following describes the LCM module aging test device in the embodiment of the present invention. Figure 2 In one embodiment of the present invention, an LCM module aging test device includes: The environment simulation module 201 is used to perform intelligent environment simulation and initialization processing on the target LCM module, obtain the initial state data and initial environment parameter combination of the target LCM module, and initialize the intelligent environment simulation system according to a preset dynamic environment adjustment algorithm; The aging test module 202 is used to use the intelligent environment simulation system to perform multi-mode alternating aging tests and real-time monitoring processing on the target LCM module according to the initial state data and the initial environment parameter combination, obtain the performance data of the target LCM module under the dynamically adjusted environment conditions, and periodically scan the surface of the target LCM module and record the working sound to obtain microstructure data and acoustic data; A fault analysis module 203 is used to perform intelligent fault detection and in-depth analysis on the performance data, microstructure data and acoustic data to obtain fault characteristics and cause analysis of the target LCM module, and use a machine learning algorithm to perform predictive analysis on the fault characteristics and cause analysis to obtain predictive fault trends and optimized environmental parameter combinations; The performance evaluation module 204 is used to perform a comprehensive performance evaluation on the target LCM module according to the fault characteristics, cause analysis, predictive fault trend and optimized environmental parameter combination to obtain a quantitative performance evaluation result.

[0054] In an embodiment of the present invention, the LCM module aging test device runs the above-mentioned LCM module aging test method. The LCM module aging test device obtains the initial state data and initial environmental parameter combination of the LCM module through intelligent environmental simulation and initialization processing, and uses the intelligent environmental simulation system to perform multi-mode alternating aging testing and real-time monitoring processing on the module, obtains performance data under a dynamic environment, performs periodic surface scanning and working sound recording, and obtains microstructure data and acoustic data. Through intelligent fault detection and in-depth analysis processing, the fault characteristics and causes of the module are identified, and the fault is predictively analyzed in combination with a machine learning algorithm to generate a predictive fault trend and an optimized environmental parameter combination. According to the fault characteristics, cause analysis, prediction trend and optimization parameters, a comprehensive performance evaluation of the module is performed. The present invention realizes accurate performance evaluation and fault prediction of the LCM module by combining intelligent environmental simulation, real-time monitoring and in-depth analysis.

[0055] above Figure 2The LCM module aging test device in the embodiment of the present invention is described in detail from the perspective of modular functional entities. The LCM module aging test equipment in the embodiment of the present invention is described in detail from the perspective of hardware processing.

[0056] Figure 3 3 is a schematic diagram of the structure of an LCM module aging test device provided by an embodiment of the present invention. The LCM module aging test device 300 may have relatively large differences due to different configurations or performances, and may include one or more processors (central processing units, CPU) 310 (for example, one or more processors) and a memory 320, and one or more storage media 330 (for example, one or more mass storage device terminals) storing application programs 333 or data 332. Among them, the memory 320 and the storage medium 330 can be short-term storage or permanent storage. The program stored in the storage medium 330 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations in the LCM module aging test device 300. Furthermore, the processor 310 may be configured to communicate with the storage medium 330, and execute a series of instruction operations in the storage medium 330 on the LCM module aging test device 300 to implement the steps of the above-mentioned LCM module aging test method.

[0057] The LCM module aging test device 300 may also include one or more power supplies 340, one or more wired or wireless network interfaces 350, one or more input and output interfaces 360, and / or one or more operating systems 331, such as Windows Serve, Mac OS X, Unix, Linux, FreeBSD, etc. Those skilled in the art will appreciate that Figure 3 The structure of the LCM module aging test equipment shown does not constitute a limitation on the LCM module aging test equipment provided by the present invention, and may include more or fewer components than shown in the figure, or a combination of certain components, or a different arrangement of components.

[0058] The present invention also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. Instructions are stored in the computer-readable storage medium. When the instructions are executed on a computer, the computer executes the steps of the LCM module aging test method.

[0059] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the above-described system, device, or unit can refer to the corresponding process in the aforementioned method embodiment and will not be repeated here.

[0060] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or the whole or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk and other media that can store program code.

[0061] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features thereof may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for testing LCM module aging, characterized in that: The LCM module aging test method comprises: Perform intelligent environment simulation and initialization processing on the target LCM module to obtain the initial state data and initial environment parameter combination of the target LCM module, and initialize the intelligent environment simulation system according to the preset dynamic environment adjustment algorithm; Using the intelligent environment simulation system, a multi-mode alternating aging test and real-time monitoring process are performed on the target LCM module according to the initial state data and the initial environment parameter combination to obtain the performance data of the target LCM module under the dynamically adjusted environment conditions, and the surface of the target LCM module is periodically scanned and the working sound is recorded to obtain microstructure data and acoustic data; Perform intelligent fault detection and in-depth analysis on the performance data, microstructure data and acoustic data to obtain fault characteristics and cause analysis of the target LCM module, and use a machine learning algorithm to perform predictive analysis on the fault characteristics and cause analysis to obtain predictive fault trends and optimized environmental parameter combinations; According to the fault characteristics, cause analysis, predictive fault trends and optimized environmental parameter combinations, a comprehensive performance evaluation is performed on the target LCM module to obtain a quantitative performance evaluation result.

2. The LCM module aging test method according to claim 1, characterized in that: The intelligent environment simulation and initialization processing is performed on the target LCM module to obtain the initial state data and initial environment parameter combination of the target LCM module, and the intelligent environment simulation system is initialized according to the preset dynamic environment adjustment algorithm, including: Performing a pre-aging process on the target LCM module to obtain an initial response characteristic curve of the target LCM module, and generating initial state data according to the initial response characteristic curve; The environmental simulation chamber is parameterized according to the initial state data to obtain an initial environmental parameter combination, and a high-precision optical scanning system is used to perform an initial scan on the surface of the target LCM module to obtain initial microstructure data; The initial response characteristic curve, the initial environmental parameter combination and the initial microstructure data are input into a preset dynamic environment adjustment algorithm, and the intelligent environment simulation system is initialized according to the output initial adjustment strategy.

3. The LCM module aging test method according to claim 1, characterized in that: The intelligent environment simulation system is used to perform multi-mode alternating aging test and real-time monitoring processing on the target LCM module according to the initial state data and the initial environment parameter combination, to obtain the performance data of the target LCM module under the dynamically adjusted environment conditions, and to periodically scan the surface of the target LCM module and record the working sound, to obtain the microstructure data and acoustic data including: According to the preset multi-mode test sequence, the target LCM module is tested for static image display, dynamic video playback and grayscale gradient to obtain display performance data in different modes; Use the intelligent environmental simulation system to dynamically adjust temperature, humidity, light intensity and electromagnetic interference intensity to obtain the response data of the target LCM module under different environmental conditions; A high-precision optical scanning system is used to periodically scan the surface of the target LCM module to obtain microstructure change data, and a high-sensitivity microphone array is used to record the acoustic characteristics of the target LCM module under different working conditions to obtain acoustic data.

4. The LCM module aging test method according to claim 1, characterized in that: The intelligent fault detection and in-depth analysis of the performance data, microstructure data and acoustic data are performed to obtain the fault characteristics and cause analysis of the target LCM module, and the machine learning algorithm is used to perform predictive analysis on the fault characteristics and cause analysis to obtain the predictive fault trend and the optimized environmental parameter combination, including: Perform multi-dimensional feature extraction on performance data, microstructure data and acoustic data to obtain the comprehensive feature vector of the target LCM module; Analyzing the comprehensive feature vector using a pre-trained fault diagnosis model to obtain the fault characteristics and fault degree assessment results of the target LCM module; According to the fault characteristics and fault degree assessment results, combined with similar cases in the historical database, the fault cause is traced and analyzed to obtain a fault cause analysis report; A machine learning algorithm is used to perform predictive analysis on the fault characteristics and cause analysis to obtain a predictive fault trend and an optimized environmental parameter combination.

5. The LCM module aging test method according to claim 4, characterized in that: According to the fault characteristics and fault degree assessment results, combined with similar cases in the historical database, the fault cause traceability analysis is performed to obtain a fault cause analysis report including: Performing feature vectorization processing on the fault characteristics and the fault degree assessment results to obtain a fault feature vector; Using the fault feature vector to perform similarity matching in the historical database to obtain a preliminary matching case set, and sorting the case set by similarity based on weighted Euclidean distance to obtain N most similar cases, where N is a natural number greater than 1; Perform multi-dimensional cluster analysis on the N most similar cases to obtain the main fault types and key influencing factors; According to the main fault types and key influencing factors, combined with the initial state data and initial environmental parameter combination of the target LCM module, a multi-layer causal relationship graph is constructed, and the multi-layer causal relationship graph is analyzed by a graph traversal algorithm to obtain the fault cause chain; Based on the fault cause chain and historical solutions, the decision tree algorithm is used to generate fault diagnosis and solution recommendations to obtain a fault cause analysis report.

6. The LCM module aging test method according to claim 4, characterized in that: The use of a machine learning algorithm to perform predictive analysis on the fault characteristics and cause analysis to obtain a predictive fault trend and an optimized environmental parameter combination includes: Perform text mining and feature extraction on fault characteristics and cause analysis reports to obtain key fault indicators and environmental factors; Using a long short-term memory network to perform time series analysis on historical data of key fault indicators and environmental factors, and obtain future trend prediction results of the key fault indicators and change patterns of the environmental factors; The time-lag correlation between fault indicators and environmental factors is evaluated by time series correlation analysis method to obtain dynamic correlation characteristics; Based on the future trend prediction results of fault indicators, the change patterns and dynamic correlation characteristics of environmental factors, a multi-objective optimization problem is constructed, and the multi-objective optimization problem is solved with the goal of minimizing the risk of faults and maximizing the performance stability, and a preliminary environmental parameter adjustment plan is obtained; Using the reinforcement learning algorithm, based on the preliminary environmental parameter adjustment plan, the optimized environmental parameter combination is obtained through virtual environment simulation and iterative optimization; The future trend prediction result and the optimized environmental parameters are combined and input into a pre-trained fault evolution model, and a plurality of possible fault development paths are generated through Monte Carlo simulation to obtain a probability-weighted predictive fault trend.

7. The LCM module aging test method according to claim 1, characterized in that: The comprehensive performance evaluation of the target LCM module is performed according to the fault characteristics, cause analysis, predictive fault trend and optimized environmental parameter combination, and the quantitative performance evaluation results obtained include: According to the fault characteristics and cause analysis, the key aging indicators of the LCM module to be tested are determined, and the aging test evaluation indicator set is obtained; Using the optimized environmental parameter combination, the LCM module to be tested is subjected to accelerated aging test, and the changes of various aging indicators over time are recorded to obtain aging curve data; Based on the predictive failure trend and aging curve data, the aging rate of the LCM module under normal use conditions is predicted by data extrapolation method to obtain the estimated aging life; Taking into account the aging test evaluation index set, aging curve data and aging life estimation, the comprehensive aging performance score of the LCM module to be tested is calculated by the weighted scoring method to obtain a quantitative performance evaluation result.

8. An LCM module aging test device, characterized in that: The LCM module aging test device comprises: The environment simulation module is used to perform intelligent environment simulation and initialization processing on the target LCM module, obtain the initial state data and initial environment parameter combination of the target LCM module, and initialize the intelligent environment simulation system according to the preset dynamic environment adjustment algorithm; An aging test module is used to use the intelligent environment simulation system to perform multi-mode alternating aging tests and real-time monitoring processing on the target LCM module according to the initial state data and the initial environment parameter combination, to obtain the performance data of the target LCM module under the dynamically adjusted environment conditions, and to periodically scan the surface of the target LCM module and record the working sound to obtain microstructure data and acoustic data; A fault analysis module is used to perform intelligent fault detection and in-depth analysis and processing on the performance data, microstructure data and acoustic data to obtain the fault characteristics and cause analysis of the target LCM module, and use a machine learning algorithm to perform predictive analysis on the fault characteristics and cause analysis to obtain predictive fault trends and optimized environmental parameter combinations; The performance evaluation module is used to perform a comprehensive performance evaluation on the target LCM module according to the fault characteristics, cause analysis, predictive fault trends and optimized environmental parameter combinations to obtain a quantitative performance evaluation result.

9. An LCM module aging test device, characterized in that: The LCM module aging test device comprises: a memory and at least one processor, wherein instructions are stored in the memory; The at least one processor calls the instructions in the memory so that the LCM module aging test device executes the steps of the LCM module aging test method as described in any one of claims 1-7.

10. A computer-readable storage medium having instructions stored thereon, characterized in that: When the instructions are executed by the processor, the steps of the LCM module aging test method as described in any one of claims 1-7 are implemented.

Images