A method and system for monitoring performance degradation of a microelectronic device

Abstract

The application belongs to the technical field of microelectronic monitoring, and particularly relates to a microelectronic device performance degradation monitoring method and system, which comprises the following steps: obtaining a strain signal, an environmental parameter and a functional test result of a microelectronic device, correcting and grid processing the strain signal to obtain a deformation distribution matrix; performing frequency domain analysis on the deformation distribution matrix and separating thermal response characteristics and load response characteristics to obtain a deformation source result; determining a damage accumulation result according to the deformation source result and the deformation distribution matrix; determining a degradation risk score in combination with the environmental parameter, historical degradation data and the functional test result, and determining a degradation evaluation result according to the degradation risk score. The application can realize deformation source differentiation, damage accumulation identification and performance influence quantitative evaluation.

Description

Technical Field

[0001] This application belongs to the field of microelectronic monitoring technology, specifically a method and system for monitoring the performance degradation of microelectronic devices. Background Technology

[0002] Microelectronic devices are fundamental units in electronic equipment, communication equipment, industrial control equipment, and smart terminals. Their operating status directly affects the stability, lifespan consistency, and maintenance costs of the entire system. During long-term operation, devices are affected by factors such as temperature changes, external loads, packaging stress, and connection fatigue, gradually leading to localized deformation, structural damage, and performance drift. Existing technologies include some solutions that focus on external environmental monitoring, making it difficult to reflect changes in the internal state of the device; others focus on functional testing, which can detect performance anomalies but struggles to trace the source and evolution of these anomalies; still others focus only on a single physical quantity, lacking corresponding analysis between it and target performance indicators. Therefore, it is difficult to achieve continuous judgment and reliable assessment of degradation states, easily leading to delayed warnings, increased misjudgments, and insufficient maintenance basis. Summary of the Invention

[0003] To address the above problems, this application provides a method and system for monitoring the performance degradation of microelectronic devices, which at least solves the problem of how to establish a correspondence between the deformation changes and environmental changes of microelectronic devices and the functional test results and output reliable degradation assessment results.

[0004] To achieve the above objectives, the technical solution adopted in this application is as follows: In a first aspect, this application provides a method for monitoring the performance degradation of microelectronic devices, the method comprising: Acquire strain signals, environmental parameters, and functional test results of microelectronic devices, with the functional test results corresponding to target performance indicators; The strain signal is corrected and meshed to obtain the deformation distribution matrix; Frequency domain analysis was performed on the deformation distribution matrix, and thermal response characteristics and load response characteristics were separated based on the calibration frequency band to obtain the deformation source results; The damage accumulation result is determined based on the deformation source result and the deformation distribution matrix; The degradation risk score is determined based on the cumulative damage results, environmental parameters, historical degradation data and functional test results. The degradation risk score is used to characterize the degree of influence of the deformation corresponding to the deformation distribution matrix on the target performance index. The degradation assessment result is determined based on the degradation risk score.

[0005] In one possible implementation, the strain signal is corrected and meshed, including: performing temperature drift correction on the strain signal based on the temperature parameter in the environmental parameters to obtain the corrected strain signal; performing spatial interpolation on the corrected strain signal to obtain meshed deformation data; and determining the deformation distribution matrix based on the meshed deformation data.

[0006] In one possible implementation, the calibration frequency band includes a thermal response frequency band and a load response frequency band; frequency domain analysis is performed on the deformation distribution matrix, and thermal response characteristics and load response characteristics are separated based on the calibration frequency band, including: performing a Fourier transform on the deformation distribution matrix to obtain spectral data; and determining the deformation source based on the energy distribution of the spectral data within the thermal response frequency band and the load response frequency band.

[0007] In one possible implementation, the deformation source results include thermally dominated deformation, load-dominated deformation, and thermally-load-coupled deformation; thermally-load-coupled deformation is used to indicate that the energy values ​​of the spectral data in both the thermal response band and the load response band are higher than the corresponding preset thresholds.

[0008] In one possible implementation, the damage accumulation result is determined based on the deformation source result and the deformation distribution matrix, including: when the deformation source result is thermally dominated deformation or thermal load coupled deformation, generating a spectrum map based on the spectrum data, using a convolutional neural network to identify the spectrum map, and obtaining the damage area and the degree of thermal damage accumulation; and determining the damage accumulation result based on the damage area and the degree of thermal damage accumulation.

[0009] In one possible implementation, a degradation risk score is determined based on cumulative damage results, environmental parameters, historical degradation data, and functional test results. This includes: using a random forest model trained on historical degradation data, inputting the cumulative damage results, environmental parameters, and functional test results into the random forest model to obtain the degradation risk score.

[0010] In one possible implementation, historical degradation data includes historical damage accumulation results, historical environmental parameters, and historical performance offsets. The historical performance offsets indicate the degree of deviation of historical functional test results from the target performance index benchmark value, which is a reference value for the target performance index under baseline operating conditions. The damage accumulation results, environmental parameters, and functional test results are input into a random forest model, including: determining the current performance offset based on the functional test results and the target performance index benchmark value; and inputting the damage accumulation results, environmental parameters, and current performance offset into the random forest model to obtain a degradation risk score.

[0011] In one possible implementation, the degradation assessment result is determined based on the degradation risk score, including: constructing a state discrimination vector based on the degradation risk score and environmental parameters; and inputting the state discrimination vector into a support vector machine classifier to obtain the degradation assessment result.

[0012] In one possible implementation, the method of this application further includes: outputting alarm information when the degradation assessment result indicates abnormal degradation; and recording the deformation distribution matrix, damage accumulation result and degradation risk score corresponding to the alarm information.

[0013] Secondly, this application provides a microelectronic device performance degradation monitoring system for implementing a microelectronic device performance degradation monitoring method. The system includes: The signal acquisition module is used to acquire strain signals, environmental parameters, and functional test results of microelectronic devices. The functional test results correspond to the target performance indicators. The matrix construction module is used to correct and mesh the strain signal to obtain the deformation distribution matrix; The source analysis module is used to perform frequency domain analysis on the deformation distribution matrix and separate thermal response characteristics and load response characteristics based on the calibration frequency band to obtain the deformation source results; The damage determination module is used to determine the cumulative damage result based on the deformation source result and the deformation distribution matrix. The risk assessment module is used to determine the degradation risk score based on the cumulative damage results, environmental parameters, historical degradation data and functional test results. The degradation risk score is used to characterize the degree of influence of the deformation corresponding to the deformation distribution matrix on the target performance index. The results output module is used to determine the degradation assessment results based on the degradation risk score.

[0014] Compared with existing technologies, the advantages and beneficial effects of this application are as follows: By acquiring strain signals, environmental parameters, and functional test results and establishing a unified data correspondence, the synchronous characterization of device structural and performance states was achieved, solving the problem of separation between physical and functional states in existing solutions. Through strain signal correction and meshing, the discrete sampling results were transformed into continuous deformation characterization, improving data consistency in subsequent analyses. Frequency domain analysis of the deformation distribution matrix and separation of thermal and load response characteristics enabled the differentiation of different causes, resolving the problem of unclear anomaly sources. Combining deformation source results with the deformation distribution matrix to determine damage accumulation results achieved a gradual mapping from deformation changes to damage states. Introducing historical degradation data and functional test results to determine degradation risk scores enabled a quantitative expression of the impact of deformation changes on target performance indicators. Outputting degradation assessment results based on degradation risk scores enabled graded judgments oriented towards actual operating conditions, providing a more direct basis for maintenance decisions and condition management. Attached Figure Description

[0015] Figure 1 This is a flowchart illustrating the method described in this application; Figure 2 This is a block diagram of the module composition of the system in this application. Detailed Implementation

[0016] To enable those skilled in the art to better understand the technical solution, the present application will be described in detail below with reference to the embodiments. The description in this section is only exemplary and explanatory, and should not be used to limit the scope of protection of the present application in any way.

[0017] Performance degradation monitoring refers to the technical process of continuously acquiring, correlating, and determining the stages of physical and functional responses reflecting changes in device state throughout the entire operation process. Its focus is not on the discovery of single anomalies, but on identifying the continuous trajectory of a device's evolution from a normal state to an unstable state. For microelectronic devices, performance degradation is usually not sudden, but rather a gradual accumulation under the combined influence of thermal effects, load disturbances, and changes in internal connection states, ultimately manifesting as deviations, fluctuations, or instability in target performance indicators. Therefore, performance degradation monitoring needs to simultaneously consider the correspondence between deformation information, environmental information, and functional test information, and on this basis, establish a comprehensive judgment mechanism for the degree, source, and trend of degradation. Based on this, this application proposes a microelectronic device performance degradation monitoring method that achieves staged identification and evaluation of device degradation states through joint processing of strain signals, environmental parameters, and functional test results.

[0018] like Figure 1 As shown, a method for monitoring the performance degradation of microelectronic devices includes: Acquire strain signals, environmental parameters, and functional test results of microelectronic devices, with the functional test results corresponding to target performance indicators; In this embodiment, after the microelectronic device enters the monitoring state, a unified acquisition window is first established. Within the same acquisition window, strain signals, environmental parameters, and functional test results are acquired. The strain signal characterizes the deformation changes of the microelectronic device during operation, the environmental parameters characterize the external conditions affecting the formation and expansion of deformation, and the functional test results characterize the current functional state of the microelectronic device. A pre-established correspondence is established between the functional test results and target performance indicators, forming a data set containing deformation information, environmental information, and performance information within the same acquisition window. After the data set is constructed, the data set within the current acquisition window is written into the monitoring buffer and then transferred to the subsequent calibration and meshing processing stages.

[0019] In one embodiment, this stage, based on the aforementioned synchronous acquisition, further defines the arrangement of the acquisition objects, the selection method of the target performance indicators, the matching rules of the functional test results, and the processing flow of abnormal acquisition windows. Specifically, strain signals are acquired through multiple strain acquisition points deployed on the surface of the microelectronic device package. These strain acquisition points are preferentially located in areas of high heat concentration, pin connection areas, solder transition areas, and package edges, ensuring that the acquisition results cover locations prone to local deformation accumulation. Environmental parameters include at least temperature and load parameters. Temperature parameters are acquired through temperature acquisition elements adjacent to the package surface, and load parameters are acquired through pressure sensors or vibration sensors at the installation location. Functional test results are generated by the functional test unit according to a preset test cycle. The functional test unit calls the corresponding test items based on the type of microelectronic device. For high-speed signal microelectronic devices, the target performance indicators can be set as propagation delay, output waveform deviation, or bit error rate; for storage microelectronic devices, the target performance indicators can be set as read / write error rate, hold state change, or access response time; for power microelectronic devices, the target performance indicators can be set as leakage current, conduction stability, or switching response deviation.

[0020] The target performance indicators are not arbitrarily specified, but are predetermined based on the main failure behaviors of the microelectronic devices in the target application and written into the device configuration table. The acquisition control unit uses a unified clock to schedule each acquisition channel, ensuring that strain signals, environmental parameters, and functional test results are time-aligned within the same acquisition window. The length of the acquisition window is set based on the thermal response rate of the microelectronic device and the external load fluctuation period. When the temperature changes slowly and the load is stable, the acquisition window is appropriately extended; when vibration is significant or the load changes frequently, the acquisition window is appropriately shortened. After each window acquisition is completed, the strain signals, environmental parameters, and functional test results within that window are merged according to time tags to generate the data group corresponding to the current window, and the correspondence between the functional test results and the target performance indicators is written into the monitoring buffer. If any acquisition channel in the current window is missing data, it is first determined whether it is a momentary communication anomaly; if re-acquisition is possible, a re-acquisition is performed; if the data is still incomplete after re-acquisition, the current window is marked as an invalid window and not directly sent to the subsequent processing stage. If there is no existing correspondence between the functional test results and the target performance indicators, the device configuration table is first called to complete the indicator mapping before generating the data group. By using the above processing method, it can be ensured that the data entering the deformation analysis link subsequently includes both strain information characterizing the structural state and test information characterizing the functional state, thus providing a consistent data foundation for establishing the correspondence between deformation changes and performance degradation.

[0021] The strain signal is corrected and meshed to obtain the deformation distribution matrix; In this embodiment, after synchronously acquiring strain signals, environmental parameters, and functional test results, the strain signals within the current acquisition window are calibrated and gridded. First, temperature parameters are extracted from the environmental parameters, and combined with the calibration information stored by the microelectronic device at a reference temperature, the original strain values ​​output at each strain acquisition point are corrected point-by-point to eliminate additional drift caused by temperature changes, resulting in a calibrated strain signal. Then, based on the spatial relationship of each strain acquisition point on the device surface, spatial interpolation is performed on the calibrated strain signal, converting the discrete sampling results into a continuous distribution, forming gridded deformation data. Next, the gridded deformation data is arranged according to a pre-defined row and column mapping rule to obtain a deformation distribution matrix corresponding to the surface region of the microelectronic device, and this deformation distribution matrix is ​​then sent to the subsequent frequency domain analysis stage.

[0022] The strain signal is corrected and gridded, including: temperature drift correction of the strain signal based on the temperature parameter in the environmental parameters to obtain the corrected strain signal; spatial interpolation of the corrected strain signal to obtain gridded deformation data; and determination of the deformation distribution matrix based on the gridded deformation data.

[0023] In one embodiment, this stage further defines the benchmark establishment method, spatial interpolation boundary constraints, matrix construction rules, and abnormal data handling method based on the above processing flow. Specifically, before the microelectronic device enters formal monitoring, a benchmark acquisition is performed first. When the ambient temperature is stable and the device is in normal working condition, the initial output value of each strain acquisition point is recorded, and the initial output value, the corresponding acquisition point position, and the current benchmark temperature are written into the calibration table. Subsequently, each time a new acquisition window is entered, the processing unit first reads the temperature parameter in the current window and determines the deviation range between the current temperature and the benchmark temperature according to the calibration table. If the temperature deviation is within the preset compensation range, the drift compensation parameter of the corresponding acquisition point is called to correct the original strain value point by point to obtain the corrected strain signal; if the temperature deviation exceeds the compensation range, a repeated acquisition is performed first, and the change amplitude of the two acquisition results is compared. When the change amplitude still exceeds the allowable range, the current window is marked as a window to be reviewed and does not directly enter the matrix construction process. After the drift correction is completed, the correction results of each acquisition point and the corresponding spatial coordinates are sent to the interpolation processing unit.

[0024] The interpolation processing unit establishes a regular grid based on the actual dimensions of the microelectronic device package surface. The grid density is set according to the spacing of the acquisition points, ensuring that the distance between adjacent grid nodes is no greater than half the average spacing between adjacent acquisition points. For grid nodes located within the effective acquisition area, a weighted interpolation method based on neighboring acquisition points is used to determine the node deformation. For grid nodes located at the package edge or in sparse acquisition point areas, a neighborhood smoothing method is used to determine the node values, avoiding abrupt gaps in edge regions. After obtaining the deformation values ​​of all grid nodes, the node values ​​are written into the two-dimensional data area in a top-down, left-to-right order to form a deformation distribution matrix.

[0025] Each row in the deformation distribution matrix corresponds to a horizontal position band on the package surface, and each column corresponds to a vertical position band on the package surface. Matrix elements are used to characterize the deformation state of the corresponding position band within the current acquisition window. The missing number threshold is preset based on the total number of nodes in the matrix, the average spacing between adjacent acquisition points, and the requirements for the integrity of continuous spatial data in frequency domain analysis. Simulated missing data is first used from the complete acquisition data during the calibration phase to determine the maximum number of consecutive missing nodes allowed without affecting the identification results of the main frequency bands and the determination of the deformation source. This number is then used as the missing number threshold. When there are differences in package size or acquisition point density among different devices, the missing number threshold is set according to the node density of the corresponding matrix region. If consecutive missing nodes occur during matrix construction, and the number of missing nodes exceeds the preset missing number threshold, the current deformation distribution matrix is ​​not directly output. Instead, boundary data of the same region in the previous valid window is used for local completion, and completion markers are written into the current matrix for subsequent frequency domain analysis to identify data integrity. By using the above processing methods, the influence of temperature changes, discrete sampling, and local missing measurements on the deformation characterization results can be kept within a controllable range, thereby ensuring that the output deformation distribution matrix has a stable spatial correspondence and a continuous data structure.

[0026] Frequency domain analysis was performed on the deformation distribution matrix, and thermal response characteristics and load response characteristics were separated based on the calibration frequency band to obtain the deformation source results; In this embodiment, after obtaining the deformation distribution matrix, frequency domain analysis is performed on the matrix data within the current acquisition window. First, the time-series data corresponding to the deformation distribution matrix is ​​read in the order of acquisition time, and the frequency components at each location are extracted. Then, thermal response features and load response features are separated based on a pre-established calibration frequency band. After feature separation, the thermal response features and load response features are comprehensively compared to determine the deformation source result corresponding to the current acquisition window. The deformation source result is used to characterize that the current deformation is mainly caused by thermal action, load action, or a combination of both, and serves as the input basis for subsequent damage accumulation judgment.

[0027] The calibration frequency band includes the thermal response frequency band and the load response frequency band; frequency domain analysis is performed on the deformation distribution matrix, and thermal response characteristics and load response characteristics are separated based on the calibration frequency band, including: performing Fourier transform on the deformation distribution matrix to obtain spectral data; and determining the deformation source based on the energy distribution of the spectral data in the thermal response frequency band and the load response frequency band.

[0028] In one embodiment, based on the above processing flow, the establishment method of the calibration frequency band, the execution process of frequency domain analysis, and the rules for determining energy distribution are further defined. Specifically, before the microelectronic device enters formal monitoring, calibration acquisition is performed first. Calibration acquisition is divided into two parts: thermal calibration acquisition and load calibration acquisition. Thermal calibration acquisition is performed under the condition that the external load is kept stable. By gradually changing the ambient temperature of the device or the device's operating heat level, the frequency response range formed by the deformation distribution matrix changing over time is recorded, and the frequency range that can stably reflect the changes in thermal expansion is written into the thermal response frequency band. Load calibration acquisition is performed under the condition that the ambient temperature is kept stable. By applying vibration, pressure, or periodic mechanical disturbance, the frequency response range formed by the deformation distribution matrix changing over time is recorded, and the frequency range that can stably reflect the changes in external load is written into the load response frequency band. The thermal response frequency band and the load response frequency band are not fixed constants, but are determined during the calibration stage according to the microelectronic device's packaging structure, mounting method, thermal inertia, and load change rate.

[0029] After entering the formal monitoring phase, the deformation distribution matrix is ​​first read from the current acquisition window and combined into continuous data segments according to time sequence. Then, a Fourier transform is performed on the continuous data segments to obtain the spectral data corresponding to each grid position. The amplitude change information at different frequency positions is stored in the spectral data. Subsequently, the amplitude distribution within the thermal response frequency band and the amplitude distribution within the load response frequency band are statistically analyzed, and thermal response characteristics and load response characteristics are formed based on the statistical results. The energy distribution here can be characterized by the cumulative amplitude result within the frequency band or the average amplitude result within the frequency band. The specific selection method is determined according to the device monitoring accuracy requirements. When there are obvious local hot spots in the device, the regional cumulative method is preferred to highlight the frequency change corresponding to local thermal expansion. When the device is subjected to relatively uniform stress, the overall averaging method is preferred to maintain the comparability between different acquisition windows. If there are obvious missing data in the current spectral data within the thermal response frequency band or the load response frequency band, the presence of completion markers during the matrix construction stage is checked first. If completion markers are present, a supplementary acquisition verification is performed on the current acquisition window to avoid local missing measurements directly affecting the source analysis results. By using the above methods, thermal response characteristics and load response characteristics can have consistent extraction rules and clear source boundaries, thereby ensuring that the deformation source results obtained subsequently are comparable and interpretable.

[0030] The deformation sources include thermally dominated deformation, load-dominated deformation, and thermally-load-coupled deformation; thermally-load-coupled deformation is used to indicate that the energy values ​​of the spectral data in both the thermal response band and the load response band are higher than the corresponding preset thresholds.

[0031] In one embodiment, after separating the thermal response characteristics and load response characteristics, the classification method of deformation source results, the setting principle of preset thresholds, and the triggering logic of thermal-load coupled deformation are further defined. Specifically, the deformation source results include thermally dominated deformation, load-dominated deformation, and thermal-load coupled deformation. Thermally dominated deformation indicates that the main frequency changes within the current acquisition window are concentrated in the thermal response frequency band, and the energy value within the thermal response frequency band reaches the thermal response judgment threshold, while the energy value within the load response frequency band does not reach the load response judgment threshold. The thermal response judgment threshold is determined during the calibration phase. Specifically, it is set based on the energy distribution range corresponding to the thermal response frequency band during thermal calibration acquisition, the upper limit of thermal fluctuations of microelectronic devices under normal operating conditions, and the stability requirements of multiple consecutive effective acquisition windows. The upper boundary of the thermal response frequency band energy value in the normal operating sample is preferentially used as the basic threshold. This basic threshold is then corrected by considering the device packaging structure, thermal inertia, and temperature rise rate, ensuring that the thermal response judgment threshold can distinguish between normal thermal fluctuations and abnormal thermal deformation without misjudging due to short-term temperature disturbances. When the device type, package size, or heat dissipation conditions change, the thermal response judgment threshold is re-determined according to the corresponding calibration results. Load-dominated deformation indicates that the main frequency changes within the current acquisition window are concentrated in the load response frequency band, and the energy value within the load response frequency band reaches the load response judgment threshold, while the energy value within the thermal response frequency band does not reach the thermal response judgment threshold.

[0032] Thermally coupled deformation indicates that the energy values ​​in both the thermal response band and the load response band simultaneously reach their respective thresholds, signifying that the current deformation is simultaneously affected by thermal expansion and external load. These two thresholds are determined during the calibration phase, based on factors including the reference energy range during calibration, the upper limit of band fluctuations under normal device operation, and the stability requirements of multiple consecutive acquisition windows. To avoid misjudgments due to occasional fluctuations in a single window, after the threshold conditions are met in the current acquisition window, consistency verification is performed using the previous or subsequent acquisition window. Only when the classification results of two consecutive valid windows are consistent is the final deformation source result output. If the energy values ​​in both the thermal response band and the load response band do not reach their corresponding thresholds, the current result is temporarily classified as a weak deformation state, and the data in this window is reserved for subsequent trend analysis, rather than directly entering the formal classification of thermally dominated deformation, load-dominated deformation, or thermally coupled deformation. If the energy values ​​of the thermal response band and the load response band are close to the threshold boundary, then the temperature change trend and load change trend in the environmental parameters are used as auxiliary criteria for discrimination: when the temperature change trend is continuously rising and the load change is stable, it is preferentially classified as thermally dominated deformation; when the load change is significant and the temperature change is gradual, it is preferentially classified as load-dominated deformation; when both increase simultaneously, it is classified as thermally-load-coupled deformation. Through the above classification method, the deformation source results can not only reflect the magnitude difference of the band energy, but also maintain a correspondence with the device operating environment, thereby providing a stable source label for subsequent damage accumulation judgment.

[0033] The damage accumulation result is determined based on the deformation source result and the deformation distribution matrix; In this embodiment, after determining the deformation source, the damage accumulation result is further determined by combining the deformation distribution matrix corresponding to the current acquisition window. When the deformation source result indicates that the current deformation is mainly affected by heat, or simultaneously affected by heat and external loads, the spectral data obtained in the frequency domain analysis stage is read and converted into a spectrum diagram to characterize the local frequency change distribution. Subsequently, the spectrum diagram is input into a convolutional neural network, which first extracts features from the local frequency enhancement region, the frequency shift continuous region, and the edge abrupt region in the spectrum diagram. Then, according to the identification rules formed by the labeled thermal damage samples, it matches the abnormal patterns corresponding to internal connection fatigue, local crack initiation, or encapsulation thermal instability, and outputs suspected abnormal regions. Subsequently, spatial continuity verification and window consistency verification are performed on the suspected abnormal regions. When the suspected abnormal regions are continuously distributed in adjacent grid regions and maintain similar abnormal features in the continuous acquisition window, the suspected abnormal regions are identified as damage regions. Further, the degree of thermal damage accumulation is determined based on the coverage area of ​​the damage region, the intensity of the abnormal response, and the changing trend in the continuous window. The larger the coverage area, the stronger the abnormal response, and the longer the duration, the higher the degree of thermal damage accumulation. The above processing method allows the identification results of the damaged area to simultaneously reflect the abnormal location and abnormal development state. Based on the spatial location, coverage area, and degree of thermal damage accumulation of the damaged area, the damage accumulation result corresponding to the current acquisition window is output and then transferred to the subsequent degradation risk assessment stage.

[0034] The damage accumulation result is determined based on the deformation source result and the deformation distribution matrix, including: when the deformation source result is thermally dominated deformation or thermal load coupled deformation, a spectrum map is generated based on the spectrum data, and a convolutional neural network is used to identify the spectrum map to obtain the damage area and the degree of thermal damage accumulation; the damage accumulation result is determined based on the damage area and the degree of thermal damage accumulation.

[0035] In one embodiment, based on the above processing flow, the construction method of the spectrogram, the recognition process of the convolutional neural network, the setting method of the thermal damage accumulation level, and the processing logic of abnormal samples are further defined. Specifically, when the deformation source result is thermally dominated deformation or thermally coupled deformation, the spectrum data corresponding to the current acquisition window is first extracted from the dataset output by the frequency domain analysis stage, and the spectrum data is rearranged according to the spatial position order in the deformation distribution matrix so that each grid position corresponds to a frequency response unit. Then, each frequency response unit is written into the two-dimensional image area to form a spectrogram. The horizontal position in the spectrogram is used to characterize the spatial distribution of the microelectronic device surface, the vertical position is used to characterize the distribution of the main frequency bands, and the image gray value or image amplitude is used to characterize the response strength of the corresponding position in the corresponding frequency band. Before formal use, the convolutional neural network is first trained using labeled thermal damage samples. The thermal damage samples are jointly established by historical detection data and disassembly verification results, wherein the sample labeling includes at least the damage location, damage boundary, and damage degree level.

[0036] To avoid mismatches between training samples and actual device types, sample groups are established during the training phase according to device packaging, mounting method, and thermal environment category. During the current monitoring process, the network parameters of the corresponding sample group are first called according to the device configuration of the microelectronic device, and then the identification is performed. During the identification process, the convolutional neural network first performs multi-layer feature extraction on the spectrogram, including local frequency enhancement regions, continuous frequency shift regions, and edge abrupt change regions, and then outputs the suspected damage location based on the extraction results. To reduce false detections caused by sporadic noise, a neighborhood consistency check is performed on the suspected damage location; only when similar frequency anomalies appear consecutively in adjacent grid areas, and the area of ​​the anomaly region reaches a preset area threshold, is the region identified as a damage region. The preset area threshold is determined during the sample calibration phase. Specifically, it is set based on the effective monitoring area of ​​the microelectronic device package surface, the spatial distribution of critical connection areas, the actual coverage of abnormal areas in historical damage samples, and the upper limit of the area of ​​local noise patches in normal samples. The smallest effective damage area in historically confirmed damage samples is preferentially used as the basic reference value. This area is then converted into the corresponding number of grid cells based on the current grid density, serving as the initial value for the preset area threshold. The initial value is corrected through comparison and verification with normal and damaged samples, ensuring that the preset area threshold can filter out small-area pseudo-abnormal areas formed by local noise, single-point drift, or short-term disturbances, while retaining continuous abnormal areas from early damage formation. When the package size, grid density, or critical connection area distribution of the microelectronic device changes, the preset area threshold is re-determined according to the calibration results of the corresponding device. The degree of thermal damage accumulation is not an abstract level, but a classification result comprehensively determined based on the duration of spectral anomalies, the coverage of abnormal areas, and the trend of abnormal amplitude changes.

[0037] The classification results can be divided into mild accumulation, moderate accumulation, and severe accumulation. The classification criteria are jointly determined during the training and calibration phases: when the abnormal region is limited to the local edge and changes slowly within a continuous window, it is judged as mild accumulation; when the abnormal region extends to the connection transition region or the hot concentration region, and the abnormal amplitude continues to increase within a continuous window, it is judged as moderate accumulation; when the abnormal region runs through multiple key connection regions, or when the abnormal amplitude rises rapidly within a continuous window and approaches the range of failed samples, it is judged as severe accumulation.

[0038] After obtaining the damaged area and the degree of thermal damage accumulation, the two are combined to generate the damage accumulation result corresponding to the current acquisition window. The damage accumulation result includes at least the damage location identifier, the extent of the damaged area, the degree of accumulation, and the time label of the current window. If the damage area output by the convolutional neural network deviates too much from the damage location in the previous effective window, a verification judgment is performed first. During the verification, the spectral data of the current window is reread and the identification is repeated. If the repeated identification results are still inconsistent, the current result is marked as a result to be confirmed and is not used for trend escalation judgment, but only reserved for subsequent continuous window comparisons. Through the above processing method, the damage accumulation result can simultaneously reflect the spatial location and development degree of thermal damage and be consistent with the deformation source result of the previous stage, thus providing a stable input for establishing a quantitative correspondence between damage state and performance degradation in the next stage.

[0039] The degradation risk score is determined based on the cumulative damage results, environmental parameters, historical degradation data and functional test results. The degradation risk score is used to characterize the degree of influence of the deformation corresponding to the deformation distribution matrix on the target performance index. In this embodiment, after determining the damage accumulation result, the degradation risk score is further determined by combining the environmental parameters corresponding to the current acquisition window, historical degradation data, and functional test results. First, the damage accumulation result corresponding to the current acquisition window is read, and the temperature parameters, load parameters, and functional test results corresponding to this damage accumulation result are extracted. Then, a current evaluation data set is formed according to pre-established feature organization rules. Subsequently, the current evaluation data set is input into the trained degradation evaluation model to obtain the degradation risk score. The degradation risk score is used to characterize the degree of influence of the current deformation state on the target performance index. The higher the degradation risk score, the more likely the current deformation change is to cause the target performance index to deviate from the normal range. After obtaining the degradation risk score, it is sent to the next stage as a direct input for determining the degradation evaluation result.

[0040] The degradation risk score is determined based on the cumulative damage results, environmental parameters, historical degradation data, and functional test results. This includes: using a random forest model trained on historical degradation data, inputting the cumulative damage results, environmental parameters, and functional test results into the random forest model to obtain the degradation risk score.

[0041] In one embodiment, based on the above processing flow, the construction method of historical degradation data, the training method of the random forest model, the composition method of the current evaluation data set, and the output rules of the degradation risk score are further defined. Specifically, the historical degradation data consists of historical damage accumulation results, historical environmental parameters, and historical functional test results, with each set of historical degradation data corresponding to a known degradation state label. The historical damage accumulation results include at least the damage location identifier, damage area range, accumulation level, and acquisition time; the historical environmental parameters include at least historical temperature parameters and historical load parameters; and the historical functional test results include at least the target performance index test values ​​of the same type as the current microelectronic device. To ensure comparability between training samples, when establishing the historical degradation data, it is first classified according to device type, packaging form, and installation method, and then the samples are organized within the same category.

[0042] The random forest model is trained offline before formal monitoring. During training, historical damage accumulation results, historical environmental parameters, and historical functional test results are sequentially converted into input features. Known degradation state labels are then used as output targets to generate a degradation assessment model for the current device. During current monitoring, damage accumulation results are first read from the current acquisition window, followed by temperature parameters, load parameters, and functional test results corresponding to that window. These are then combined into the current assessment data set in the same order as during training. If the current assessment data set contains missing items, it is first checked whether it is a single-channel transient anomaly. If re-acquisition is possible, a re-acquisition is performed. If missing items still exist after re-acquisition, the current window is marked as a pending confirmation window, and the formal degradation risk score is not output temporarily. If the current assessment data set is complete, it is input into the random forest model, which outputs a continuously changing degradation risk score.

[0043] The degradation risk score range can be pre-set as a continuous interval from low to high, and calibrated according to the distribution of normal operation samples, warning samples, and failure samples. To prevent occasional fluctuations in a single acquisition window from being directly amplified into abnormal risks, after obtaining the current degradation risk score, it can be smoothed by combining it with the degradation risk score of the previous valid window. When the degradation risk scores of two consecutive valid windows change in the same direction, the current output value is retained; when the changes are significantly opposite, the current window is marked as a verification window, and it waits for confirmation in the next window. Through the above processing method, the degradation risk score can reflect both the current damage state and the current environmental conditions and functional test changes, thus forming a risk quantification result that can directly proceed to the next stage.

[0044] Historical degradation data includes historical damage accumulation results, historical environmental parameters, and historical performance offsets. The historical performance offsets indicate the degree of deviation of historical functional test results from the target performance index benchmark value, which is the reference value of the target performance index under the baseline operating conditions. The damage accumulation results, environmental parameters, and functional test results are input into the random forest model, including: determining the current performance offset based on the functional test results and the target performance index benchmark value; and inputting the damage accumulation results, environmental parameters, and current performance offset into the random forest model to obtain the degradation risk score.

[0045] In one embodiment, based on the completion of the random forest model invocation, the methods for establishing historical performance offsets, determining the target performance index benchmark value, forming the current performance offset, and the correspondence between degradation risk scores and target performance indices are further defined. Specifically, historical degradation data includes not only historical damage accumulation results and historical environmental parameters, but also historical performance offsets. Historical performance offsets are used to characterize the degree of deviation of historical functional test results from the target performance index benchmark value. The target performance index benchmark value is not a fixed constant, but a reference value obtained through benchmark acquisition under benchmark conditions.

[0046] The baseline operating condition is preferentially selected when the microelectronic device is operating under stable temperature, stable load, and normal working conditions. Multiple functional tests are performed continuously under these conditions, and the test results within the stable range are taken as the baseline value for the target performance index. Upon entering formal monitoring, the current performance offset is first determined based on the functional test results in the current acquisition window and the baseline value for the target performance index. If the target performance index is a continuous numerical index such as propagation delay, response time, or leakage current, the direction and magnitude of the deviation between the current test value and the baseline value are used as components of the current performance offset. If the target performance index is a discrete index such as error rate, stability level, or pass / fail result, the test results are first converted into a unified graded value before determining the current performance offset.

[0047] After obtaining the current performance offset, the accumulated damage result, environmental parameters, and the current performance offset are input into the random forest model, outputting a degradation risk score. Here, the introduction of the current performance offset does not replace the functional test results, but rather transforms the functional test results into an offset representation more suitable for cross-window comparison, enabling the model to directly identify the correspondence between deformation changes, environmental changes, and performance deviations. To avoid distortion of offset judgments due to long-term outdated target performance index benchmark values, benchmark acquisition is re-executed and the target performance index benchmark values ​​are updated after maintenance, device replacement, or significant changes in operating conditions of the microelectronic device. Through this method, the degradation risk score not only reflects the damage state itself but also directly corresponds to the deviation trend of the target performance index, thus more accurately characterizing the degree of influence of deformation corresponding to the deformation distribution matrix on the target performance index.

[0048] The degradation assessment result is determined based on the degradation risk score.

[0049] In this embodiment, after obtaining the degradation risk score, the degradation assessment result corresponding to the current acquisition window is determined based on the degradation risk score. First, the degradation risk score in the current acquisition window is read, and then combined with the environmental parameters corresponding to that acquisition window to form the current discrimination input. Subsequently, the current discrimination input is sent to the degradation assessment unit, which outputs the degradation assessment result corresponding to the current operating state. The degradation assessment result is used to characterize whether the current microelectronic device is within the normal degradation range, the degradation range of concern, or the abnormal degradation range. After obtaining the degradation assessment result, it is written into the monitoring record of the current window and serves as the basis for subsequent alarm output and operation record management.

[0050] The degradation assessment result is determined based on the degradation risk score, including: constructing a state discrimination vector based on the degradation risk score and environmental parameters; and inputting the state discrimination vector into a support vector machine classifier to obtain the degradation assessment result.

[0051] In one embodiment, based on the above processing flow, the construction method of the state discrimination vector, the training method of the support vector machine classifier, and the output rules of the degradation evaluation results are further defined. Specifically, after the degradation risk score is determined, the environmental parameters corresponding to the degradation risk score are first read from the current acquisition window. The environmental parameters include at least temperature parameters and load parameters. In this embodiment, the state discrimination vector refers to a set of input data used to characterize the current device degradation state. The state discrimination vector includes at least the degradation risk score, the current temperature parameter, the current load parameter, and a trend marker corresponding to the current temperature parameter.

[0052] Trend markers characterize the directional relationship between temperature and load changes between the current acquisition window and the previous effective acquisition window, enabling the degradation assessment unit to identify not only the current state but also the trend of state change. The support vector machine classifier is trained before formal monitoring, with training samples consisting of historical degradation data and confirmed degradation state labels. Degradation state labels include at least three categories: normal degradation, concern degradation, and abnormal degradation. Normal degradation indicates that both functional test results and damage accumulation results are within acceptable limits; concern degradation indicates that the current state has deviated from the normal range but has not yet reached the triggering conditions for abnormal degradation; abnormal degradation indicates that the current state has already significantly affected the target performance indicators, and shows a continuously worsening trend within consecutive windows.

[0053] During the training phase, historical degradation risk scores, historical environmental parameters, and historical trend labels are first combined to form a historical state discrimination vector. This vector, along with the corresponding degradation state label, is then input into a support vector machine (SVM) classifier to complete training. Upon entering the formal monitoring phase, the degradation risk score corresponding to the current acquisition window and the environmental parameters acquired synchronously with that score are first read. Then, a state discrimination vector corresponding to the current acquisition window is constructed according to the feature arrangement consistent with the training phase. Subsequently, the state discrimination vector is input into the SVM classifier, which determines the current state based on the established classification boundaries and outputs the current degradation assessment result. The degradation assessment result can be characterized as normal degradation, concern degradation, or abnormal degradation. To avoid misjudgments caused by instantaneous fluctuations in a single acquisition window, consistency verification can be performed by combining the current degradation assessment result with the discrimination result of the previous valid acquisition window. When the discrimination results of two consecutive valid acquisition windows are consistent, the current degradation assessment result is confirmed. When the results are inconsistent, the current acquisition window is marked as a window awaiting review, and further confirmation is awaited from subsequent acquisition windows. To avoid classification result jumps caused by occasional perturbations in a single acquisition window, after the support vector machine classifier outputs the current degradation assessment result, a consistency check is performed by combining it with the assessment result of the previous valid acquisition window. When the assessment results of two consecutive valid windows are consistent, the current degradation assessment result is directly confirmed. When the assessment results of two valid windows are inconsistent, the current acquisition window is marked as a window to be reviewed, and it waits for the next acquisition window to be judged again. If the assessment result of the next acquisition window is consistent with the current acquisition window, the current classification result is confirmed; if the next acquisition window reverts to the classification result of the previous valid acquisition window, the assessment conclusion of the previous valid acquisition window is retained. Through the above processing method, the degradation assessment result can simultaneously reflect the risk quantification result and the state of environmental change, thereby avoiding the misjudgment problem caused by directly outputting the assessment conclusion based on a single degradation risk score.

[0054] The method in this application also includes: outputting alarm information when the degradation assessment result indicates abnormal degradation; and recording the deformation distribution matrix, damage accumulation result and degradation risk score corresponding to the alarm information.

[0055] In one embodiment, based on the completed degradation assessment result output, the alarm output method, recording content, and recording triggering rules under abnormal degradation conditions are further defined. Specifically, when the degradation assessment result indicates that the current acquisition window is within the abnormal degradation range, the alarm processing unit first generates alarm information corresponding to the current acquisition window. The alarm information includes at least the device identifier, acquisition window timestamp, degradation assessment result level, and alarm triggering reason. The alarm triggering reason is determined based on the current degradation risk score, current damage accumulation result, and current environmental parameters, and is used to explain that the abnormal degradation mainly originates from enhanced thermal effects, enhanced load effects, or a combination of both.

[0056] To avoid repeated alarms caused by short-term noise, this embodiment uses a window confirmation method for alarm output: when a single acquisition window is first determined to be abnormally degraded, a pending alarm is generated; when the next valid acquisition window also remains abnormally degraded, the pending alarm is converted into a formal alarm and output. Formal alarms can be written to the local monitoring terminal or sent to a host computer or maintenance management platform. After alarm output, the monitoring data corresponding to the alarm information also needs to be recorded synchronously. The recorded content includes at least the deformation distribution matrix, damage accumulation result, and degradation risk score corresponding to the current acquisition window. The deformation distribution matrix is ​​used to retain the current spatial deformation state, the damage accumulation result is used to retain the current damage location and accumulation degree, and the degradation risk score is used to retain the current risk quantification level. To ensure that subsequent review and trend analysis can directly call historical records, the recording unit binds the above data with the alarm information and writes it into the degradation event recording area, maintaining the recording order of the same device in consecutive abnormal windows. If the same device outputs multiple formal alarms consecutively within a preset time period, the records are appended in chronological order without overwriting the previous record. If the same device recovers to a state of concern or normal degradation after abnormal degradation, a recovery marker is written to the recovery window, allowing subsequent maintenance personnel to directly view the start, development, and mitigation process of a complete degradation event. Through the above processing method, the degradation assessment phase can not only output classification results but also form a traceable monitoring record closed loop when abnormal degradation occurs.

[0057] like Figure 2 As shown, a microelectronic device performance degradation monitoring system is used to implement a microelectronic device performance degradation monitoring method. The system includes: The signal acquisition module is used to acquire strain signals, environmental parameters, and functional test results of microelectronic devices. The functional test results correspond to target performance indicators. The signal acquisition module consists of a strain sensor array, an environmental parameter acquisition unit, a functional test unit, an analog-to-digital conversion circuit, a clock synchronization circuit, and a data interface circuit. The strain sensor array is used to acquire strain signals of microelectronic devices, the environmental parameter acquisition unit is used to acquire temperature and load parameters, the functional test unit is used to output functional test results corresponding to the target performance indicators, the analog-to-digital conversion circuit is used to convert the analog signals output from each acquisition channel into digital signals, the clock synchronization circuit is used to uniformly control the acquisition timing of each channel, and the data interface circuit is used to transmit the acquisition results to the subsequent processing unit.

[0058] The matrix construction module is used to correct and mesh the strain signal to obtain the deformation distribution matrix. The matrix construction module consists of a signal correction circuit, a buffer unit, a coordinate mapping unit, an interpolation processing unit, and a matrix generation unit. The signal correction circuit is used to correct the drift of the strain signal based on the temperature parameter in the environmental parameters. The buffer unit is used to store the strain data before and after correction. The coordinate mapping unit is used to establish the correspondence between the strain acquisition point and the position of the device surface. The interpolation processing unit is used to perform spatial interpolation on the corrected strain signal. The matrix generation unit is used to form the deformation distribution matrix based on the interpolation result.

[0059] The source analysis module is used to perform frequency domain analysis on the deformation distribution matrix and separate thermal response features and load response features based on the calibration frequency band to obtain the deformation source results. The source analysis module consists of a frequency domain transformation unit, a frequency band storage unit, a feature separation unit, and a source discrimination unit. The frequency domain transformation unit is used to perform frequency domain analysis on the deformation distribution matrix, the frequency band storage unit is used to store the thermal response frequency band and the load response frequency band in the calibration frequency band, the feature separation unit is used to separate thermal response features and load response features based on the calibration frequency band, and the source discrimination unit is used to output the deformation source results based on the separation results.

[0060] The damage determination module is used to determine the cumulative damage result based on the deformation source result and the deformation distribution matrix. The damage determination module consists of a feature map generation unit, a neural network operation unit, a region extraction unit, and a cumulative determination unit. The feature map generation unit is used to generate recognition input data based on the spectral data corresponding to the deformation source result. The neural network operation unit is used to perform damage recognition operations. The region extraction unit is used to output the damage region. The cumulative determination unit is used to combine the damage region and the degree of thermal damage accumulation to form the cumulative damage result.

[0061] The risk assessment module is used to determine the degradation risk score based on the cumulative damage results, environmental parameters, historical degradation data, and functional test results. The degradation risk score is used to characterize the degree of influence of the deformation corresponding to the deformation distribution matrix on the target performance index. The risk assessment module consists of a historical data storage unit, a feature organization unit, a model calculation unit, and a score output unit. The historical data storage unit is used to store historical degradation data. The feature organization unit is used to organize the cumulative damage results, environmental parameters, and functional test results into assessment inputs. The model calculation unit is used to call the degradation assessment model for calculation. The score output unit is used to output the degradation risk score, which characterizes the degree of influence of the deformation corresponding to the deformation distribution matrix on the target performance index.

[0062] The result output module is used to determine the degradation assessment result based on the degradation risk score. The result output module consists of a status discrimination unit, a display interface unit, an alarm interface unit, and a recording unit. The status discrimination unit is used to determine the degradation assessment result based on the degradation risk score, the display interface unit is used to output the assessment result information, the alarm interface unit is used to output alarm information when the assessment result meets the conditions, and the recording unit is used to save the degradation assessment result and related monitoring data.

[0063] It should be noted that, in this document, the terms "comprising," "including," and any other variations are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Specific examples have been used in this document to illustrate the principles and implementation methods of the technical solutions of this application. The above examples are only for the purpose of helping to understand the methods and core ideas of this application. The above descriptions are merely preferred embodiments of this application. It should be pointed out that, due to the limitations of written expression and the objective existence of infinite specific structures, those skilled in the art can make several improvements, modifications, or changes without departing from the principles of this application, and can also combine the above technical features in an appropriate manner; these improvements, modifications, changes, or combinations, or the direct application of the concept and technical solutions of this application to other situations without modification, should all be considered within the scope of protection of this application.

Claims

1. A method for monitoring the performance degradation of microelectronic devices, characterized in that, The method includes: Acquire strain signals, environmental parameters, and functional test results of microelectronic devices, wherein the functional test results correspond to target performance indicators; The strain signal is corrected and meshed to obtain the deformation distribution matrix; Frequency domain analysis was performed on the deformation distribution matrix, and thermal response characteristics and load response characteristics were separated based on the calibration frequency band to obtain the deformation source results; The damage accumulation result is determined based on the deformation source result and the deformation distribution matrix; A degradation risk score is determined based on the accumulated damage results, the environmental parameters, historical degradation data, and the functional test results. The degradation risk score is used to characterize the degree of influence of the deformation corresponding to the deformation distribution matrix on the target performance index. The degradation assessment result is determined based on the degradation risk score.

2. The method according to claim 1, characterized in that, The correction and meshing process for the strain signal includes: The strain signal is temperature drift corrected based on the temperature parameter in the environmental parameters to obtain the corrected strain signal. Spatial interpolation is performed on the corrected strain signal to obtain gridded deformation data; The deformation distribution matrix is ​​determined based on the gridded deformation data.

3. The method according to claim 1, characterized in that, The calibration frequency band includes the thermal response frequency band and the load response frequency band; The step of performing frequency domain analysis on the deformation distribution matrix and separating thermal response characteristics and load response characteristics based on the calibration frequency band includes: Perform a Fourier transform on the deformation distribution matrix to obtain the spectral data; The deformation source is determined based on the energy distribution within the thermal response band and the load response band using the spectral data.

4. The method according to claim 3, characterized in that, The deformation sources include thermally dominated deformation, load-dominated deformation, and thermal-load coupled deformation. The thermal load coupling deformation is used to indicate that the energy value of the spectral data in the thermal response frequency band and the energy value in the load response frequency band are both higher than the corresponding preset threshold.

5. The method according to claim 4, characterized in that, The step of determining the damage accumulation result based on the deformation source result and the deformation distribution matrix includes: When the deformation source is determined to be thermally dominated deformation or thermal load coupled deformation, a spectrum diagram is generated based on the spectrum data, and a convolutional neural network is used to identify the spectrum diagram to obtain the damage area and the degree of thermal damage accumulation. The damage accumulation result is determined based on the damaged area and the degree of thermal damage accumulation.

6. The method according to claim 1, characterized in that, The determination of the degradation risk score based on the accumulated damage results, the environmental parameters, historical degradation data, and the functional test results includes: Using a random forest model trained based on the historical degradation data, the cumulative damage result, the environmental parameters, and the functional test results are input into the random forest model to obtain the degradation risk score.

7. The method according to claim 6, characterized in that, The historical degradation data includes historical damage accumulation results, historical environmental parameters, and historical performance offsets. The historical performance offsets are used to indicate the degree of deviation of historical functional test results from the target performance index benchmark value. The target performance index benchmark value is a reference value of the target performance index under the benchmark operating conditions. The step of inputting the damage accumulation result, the environmental parameters, and the functional test results into the random forest model includes: The current performance offset is determined based on the functional test results and the target performance index benchmark value. The cumulative damage result, the environmental parameters, and the current performance offset are input into the random forest model to obtain the degradation risk score.

8. The method according to claim 1, characterized in that, The determination of the degradation assessment result based on the degradation risk score includes: A state discrimination vector is constructed based on the degradation risk score and the environmental parameters; The state discrimination vector is input into a support vector machine classifier to obtain the degradation evaluation result.

9. The method according to claim 8, characterized in that, The method further includes: If the degradation assessment result indicates abnormal degradation, an alarm message will be output; Record the deformation distribution matrix, the damage accumulation result, and the degradation risk score corresponding to the alarm information.

10. A microelectronic device performance degradation monitoring system, used to implement the microelectronic device performance degradation monitoring method according to any one of claims 1-9, characterized in that, The system includes: The signal acquisition module is used to acquire strain signals, environmental parameters, and functional test results of microelectronic devices, wherein the functional test results correspond to target performance indicators. A matrix construction module is used to correct and mesh the strain signal to obtain a deformation distribution matrix; The source analysis module is used to perform frequency domain analysis on the deformation distribution matrix and separate thermal response characteristics and load response characteristics based on the calibration frequency band to obtain the deformation source results; The damage determination module is used to determine the cumulative damage result based on the deformation source result and the deformation distribution matrix. The risk assessment module is used to determine a degradation risk score based on the damage accumulation results, the environmental parameters, historical degradation data, and the functional test results. The degradation risk score is used to characterize the degree of influence of the deformation corresponding to the deformation distribution matrix on the target performance index. The results output module is used to determine the degradation assessment result based on the degradation risk score.

Images