Life evaluation method for power device working condition data fusion

Abstract

The present application relates to the technical field of power device life data analysis, and particularly relates to a life evaluation method for power device working condition data fusion. The method comprises the following steps: performing device operation working condition segment analysis based on historical power device multi-source operation monitoring data to generate device operation working condition segment data; analyzing working condition stress migration path data based on the device operation working condition segment data; analyzing working condition damage propagation trajectory data through the working condition stress migration path data; analyzing device working condition life degradation evolution data based on the working condition damage propagation trajectory data; generating a device working condition-life degradation prediction model through the device working condition life degradation evolution data; and transmitting target power device operation cycle data to the device working condition-life degradation prediction model to perform intelligent evaluation of the life of the target power device. The present application realizes intelligent evaluation of the life of the power device by analyzing the fusion working condition of the power device.

Description

Technical Field

[0001] This invention relates to the field of power device lifetime data analysis technology, and in particular to a lifetime evaluation method for power device operating condition data fusion. Background Technology

[0002] Power devices are core components in electrical equipment such as power electronics, new energy, and industrial frequency converters. They are widely used in key areas such as smart grids, rail transit, and new energy power generation. Their reliability and service life directly determine the safe and stable operation of the power system. Under actual operating conditions, power devices are constantly exposed to load fluctuations, temperature variations, electrical shocks, and frequent operating condition switching, which can easily lead to cumulative internal damage, performance degradation, parameter deviations, and even device failure. In severe cases, this can cause equipment malfunctions and system shutdowns. Therefore, accurate life assessment and degradation prediction are crucial for preventative maintenance of devices and ensuring the reliable operation of the power system. However, existing power device lifetime assessment methods lack the ability to fuse multi-source operational monitoring data, making it impossible to accurately characterize the device's operating status and stress characteristics under complex dynamic conditions. They neglect the continuous dynamic switching characteristics of operating conditions, making it difficult to accurately classify fragmented operating conditions, analyze the operating condition transition mechanism and stress migration law. At the same time, they cannot quantify the abnormal stress activation, diffusion and superposition damage process, and do not consider the correlation, uncertainty and multi-factor synergistic degradation effect of damage propagation. As a result, existing prediction models have poor operating condition adaptability, low accuracy and weak generalization ability, making it difficult to adapt to complex actual operating conditions and unable to achieve accurate and intelligent lifetime assessment of power devices throughout their entire life cycle. Summary of the Invention

[0003] Based on this, the present invention provides a lifetime evaluation method for power device operating condition data fusion to solve at least one of the above-mentioned technical problems.

[0004] To achieve the above objectives, a lifespan assessment method based on power device operating condition data fusion includes the following steps: Step S1: Obtain historical multi-source operation monitoring data of power devices; perform device operation condition segment analysis based on historical multi-source operation monitoring data of power devices to generate device operation condition segment data; Step S2: Analyze the migration path of stress under operating conditions based on the device operating condition segment data, and generate stress migration path data under operating conditions; Step S3: Perform damage propagation trajectory analysis on the stress migration path data under the working condition to generate damage propagation trajectory data under the working condition; Step S4: Based on the damage propagation trajectory data under operating conditions, perform device operating condition life degradation evolution analysis to generate device operating condition life degradation evolution data; establish a prediction mapping relationship model between device operating conditions and life degradation using the device operating condition life degradation evolution data to generate a device operating condition-life degradation prediction model. Step S5: Obtain the target power device operating cycle data; transmit the target power device operating cycle data to the device operating condition-lifetime degradation prediction model for intelligent lifetime evaluation of the target power device, and generate target power device lifetime evaluation data.

[0005] Furthermore, step S1 includes the following steps: Step S11: Obtain historical multi-source operation monitoring data of power devices; Step S12: Perform time-series identification and monitoring data preprocessing on historical power device multi-source operation monitoring data to obtain standard device multi-source operation monitoring data; Step S13: Perform multi-source operation characteristic analysis on the multi-source operation monitoring data of standard devices to generate multi-source operation characteristic data of devices; Step S14: Perform continuous analysis of operating conditions based on multi-source operating characteristic data of the device to generate continuous operating condition data; Step S15: Divide the standard device multi-source operation monitoring data into device operation condition segments using continuous operation status data to generate device operation condition segment data.

[0006] Furthermore, step S14 includes the following steps: Step S141: Based on the device multi-source operation characteristic data, perform the change boundary identification processing of the device multi-source operation characteristics to generate device multi-source operation characteristic change boundary data; Step S142: Analyze the operating status of the standard device multi-source operation monitoring data by using the boundary data of the device multi-source operation characteristic change, and generate device operating status data. Step S143: Perform a continuity analysis of the device operating condition status data to generate continuous operating condition status data.

[0007] Furthermore, step S2 includes the following steps: Step S21: Perform conversion drive relationship analysis on the device operating condition segment data to generate operating condition conversion drive data; Step S22: Design the operating condition migration node based on the operating condition transition driving data to obtain the device operating condition migration node data; Step S23: Analyze the migration connection relationship of operating stress based on the device operating condition migration node data, and generate operating stress migration connection relationship data; Step S24: Perform stress migration chain combination processing on the stress migration connection relationship data under working conditions to generate stress migration chain data under working conditions; Step S25: Perform stress migration chain correlation analysis based on the stress migration chain data under the working conditions, and generate stress migration chain correlation data under the working conditions; Step S26: Reconstruct the migration path of the stress under the working condition using the associated data of the stress migration chain to generate stress migration path data.

[0008] Furthermore, step S23 includes the following steps: Step S231: Perform multi-domain feature analysis on the device condition migration node data to generate multi-domain feature data of device condition migration node operation. Step S232: Calculate the stress change of time-adjacent operating condition migration nodes based on the multi-domain feature data of the device operating condition migration nodes, and generate operating condition migration node stress change data. Step S233: Analyze the migration connection relationship of working condition stress based on the stress change data of the working condition migration nodes, and generate working condition stress migration connection relationship data.

[0009] Furthermore, step S3 includes the following steps: Step S31: Perform stress multi-source incremental characteristic analysis on the migration path based on the stress migration path data under the working condition, and generate stress multi-source incremental characteristic data of the migration path. Step S32: Perform an analysis of the activated region of abnormal stress under working conditions on the incremental characteristic data of multi-source stress in the migration path, and generate data of activated region of abnormal stress under working conditions. Step S33: Perform abnormal stress state analysis based on the data of the activated area of ​​abnormal stress under working conditions, and generate abnormal stress state data under working conditions. Step S34: Based on the data of the activated region of abnormal stress under working conditions and the data of the state of abnormal stress under working conditions, perform damage propagation trajectory analysis under working conditions to generate damage propagation trajectory data under working conditions.

[0010] Furthermore, step S34 includes the following steps: Step S341: Analyze the abnormal stress diffusion process of the activated region data under abnormal working conditions to generate abnormal stress diffusion process data. Step S342: Analyze the overlap characteristics of abnormal stress diffusion based on the abnormal stress diffusion process data under the working condition, and generate abnormal stress diffusion overlap characteristic data. Step S343: Analyze the damage propagation trajectory of the working condition by using the abnormal stress diffusion overlap characteristic data and the abnormal stress state data of the working condition, and generate the working condition damage propagation trajectory data.

[0011] Furthermore, step S4 includes the following steps: Step S41: Divide the working condition damage propagation trajectory data into working condition damage propagation units to generate working condition damage propagation unit data; Step S42: Perform condition damage degradation mapping feature analysis based on the condition damage propagation unit data to generate condition damage degradation mapping feature data. Step S43: Design the contribution factor matrix of working condition life degradation based on the working condition damage degradation mapping feature data to obtain the working condition life degradation contribution matrix data. Step S44: Perform device operating life degradation evolution analysis on the operating life degradation contribution matrix data to generate device operating life degradation evolution data. Step S45: Establish a prediction mapping model between device operating conditions and lifetime degradation using device operating condition lifetime degradation evolution data, and generate a device operating condition-lifetime degradation prediction model.

[0012] Furthermore, step S42 includes the following steps: Step S421: Perform element damage propagation dependency analysis on the element damage propagation data under working conditions to generate element damage propagation dependency data; Step S422: Perform a specificity analysis of the impact of unit damage propagation on the unit based on the unit damage propagation data, and generate specificity data of the impact of unit damage propagation on the unit. Step S423: Perform damage uncertainty propagation factor analysis on the specific data of damage propagation impact of unit elements, and generate damage uncertainty propagation factor data of unit elements. Step S424: Perform condition damage degradation mapping feature analysis based on the unit damage propagation dependency data and the condition unit damage uncertainty propagation factor data to generate condition damage degradation mapping feature data.

[0013] Furthermore, step S44 includes the following steps: Step S441: Perform a collaborative relationship analysis of the working condition damage degradation mapping feature data to generate collaborative relationship data of working condition damage degradation. Step S442: Based on the data on the synergistic relationship between working condition damage and degradation, perform cross-correlation analysis on the evolution of working condition damage and degradation factors, and generate cross-correlation data on the evolution of working condition damage and degradation factors. Step S443: Analyze the constraint relationship of working condition damage degradation behavior based on the working condition damage degradation mapping feature data, and generate working condition damage degradation behavior constraint relationship data. Step S444: Analyze the device's operating life degradation evolution by using the cross-correlation factor data of operating condition damage degradation evolution and the constraint relationship data of operating condition damage degradation behavior on the operating life degradation contribution matrix data, and generate device operating life degradation evolution data.

[0014] The beneficial effects of this application are as follows: By performing time-series identification, preprocessing, multi-source feature analysis, and continuous operation status analysis on historical multi-source operation monitoring data of power devices, this invention can effectively unify multi-source monitoring data standards, uncover multi-dimensional operation characteristics of devices, and accurately identify the boundaries of multi-source operation characteristic changes and determine the real-time operating status of devices. This achieves refined segmentation of operating conditions based on the continuity of operating conditions, enabling a true reconstruction of the actual operating conditions details of power devices throughout their entire lifecycle. Based on this refined segmented data, through operating condition transition driving relationship analysis, operating condition migration node design, multi-domain characteristic stress calculation, and migration chain combination and correlation analysis, the precise reconstruction of the operating condition stress migration path is achieved. This solves the problems of being unable to analyze the internal driving mechanism of operating condition transitions, difficulty in quantifying the stress change patterns between time-series operating conditions, and inability to effectively construct stress migration correlations. It captures the stress evolution characteristics during different operating condition switching processes, quantifies the stress fluctuations and transmission relationships between adjacent operating condition nodes, constructs a complete operating condition stress migration link and migration path, and reveals the dynamic transmission and evolution laws of stress in power devices under complex dynamic operating conditions. Based on stress migration path data under operating conditions, multi-dimensional and refined damage simulation is conducted. By analyzing the multi-source incremental characteristics of stress along the migration path, the dynamic incremental changes in stress under operating conditions are accurately captured. Combined with the identification of abnormal stress activation regions and the assessment of abnormal stress states, the location and state of abnormal stress in power devices under complex operating conditions are clarified. Simultaneously, the analysis of abnormal stress diffusion process and diffusion overlap characteristics is further refined. Combined with stress activation regions and stress states, the accurate construction of the damage propagation trajectory under operating conditions is completed. This accurately recreates the entire process of damage accumulation, diffusion, and superposition evolution of power devices under alternating multiple operating conditions, characterizing the spatiotemporal characteristics and correlation properties of device damage propagation, and solving the problems of ambiguous damage mechanism characterization and distorted damage trajectory assessment. Based on the damage propagation trajectory data under operating conditions, the device damage region is finely decomposed through damage propagation unit division. Combined with the analysis of unit damage propagation dependency, influence specificity, and uncertainty propagation factors, the core features of the damage degradation mapping under operating conditions are accurately extracted. Building upon this foundation, a matrix of factors contributing to device lifespan degradation under different operating conditions is constructed to quantify the differentiated impact weights of various operating conditions and damage units on device lifespan degradation. Simultaneously, lifespan degradation evolution analysis is conducted by combining damage-degradation synergy, evolutionary cross-correlation factors, and behavioral constraints. This approach fully considers multi-factor coupled degradation, damage uncertainty, and synergistic evolution characteristics, effectively improving the model's adaptability to complex real-world operating conditions and the accuracy of degradation prediction. It achieves mechanistic and quantitative modeling of the lifespan degradation evolution law of power devices, significantly enhancing the reliability and generalization performance of the lifespan prediction model. Based on the trained device operating condition-lifespan degradation prediction model, intelligent lifespan evaluation processing is performed by incorporating actual operating cycle data of the target power device. Relying on a complete technical chain including detailed analysis of operating conditions, stress migration reconstruction, damage trajectory extrapolation, and degradation evolution modeling, real-time, full-cycle intelligent lifespan status evaluation of the target power device is achieved.Based on multi-source data fusion, stress damage evolution mechanism and intelligent mapping model, dynamic and accurate life assessment is completed. It can adapt to the complex and ever-changing actual operating conditions of power devices, and accurately output life assessment data that fits the actual service status of the devices. It realizes intelligent, automated and high-precision driving application of power device life assessment, and provides reliable data support and technical basis for power device health management, fault early warning and preventive operation and maintenance strategy formulation.

[0015] Therefore, the life evaluation method for power device operating condition data fusion of this invention can comprehensively and accurately depict the actual operating status and stress characteristics of power devices under complex dynamic operating conditions; analyze the operating condition transformation mechanism and stress migration evolution law, and realize the refined quantitative characterization of abnormal stress activation, diffusion and superimposed damage process. It takes into account the correlation, uncertainty and multi-factor synergistic degradation effect of device damage propagation. By constructing a predictive mapping model between operating conditions and life degradation, it greatly improves the adaptability, prediction accuracy and generalization ability of the model to actual complex operating conditions. It can efficiently realize intelligent and accurate life evaluation of the entire life cycle of power devices, and provide reliable technical support for power device health status management, preventive maintenance strategy formulation and stable operation and maintenance of power systems. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the steps of a life evaluation method based on power device operating condition data fusion according to the present invention. Figure 2 for Figure 1 A detailed flowchart illustrating the implementation steps of step S4. The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0017] The technical method of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0018] Furthermore, the accompanying drawings are merely illustrative of the invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. Functional entities may be implemented in software, or in one or more hardware modules or integrated circuits, or in different network and / or processor methods and / or microcontroller methods. The term "and / or" as used herein includes any and all combinations of one or more of the associated items listed.

[0019] To achieve the above objectives, please refer to Figures 1 to 2 This invention provides a lifetime assessment method based on power device operating condition data fusion. In the embodiments of this invention, please refer to... Figure 1 The diagram shown is a flowchart illustrating the steps of a lifespan assessment method based on power device operating condition data fusion according to the present invention. The lifespan assessment method based on power device operating condition data fusion includes the following steps: Step S1: Obtain historical multi-source operation monitoring data of power devices; perform device operation condition segment analysis based on historical multi-source operation monitoring data of power devices to generate device operation condition segment data; In this embodiment of the invention, the historical operational information of power devices across all dimensions is collected and the structured segments of operating conditions are standardized. Multi-source operational monitoring information generated during the long-term service of power devices is used to perform comprehensive operational condition analysis and segment decomposition. Full historical operational monitoring information across the electrothermal, electrical, mechanical, and environmental dimensions of the power devices' service life is collected, covering continuous time-series records of device temperature operation, electrical load status, mechanical vibration status, and external environmental status. The collected multi-source monitoring information undergoes unified time-series calibration and structured preprocessing to achieve time-series alignment and format regularization of monitoring information with different acquisition frequencies, different unit types, and different time-series precisions. Invalid monitoring content caused by equipment maintenance, start-up / shutdown transients, and abnormal disturbances is removed, forming a standardized monitoring dataset with regular time-series data and valid status. Multi-dimensional operational feature analysis is performed on the standardized monitoring data, extracting core operational features corresponding to steady-state operation, dynamic fluctuations, load switching, and temperature shocks, forming a device feature dataset covering multiple operational dimensions. Based on the temporal variation patterns of multi-source operating characteristics, we conduct continuous analysis of operating conditions, identify characteristic change boundaries and operating condition switching nodes across the entire time range, and distinguish between continuous steady-state operating condition intervals and operating condition transition intervals. Based on the continuous distribution patterns of operating conditions, we perform precise temporal segmentation of standardized monitoring data, defining time intervals with consistent operating conditions and continuous evolution of operating characteristics as independent operating condition segments. We then perform operating condition attribute labeling and invalid segment removal on all operating condition segments, forming standardized device operating condition segment data with clear temporal boundaries, unique operating condition attributes, and complete operating characteristics. This provides standardized basic analysis units for subsequent stress evolution analysis.

[0020] Step S2: Analyze the migration path of stress under operating conditions based on the device operating condition segment data, and generate stress migration path data under operating conditions; In this embodiment of the invention, the migration path analysis of operating condition stress is performed based on device operating condition segment data. Each independent operating condition segment is used as the basic analysis unit to analyze the inherent driving mechanism of operating condition switching between different segments, uncovering the driving characteristics of operating condition transitions caused by load changes, temperature fluctuations, environmental disturbances, and mechanical shocks. The driving types and intensity distribution patterns of the full-time operating condition switching are identified, forming complete information on the driving correlation of operating condition transitions. Based on the driving characteristics of operating condition transitions, full-domain operating condition migration nodes are calibrated. The time-series locations where the operating condition state undergoes a substantial switch and the stress level changes effectively are determined as core migration nodes. Simultaneously, the annotations of the preceding and following operating condition attributes and driving characteristics of each migration node are improved, constructing a hierarchical and well-organized device operating condition migration node system. Multi-domain operating characteristic analysis is performed on all operating condition migration nodes, extracting the electrothermal stress, electrical stress, and mechanical stress state information at each node location. Combined with the stress change trends of temporally adjacent nodes, stress evolution correlation analysis between nodes is conducted, establishing a temporally progressive and stress-continuous operating condition stress migration connection topology. Chain-like aggregation is performed on discrete node stress connection relationships to integrate adjacent connection units with consistent temporal sequence and stress evolution trends into complete stress migration chains, distinguishing between single-dimensional independent stress migration chains and multi-dimensional coupled stress migration chains. The temporal linkage, amplitude coupling, and trend synergy characteristics between different stress migration chains are further analyzed to clarify the synchronous evolution, lagging evolution, and indirect coupling evolution laws among multiple chains. Based on the chain association characteristics, the topology of the entire stress migration chain is reconstructed and temporally aligned, redundant branches are merged, and the main evolution path is sorted out to form working condition stress migration path data that is temporally continuous, dimensionally comprehensive, and whose evolutionary logic closely matches the actual stress change law of the device. This provides a complete stress evolution basis for subsequent abnormal stress identification and damage propagation analysis.

[0021] Step S3: Perform damage propagation trajectory analysis on the stress migration path data under the working condition to generate damage propagation trajectory data under the working condition; In this embodiment of the invention, damage propagation trajectory analysis is performed on the stress migration path data under operating conditions. Multi-source stress increment feature mining is conducted on the complete stress migration time-series path to analyze the temporal increment variation patterns of thermal stress, electrical stress, and mechanical stress across the entire time-series. The differentiated characteristics of steady-state small increments, dynamic increments during operating condition switching, and abrupt increments due to disturbances are distinguished, and the dynamic evolution characteristics of stress at different operating stages are analyzed. Based on the fluctuation patterns and evolution levels of the multi-source stress increment features, the location and definition of abnormal stress activation regions across the entire operating condition are carried out. Time-series regions where multi-dimensional stress synchronously rises and incremental evolution deviates from the normal steady-state range are defined as abnormal stress activation regions. The types of regions are distinguished between single-stress abnormal activation and multi-stress coupled abnormal activation, and the collection and classification of abnormal regions across the entire time-series are completed. Refined stress state analysis is performed on each abnormal stress activation region to quantify the stress evolution intensity, duration, rate of change, and coupling degree of different abnormal regions. Different levels of abnormal stress operating states are classified, and the evolution trend and damage induction potential of abnormal stress in each region are clarified. Dynamic diffusion process analysis is conducted for all abnormal stress regions to deduce the extension and expansion of abnormal stress across time intervals, characterizing the evolution patterns of independent diffusion of single stress and synergistic diffusion of multiple stresses. The temporal coverage and boundary distribution characteristics of the stress diffusion intervals across the entire domain are compared to analyze the overlapping and confluence patterns of different diffusion intervals, quantifying the coupling enhancement effect brought about by the superposition of stresses in multiple regions. Combining the abnormal stress state level and stress diffusion overlap characteristics, the damage initiation and diffusion process of device material fatigue and structural aging induced by abnormal stress is deduced layer by layer, distinguishing between independent damage evolution and superimposed accelerated damage evolution modes. The initiation location, diffusion range, cumulative intensity, and evolution trend of damage under all time-series operating conditions are comprehensively summarized, ultimately generating temporally continuous, mechanism-matched, and hierarchically clear damage propagation trajectory data, providing accurate damage evolution basis for device lifetime degradation evolution analysis.

[0022] Step S4: Based on the damage propagation trajectory data under operating conditions, perform device operating condition life degradation evolution analysis to generate device operating condition life degradation evolution data; establish a prediction mapping relationship model between device operating conditions and life degradation using the device operating condition life degradation evolution data to generate a device operating condition-life degradation prediction model. In this embodiment of the invention, based on the damage propagation trajectory data under operating conditions, the unit decomposition, feature mapping, contribution quantification, and lifetime degradation evolution modeling of power device operating condition damage are completed, constructing a complete operating condition-lifetime degradation prediction system. Standardized damage propagation unit division is performed on continuous operating condition damage propagation trajectories. Based on the gradient of damage increment changes, damage evolution trends, and the switching positions of dominant damage mechanisms, time-series segmentation is completed, forming independent damage propagation units with multiple differentiated characteristics, achieving discrete and standardized decomposition of the continuous damage evolution process. In-depth damage degradation mapping feature analysis is conducted on each damage propagation unit, analyzing the damage propagation dependency and transmission relationships between units, the specific differences in the impact of unit damage, and the uncertainty characteristics of damage propagation caused by operating condition fluctuations. By integrating correlation patterns, specific characteristics, and uncertainty correction characteristics, a precise mapping feature system between damage state and device lifetime degradation state is constructed. Based on the global damage degradation mapping features, a structured matrix of multi-dimensional degradation contribution factors is constructed. The lifetime degradation contribution degree of different damage characteristics and different damage units is structured, quantified, and hierarchically distinguished, clarifying the weight ratio and mechanism of action of various damage factors on the overall aging degradation of the device. Based on the contribution matrix, this study analyzes the synergistic relationships of damage degradation across all operating conditions, traces the sources of cross-correlation factors, and resolves degradation behavior constraints. It identifies the synergistic effects of multiple damages, the coupling influence of multiple factors, and the degradation constraints imposed by device material fatigue saturation and stress tolerance boundaries. Combining synergistic gain characteristics and boundary constraint characteristics, the study performs a comprehensive correction of the basic degradation contribution results, enabling a refined extrapolation of the device's full-time-series lifespan degradation evolution and obtaining continuous lifespan degradation evolution information throughout the device's entire lifecycle. Based on complete lifespan degradation evolution data, a stable predictive mapping relationship is established between operating parameters, stress evolution characteristics, damage propagation characteristics, and lifespan degradation states. This leads to the construction of a multi-input, interpretable, and adaptable device operating condition-lifespan degradation prediction model. The model can adapt to degradation prediction needs in various operating conditions, including steady-state, fluctuating, and shock scenarios, achieving a stable mapping output from front-end operating condition data to back-end lifespan degradation states.

[0023] Step S5: Obtain the target power device operating cycle data; transmit the target power device operating cycle data to the device operating condition-lifetime degradation prediction model for intelligent lifetime evaluation of the target power device, and generate target power device lifetime evaluation data.

[0024] In this embodiment of the invention, comprehensive operational data of the target power device throughout its actual operating cycle is collected. The data dimensions are consistent with those of historical training and analysis data, fully covering the time-series operational information of the target device's real-time electrothermal operation, electrical load status, mechanical operation status, and external environment operation status, ensuring dimensional compatibility between the target data and the model input system. The full-cycle operational data of the target device is integrated into a mature device condition-lifetime degradation prediction model. Internally, the model performs a chain-like processing of the target condition data according to standardized processing logic, including time-series normalization, feature extraction, condition segment division, stress evolution analysis, and damage trajectory deduction, replicating the complete analysis mechanism and evolutionary logic of historical data. Based on a built-in degradation contribution weight system, multi-factor coupling correction mechanism, and degradation boundary constraint mechanism, the model quantifies the current damage accumulation level and lifetime degradation degree of the target device in each time interval, and deduces the current remaining lifespan and subsequent degradation trend of the target device by combining the full life-cycle degradation evolution law. The model output includes the degradation stage distribution, damage accumulation gradient, lifetime degradation quantification level, and current lifetime status evaluation conclusions for the target device throughout its entire operating cycle, forming a lifetime evaluation result for the target device under actual service conditions. It generates lifetime evaluation data for target power devices covering real-time degradation status, staged degradation characteristics, and overall lifetime level, enabling intelligent and refined evaluation of the lifetime status of power devices under complex operating conditions.

[0025] Furthermore, step S1 includes the following steps: Step S11: Obtain historical multi-source operation monitoring data of power devices; In this embodiment of the invention, targeted collection of full-dimensional historical operational monitoring data is performed on power devices in power electronic equipment. The collected data types cover multi-source heterogeneous operating condition data during the operation of power devices, specifically including core operating parameters such as device junction temperature, case temperature, on-state voltage drop, switching frequency, load current, bus voltage, operating power consumption, ambient temperature and humidity, and vibration amplitude. The data collection cycle matches the actual industrial operating sequence of the power devices, and long-term continuous collection is achieved by relying on monitoring terminals deployed at the equipment maintenance end. The collection objects cover the same model of power devices with different service years, different load conditions, and different operating environments, thereby constructing a historical dataset with comprehensive operating condition coverage, temporal continuity, and sample diversity. All collected multi-source operational monitoring data are bound to the power device equipment number, operating time period, and equipment operating scenario attributes, realizing accurate association between individual data and device operating conditions. This provides the original data source for subsequent operating condition segment analysis, feature mining, and life degradation analysis, ensuring the integrity of the basic data and the adaptability of the operating conditions for subsequent data fusion analysis.

[0026] Step S12: Perform time-series identification and monitoring data preprocessing on historical power device multi-source operation monitoring data to obtain standard device multi-source operation monitoring data; In this embodiment of the invention, historical multi-source operation monitoring data of power devices undergoes time-series identification and monitoring data preprocessing. First, time-series identification is performed, assigning a unique time-series label to each heterogeneous monitoring data point based on a unified time scale. This aligns the acquisition timelines of different parameters, eliminating time-series misalignment caused by inconsistent acquisition frequencies of multi-source parameters and achieving time-series synchronization of all monitoring data. Based on this, multi-dimensional data preprocessing is performed. Missing data, abnormal jump data, and redundant / repeated data in the original monitoring data are cleaned in layers, removing invalid interference data generated during device start-up / shutdown and equipment maintenance. Simultaneously, multi-source data dimension unification is achieved. For operating parameters with different dimensions such as temperature, current, voltage, frequency, and vibration, an extreme value normalization algorithm is used to complete data mapping and transformation. Through the entire process of time-series alignment, data cleaning, and dimension unification, standard multi-source operation monitoring data of devices with a unified format, regular time sequence, and valid operating conditions are output.

[0027] Step S13: Perform multi-source operation characteristic analysis on the multi-source operation monitoring data of standard devices to generate multi-source operation characteristic data of devices; In this embodiment of the invention, multi-source operational characteristic analysis is performed on the multi-source operational monitoring data of standard devices. The standard monitoring data is divided into four major feature dimensions: device electrothermal characteristics, electrical operational characteristics, environmental condition characteristics, and mechanical operational characteristics. Electrothermal characteristics include junction temperature fluctuation amplitude, case temperature steady-state average, and temperature rise / fall rate. Electrical characteristics include current load rate, voltage fluctuation range, and switching loss fluctuation. Environmental characteristics include ambient temperature change gradient and ambient humidity fluctuation range. Mechanical characteristics include device operating vibration frequency and vibration amplitude fluctuation. A multi-source feature correlation analysis model is constructed. The model uses the original monitoring data of each dimension as input samples and classifies the device into two operating states: steady-state operation and dynamic fluctuation operation. A correlation coefficient algorithm is used to calculate the correlation degree between each parameter. Based on the correlation coefficient, redundant features are eliminated, core features are aggregated, and integrated to obtain multi-source operational characteristic data that comprehensively characterizes the operating state of power devices.

[0028] Step S14: Perform continuous analysis of operating conditions based on multi-source operating characteristic data of the device to generate continuous operating condition data; In this embodiment of the invention, relying on the aggregated multi-source operating characteristic data of devices, a continuous analysis of the operating state under all time-series conditions is carried out to achieve the determination of the state boundaries and continuous characterization of the operating conditions of power devices. First, based on the time-series fluctuation patterns of the multi-source operating characteristic data, the state abrupt boundary and steady-state duration interval of each dimension of operating characteristics are identified, distinguishing four basic operating conditions: steady-state operation, light load fluctuation, heavy load impact, and abnormal disturbance. Using a time-series sliding window as the analysis carrier, a time-series analysis window of fixed duration is set, and the full time-series characteristic data is traversed segment by segment. The fluctuation amplitude, duration, and trend of each operating characteristic within the window are statistically analyzed to determine the device operating condition state corresponding to a single window. When the operating condition states of adjacent time-series windows are consistent, the corresponding operating segment is determined to be a continuous operating condition interval; when the operating condition states of adjacent time-series windows switch, the continuity of the operating condition is determined to be interrupted, and the operating condition state switching node is marked. By analyzing the synergistic change patterns of multi-dimensional operational characteristics, the continuity of operating conditions in different operating segments is quantified. Ultimately, continuous operating condition data that can accurately divide continuous operating condition intervals and operating condition switching nodes is generated, providing a core basis for subsequent operating condition segmentation.

[0029] Step S15: Divide the standard device multi-source operation monitoring data into device operation condition segments using continuous operation status data to generate device operation condition segment data.

[0030] In this embodiment of the invention, continuous operating condition data is used as the dividing benchmark to perform time-series segmentation of multi-source operating monitoring data of standard devices, thereby completing the precise division of operating condition segments for power devices. Using the operating condition switching node as the segmentation threshold, the full-time continuous standard monitoring dataset is segmented, defining continuous time-series data segments with consistent operating conditions between adjacent switching nodes as single, independent device operating condition segments. Each segment possesses unified operating condition attributes; within a single segment, there are no state transitions in the device's electrothermal, electrical, environmental, and mechanical operating states, only normal parameter fluctuations within the steady-state range. Clear differences in operating conditions exist between different segments. Simultaneously, each segment is bound to a corresponding operating condition type, runtime, and core feature attribute label, completing the structured annotation of the segment data and eliminating transient disturbance segments that are too short or lack effective operating condition representation. Through the above-mentioned segmentation, attribute labeling, and invalid segment screening process, the device operation condition segment data with independent time sequence, unique operating condition attributes, and clear state characteristics is finally generated. This realizes the transformation of the original continuous monitoring data into discrete standard operating condition segments, providing standardized basic data units for subsequent stress migration path analysis.

[0031] Furthermore, step S14 includes the following steps: Step S141: Based on the device multi-source operation characteristic data, perform the change boundary identification processing of the device multi-source operation characteristics to generate device multi-source operation characteristic change boundary data; In this embodiment of the invention, relying on the constructed multi-source operating characteristic data of the device, the quantitative identification of the temporal change boundaries of multi-dimensional operating characteristics is completed, and the fluctuation inflection points and interval boundaries of each operating characteristic of the power device within the entire service time series are identified. The core operating characteristics of four dimensions—electrothermal, electrical, environmental, and mechanical—are used as the analysis objects, targeting each type of time-continuous feature sequence. Based on the full-time-series characteristic data, the gradient calculation of the features at each time step is completed, obtaining the temporal gradient change sequence of each feature. Timing nodes where the feature gradient values ​​undergo significant jumps are defined as candidate boundary nodes for feature changes. All candidate boundary nodes are aggregated and screened, and the number of synchronous jumps of multi-dimensional features at a single node is counted. Timing nodes with coordinated changes of multiple features are determined as valid feature change boundaries, while invalid candidate nodes caused by slight fluctuations in a single feature are eliminated. All valid boundary nodes within the entire time series are summarized, and the temporal position, jump feature type, and feature change amplitude attribute corresponding to each boundary node are recorded. This data is integrated to form multi-source operating characteristic change boundary data of the device covering all operating characteristic fluctuation intervals and abrupt change nodes, providing a boundary determination basis for subsequent refined analysis of operating conditions.

[0032] Step S142: Analyze the operating status of the standard device multi-source operation monitoring data by using the boundary data of the device multi-source operation characteristic change, and generate device operating status data. In this embodiment of the invention, the operating conditions of standard devices are analyzed by using boundary data of multi-source operating characteristics of devices to analyze the operating conditions of multi-source operating monitoring data. Using the boundary nodes of characteristic changes as segmentation nodes, the full-time monitoring data is divided into several characteristic stable intervals and characteristic abrupt change intervals. Operating condition matching analysis is performed for each interval. A multi-feature coupled operating condition discrimination model is constructed. The model input consists of a combination of device junction temperature fluctuations, load current levels, switching operating states, environmental parameters, and vibration parameters within each time interval. The model discrimination dimensions include four core operating condition states: steady-state operation, light load fluctuations, heavy load impacts, and abnormal disturbances. The model relies on a massive historical set of labeled operating condition samples to solidify feature thresholds and discrimination rules. The sample set includes standard operating condition data of power devices under different load levels, different temperature environments, and different operating durations. The classification and determination of operating condition states for each time interval are completed through multi-dimensional feature combination matching. Characteristic stable intervals correspond to steady-state or conventional fluctuation conditions with smooth device parameter fluctuations and no stress abrupt changes. Characteristic abrupt change intervals correspond to operating condition switching states caused by device load switching, sudden temperature changes, or external disturbances. After completing the labeling of the operating conditions of the full time series data segment by segment, the operating condition types, interval durations, and core operating condition characteristics of all time series intervals are integrated to generate device operating condition status data with continuous time series and complete status labeling, providing a standardized status data source for continuous operating condition analysis.

[0033] Step S143: Perform a continuity analysis of the device operating condition status data to generate continuous operating condition status data.

[0034] In this embodiment of the invention, the operating condition status data of the device is analyzed for continuity. Using the temporal progression direction as the main analysis line, the operating condition status types of adjacent time intervals are compared and matched one by one. Continuity judgment rules are set: time intervals where the operating condition status types of adjacent time intervals are consistent and there are no abrupt changes in features at the interval transitions are classified as continuous operating condition intervals; locations where the operating condition status types of adjacent time intervals switch or where there are multiple feature-related abrupt changes at the transitions are identified as continuity breakpoints. Simultaneously, a duration constraint mechanism for operating condition status is introduced, statistically analyzing the duration of each continuous operating condition interval, retaining valid continuous operating condition intervals that meet the basic duration threshold, and merging fragmented intervals of the same state formed by short-term operating condition fluctuations to avoid interference from instantaneous parameter fluctuations on the overall operating condition continuity judgment. The entire process relies on multi-dimensional operating condition characteristics and status labeling results to make comprehensive judgments, quantify the continuous attributes of operating conditions in each operating segment, mark all operating condition switching breakpoints and continuous operating condition intervals, and generate continuous operating condition data containing the distribution of continuous operating condition intervals throughout the time series and the location of operating condition breakpoints, providing core judgment support for the accurate division of upper-level operating condition segments.

[0035] Furthermore, step S2 includes the following steps: Step S21: Perform conversion drive relationship analysis on the device operating condition segment data to generate operating condition conversion drive data; In this embodiment of the invention, a transformation-driven relationship analysis is performed on the device operating condition segment data. The operating condition segments cover various standardized operating condition types, including steady-state light load, steady-state heavy load, dynamic ramp-up, load drop, temperature disturbance, and environmental disturbance. Each operating condition segment has an independent time series interval and characteristic attributes. A quantitative analysis model driven by operating condition transformation is constructed. The model uses the core feature differences between pairs of adjacent operating condition segments as input variables. For example, the input variables include three types of core parameters: electrothermal feature difference, electrical load feature difference, and environmental disturbance feature difference. The model calculation structure is as follows: ,in Let be the driving intensity for the transition from the i-th working condition segment to the j-th working condition segment. This represents the difference in temperature characteristics between segments. This represents the difference in load current characteristics between segments. This represents the difference in environmental characteristics between segments. The weighting coefficients for the corresponding feature dimensions are obtained by statistically analyzing the feature contribution ratios of all historical operating condition samples. The model iterates through all temporally adjacent operating condition segments, calculating the driving intensity and type corresponding to each operating condition switch. It distinguishes between three main conversion mechanisms: load-driven, thermal stress-driven, and environmental disturbance-driven, and statistically analyzes the frequency and intensity distribution of pairwise transitions between different operating conditions. Based on the driving attributes, driving intensity, and timing sequence of all operating condition transitions, it integrates to form complete operating condition transition driving data covering all temporal operating condition switching behaviors. This data includes the driving factor composition, driving intensity value, operating condition type before and after the transition, and timing position for each operating condition transition, providing a quantitative basis for the structured design of subsequent operating condition migration nodes.

[0036] Step S22: Design the operating condition migration node based on the operating condition transition driving data to obtain the device operating condition migration node data; In this embodiment of the invention, operating condition migration nodes are designed based on operating condition transition driving data. The marked operating condition switching time sequence positions in the driving data serve as candidate base nodes. Effective migration nodes are selected by combining driving strength values, retaining operating condition switching positions with driving strengths higher than the general level as core migration nodes, and eliminating invalid switching points caused by minor parameter fluctuations without substantial operating condition changes. All selected operating condition migration nodes are annotated with attribute dimensions, binding node time sequence coordinates, preceding operating condition type, subsequent operating condition type, node driving factor composition, and node driving strength attribute to each individual migration node. Operating condition migration hierarchy classification rules are constructed, dividing all migration nodes into three levels according to different driving types: electrical load migration nodes, thermal stress migration nodes, and environmental disturbance migration nodes. Nodes of the same type are grouped to form same-dimensional operating condition migration sequences, while nodes of different types form cross-dimensional migration topology points. Based on the full-time-series operating condition distribution pattern, intermediate reference nodes are supplemented for long-time steady-state operating condition intervals, achieving node-based representation of long-term steady-state operation processes and avoiding migration path discontinuity problems caused by lack of node coverage in long steady-state intervals. All categorized and hierarchically divided working condition migration nodes are aggregated, and the node attribute fields and organizational structure are unified to form device working condition migration node data containing node location, node type, drive attribute, and correlation between previous and subsequent working conditions. This provides standardized node units for subsequent stress migration connection relationship analysis between multiple nodes.

[0037] Step S23: Analyze the migration connection relationship of operating stress based on the device operating condition migration node data, and generate operating stress migration connection relationship data; In this embodiment of the invention, the migration connection relationship of operating condition stress is analyzed based on device operating condition migration node data. This completes the reconstruction of multi-domain operational characteristics of each operating condition migration node, extracts multi-dimensional parameters of instantaneous electrical stress, thermal stress, and mechanical stress corresponding to each migration node, and forms a single-node multi-domain stress feature set to characterize the comprehensive stress state of the power device at the node location. A time-series adjacent node stress evolution correlation model is constructed, with the following model structure: ,in Let be the rate of stress change from the j-th migration node to the k-th migration node. These are the combined stress quantification values ​​for two adjacent nodes, respectively. The model uses the temporal coordinates of the corresponding nodes. Stress changes are calculated for each temporally adjacent node pair, yielding the rate and direction of stress evolution between nodes. Based on the direction of stress change, three connection attributes are identified: stress accumulation connection, stress release connection, and stable stress transfer connection. Effective stress migration connections are established for node pairs with continuous temporal relationships and positive correlations in stress evolution. Connections between nodes with abrupt stress changes, disordered patterns, or no continuous evolution are disconnected, completing the topological connection screening for all nodes. The model records the first and last node numbers, stress change rate, stress evolution type, and temporal span information for each effective connection, integrating these to form a complete set of stress migration connection relationship data covering all effective node connections. This provides a topological foundation for the subsequent construction of combined stress migration chains.

[0038] Step S24: Perform stress migration chain combination processing on the stress migration connection relationship data under working conditions to generate stress migration chain data under working conditions; In this embodiment of the invention, stress migration connection relationship data under operating conditions is processed by stress migration chain combination. Using a temporal progression as the core combination logic and a single stress migration connection as the basic unit, multiple adjacent connections with interconnected start and end nodes and a continuous stress evolution trend are chained together to form a preliminary temporal stress migration chain. Continuity verification is performed on the completed preliminary chain. For segmented structures with short-term steady-state intervals in the middle of the chain, smooth transition connections are added based on the stress stability characteristics of the steady-state intervals, achieving the integration of segmented chains. For multiple parallel stress migration links of different dimensions, three independent chain systems are distinguished: electrical stress migration chain, thermal stress migration chain, and mechanical stress migration chain, achieving chain-specific characterization of multi-dimensional stress migration behavior. A chain segmentation threshold is set to reasonably segment ultra-long chains with extremely large temporal spans and multiple operating condition transitions in the middle, ensuring the overall consistency of the stress evolution trend of a single migration chain. After completing the chain combination, classification, and segmentation optimization of all connecting units, multiple time-ordered, dimensionally clear, and evolution-continuous stress migration chains are formed. All chains uniformly store the chain start and end time sequence, included node sequence, stress evolution trend, and chain dimension attributes, generating structured stress migration chain data, providing chain-level analysis units for subsequent chain relationship mining.

[0039] Step S25: Perform stress migration chain correlation analysis based on the stress migration chain data under the working conditions, and generate stress migration chain correlation data under the working conditions; In this embodiment of the invention, the correlation analysis of stress migration chains under working conditions is performed based on the stress migration chain data. Four core chain features are extracted for each stress migration chain: time-series coverage interval, stress change amplitude, evolution rate, and peak stress location. A migration chain correlation calculation model is then constructed. For example, ,in Let m be the correlation between the m-th migration chain and the n-th migration chain. For chain temporal overlap, For stress amplitude matching degree, For evolution rate similarity, The chain association weight coefficient is determined by statistical analysis of historical multi-stress coupling working condition samples. The model iterates through all migration chain combinations of different dimensions and time series, calculating the association degree between each pair of chains. Chains with an association degree higher than a set level are identified as coupled chains, forming a multi-chain linkage structure. Three association types are distinguished: synchronous association, delayed association, and indirect association. Synchronous association represents the coordinated evolution of stress in multiple domains within the same time series; delayed association represents the delayed change of stress in other dimensions triggered by the evolution of a single stress; and indirect association represents multiple chains forming indirect coupling through common working condition nodes. The associated objects, association types, association strengths, and temporal linkage characteristics of all chains are recorded and integrated to form complete working condition stress migration chain association data.

[0040] Step S26: Reconstruct the migration path of the stress under the working condition using the associated data of the stress migration chain to generate stress migration path data.

[0041] In this embodiment of the invention, the migration path of stress under operating conditions is reconstructed by associating stress migration chain data with operating condition stress. Using each stress migration chain as the basic framework and the chain relationships as fusion constraints, multi-dimensional migration chains with coupling relationships are topologically superimposed and temporally aligned, eliminating temporal misalignment and topological fragmentation between chains of different dimensions. For synchronously associated chains, a parallel fusion approach is used to construct multi-stress collaborative migration paths; for lagging associated chains, stress progressive evolution migration paths are constructed according to temporal sequence; for indirectly associated chains, paths are connected and spliced ​​through common operating condition migration nodes. Path intersections and branch structures occurring during the reconstruction process are standardized and streamlined, retaining the main paths with continuous stress evolution characteristics and merging redundant branch paths with consistent trends. After completing the global path splicing and streamlining, stress evolution characteristics are labeled for the entire migration path, recording the distribution and temporal intervals of stress accumulation, stress stabilization, and stress abrupt change stages throughout the path, clarifying the overall migration trend and evolution law of power device operating condition stress throughout its entire lifecycle. By integrating all reconstructed global path topology, path stage characteristics, and multi-stress coupling evolution information, we generate working condition stress migration path data that is time-complete, dimensionally comprehensive, and logically continuous, providing complete stress evolution foundation data for subsequent working condition damage propagation trajectory analysis.

[0042] Furthermore, step S23 includes the following steps: Step S231: Perform multi-domain feature analysis on the device condition migration node data to generate multi-domain feature data of device condition migration node operation. In this embodiment of the invention, multi-domain feature analysis of device condition migration node data is performed to mine multi-dimensional operational features of a single node in the electrothermal, electrical, mechanical, and environmental domains. Device condition migration nodes include three core node types: electrical load migration nodes, thermal stress migration nodes, and environmental disturbance migration nodes. Different node types correspond to key condition switching moments during device operation. For each condition migration node, multi-domain core operational features of the corresponding time-series cross-section and the adjacent steady-state range are extracted. Electrothermal domain features include the average junction temperature, junction temperature change rate, case temperature distribution characteristics, and heat accumulation characteristics. Electrical domain features include load current amplitude, bus voltage fluctuation, switching loss level, and conduction impedance characteristics. Mechanical domain features include device operating vibration amplitude and vibration frequency distribution characteristics. Environmental domain features include ambient temperature, ambient humidity, and airflow disturbance characteristics. For a single operating condition migration node, the preprocessed standardized feature components of the electrothermal, electrical, mechanical, and environmental domains are assigned corresponding feature weights. A weighted summation operation is then performed on these four dimensional features to obtain the comprehensive multi-domain feature quantification result for the node. The weight configuration for each feature dimension is determined based on statistical fitting of a sample set of multi-domain operating condition feature contributions throughout the power device's lifecycle. This sample set covers device operating characteristic data under different load conditions and temperature environments, and the weight values ​​match the actual contribution ratio of each dimension feature to the device's operating condition status. During operation, the original domain features and the weighted fused comprehensive features of each node are uniformly retained. By comparing the magnitudes of the weights of the four dimensions, the dominant and secondary auxiliary feature dimensions of each migration node are determined. Weak and stray features with excessively low amplitudes and no operating condition characterization value are filtered out, generating multi-domain feature data for the device operating condition migration node with complete dimensions, hierarchical features, and clear attributes.

[0043] Step S232: Calculate the stress change of time-adjacent operating condition migration nodes based on the multi-domain feature data of the device operating condition migration nodes, and generate operating condition migration node stress change data. In this embodiment of the invention, stress changes of temporally adjacent working condition migration nodes are calculated based on multi-domain feature data of device working condition migration nodes. Using the temporal progression order as the sorting criterion, all working condition migration nodes are temporally sorted to construct a continuous temporal node sequence, and pairs of adjacent nodes are matched to form temporally adjacent node pairs. For each pair of adjacent nodes, three core stress components—thermal stress, electrical stress, and mechanical stress—are extracted and decomposed. The amplitude and evolution rate of the changes in different stress dimensions within the node's temporal interval are calculated separately, avoiding the analytical bias caused by calculations of a single stress dimension. First, the differences in multi-domain features between adjacent nodes are used to convert the thermal stress change, electrical stress change, and mechanical stress change into three single-dimensional stress change parameters. Then, the three single-dimensional stress change parameters are squared, the squared results are accumulated, and the square root of the accumulated results is performed to obtain the comprehensive stress change between adjacent nodes. This achieves the coupling and fusion of multi-dimensional stress changes, objectively reflecting the overall working condition stress fluctuation level under the synergistic effect of multiple stresses. The calculation process combines the time intervals between nodes to synchronously calculate the stress evolution rate over time, comprehensively characterizing the magnitude and speed of stress changes. It statistically analyzes the stress increase / decrease states, overall stress change level, and stress evolution trend of each pair of adjacent nodes, distinguishing between three basic change states: continuous stress accumulation, slow stress release, and stable stress transition. It fully records the stress evolution parameter information of each pair of temporally adjacent nodes, summarizing the stress calculation results of all node pairs to form stress change data covering all temporal nodes and multiple stress dimensions of the working condition migration nodes. This provides quantitative support for determining and analyzing the connection relationships of working condition stress migration.

[0044] Step S233: Analyze the migration connection relationship of working condition stress based on the stress change data of the working condition migration nodes, and generate working condition stress migration connection relationship data.

[0045] In this embodiment of the invention, the migration connection relationship of stress under working conditions is analyzed based on the stress change data of the migration nodes. The stress change state of temporally adjacent nodes is used as the core judgment criterion. Combined with the multi-domain stress co-evolution law, standardized stress migration connection judgment rules are established. For adjacent node pairs with continuous changes in comprehensive stress and unified multi-dimensional stress evolution trends, effective working condition stress migration connection relationships are constructed. For node pairs with abrupt and disordered stress changes and contradictory multi-dimensional stress evolution trends, they are judged as having no effective migration association, and invalid topological connections are blocked. Based on the specific characteristics of stress changes between nodes, effective migration connections are classified into types: node connections with continuously increasing stress are defined as stress accumulation migration connections; node connections with continuously decreasing stress are defined as stress release migration connections; and node connections with small stress fluctuations and overall stability are defined as stress stability migration connections. The correspondence between the first and last nodes, stress change characteristics, connection types, and temporal span attributes of all effective migration connections are analyzed to construct a full-domain working condition stress migration topological network, realizing the structured integration of stress associations of discrete working condition migration nodes. Invalid connections with extremely short temporal spans and extremely low stress change levels are simultaneously eliminated to simplify the topological network structure. Integrate the complete attribute information of all valid connections to form working condition stress migration connection relationship data that covers the entire time series, multiple dimensions, and complete topological logic.

[0046] Furthermore, step S3 includes the following steps: Step S31: Perform stress multi-source incremental characteristic analysis on the migration path based on the stress migration path data under the working condition, and generate stress multi-source incremental characteristic data of the migration path. In this embodiment of the invention, the stress migration path is analyzed for multi-source incremental stress characteristics based on the stress migration path data. This data includes a full-time stress migration node sequence, multi-dimensional stress evolution trends, temporal distribution of stress accumulation and release, and multi-domain stress coupling correlation information. Based on this, the analysis carrier is used to separate the temporal evolution sequences of three core stress dimensions: thermal stress, electrical stress, and mechanical stress. A multi-source stress incremental temporal analysis model is constructed, with the following structure: ,in Let be the multi-source stress increment eigenvector at time t. This represents the time increment of thermal stress. For electrical stress time-series increment, The mechanical stress time-series increments are calculated by converting the stress parameter differences between adjacent time-series nodes. Model training and feature fitting are completed using a time-series sample set of stress under long-term service conditions of power devices. The sample set covers stress migration time-series data under different load switching frequencies, different temperature shock intensities, and different mechanical vibration environments. The model calculates and collects features of multi-dimensional stress increments node by node, simultaneously extracting the continuous evolution duration, increment fluctuation law, multi-dimensional increment synergy, and increment peak distribution characteristics of stress increments. It distinguishes three types of increment forms: steady-state small increments, dynamic increments during operating condition switching, and abrupt increments due to external disturbances, thus fully characterizing the dynamic evolution characteristics of each time-series stage in the stress migration path. The multi-dimensional stress increment parameters, feature attributes, and evolution laws of all time-series intervals are summarized and integrated to form time-continuous, dimensionally comprehensive multi-source increment feature data of migration path stress, providing refined feature support for subsequent identification of abnormal stress regions.

[0047] Step S32: Perform an analysis of the activated region of abnormal stress under working conditions on the incremental characteristic data of multi-source stress in the migration path, and generate data of activated region of abnormal stress under working conditions. In this embodiment of the invention, an abnormal stress activation region analysis is performed on the multi-source incremental characteristic data of migration path stress. Based on the full-time stress increment characteristic sequence, and combined with the physical service characteristics of power devices, the activation criteria for abnormal stress are clarified. Time-series intervals showing synchronous increases in multi-dimensional stress increments, deviations from the normal steady-state fluctuation range in increment evolution trend, and significant increases in multi-domain stress coupling strength are defined as the basic intervals for the initiation of abnormal stress. An abnormal stress activation discrimination model is constructed, with the following model structure: ,in For abnormal stress activation degree, Let be the magnitude of the multi-source stress increment at time t. This represents the average stress increment modulus under normal operating conditions throughout the entire time series. To activate the weighting coefficients, the coefficients are determined by fitting statistical samples of stress increments under the device's rated operating conditions. The abnormal stress activation degree is calculated sequentially across time intervals using the model. Time intervals with activation degrees exceeding the normal fluctuation range are identified as abnormal stress activation regions. The start and end times, activation intensity, dominant stress dimension, and multi-stress coupling characteristics of each activation region are marked. Adjacent scattered abnormal activation intervals are aggregated and integrated, merging continuous abnormal stress intervals generated during fluctuations under the same operating condition, and separating isolated abnormal regions caused by independent disturbances, thus completing the structured division of all-time-series abnormal stress activation regions. The anomaly type of stress increment in each activation region is simultaneously labeled, including four core types: thermal stress overload activation, electrical stress impact activation, mechanical stress disturbance activation, and multi-stress coupling activation, fully recording the abnormal stress initiation characteristics and evolution basis of each region. All divided and attribute-complete abnormal stress interval information is integrated to generate abnormal stress activation region data for operating conditions.

[0048] Step S33: Perform abnormal stress state analysis based on the data of the activated area of ​​abnormal stress under working conditions, and generate abnormal stress state data under working conditions. In this embodiment of the invention, abnormal stress state analysis is performed based on data from activated regions of abnormal stress under operating conditions. Each activated region is treated as an independent analysis unit, and the multi-source stress increment time-series data, stress coupling characteristics, and increment fluctuation patterns within that region are extracted. The abnormal stress operating state is quantified from four dimensions: stress amplitude, evolution rate, duration, and coupling degree. A quantitative evaluation model for abnormal stress state under operating conditions is constructed, with the following model structure: ,in The quantized values ​​of the abnormal stress state in the time series interval from a to b are given. The magnitude of stress increment within the region. The average rate of stress evolution, The duration of abnormal stress, For multi-dimensional stress coupling degree, Weighting coefficients for each dimension are determined by training on a sample set of stress states from device aging tests. This sample set covers full-gradient stress condition data for mild, moderate, and severe anomalies. The model quantifies the abnormal stress states in each activated region, classifying them into three levels: mild disturbance anomaly, moderate stress shift, and severe stress impact. It also marks the evolution trend of abnormal stress in each region, distinguishing between three evolutionary trends: continuous aggravation, stable maintenance, and gradual attenuation. The model comprehensively collects the stress state level, evolution trend, damage induction potential, and dominant anomaly factors for each anomaly region. Transient, mild anomalies without sustained evolution or damage induction characteristics are removed, while effective abnormal stress state information with device degradation induction effects is retained. This integrated data forms comprehensive, hierarchical, and attribute-complete abnormal stress state data for operating conditions, providing core state parameters for the prediction and analysis of damage propagation trajectories.

[0049] Step S34: Based on the data of the activated region of abnormal stress under working conditions and the data of the state of abnormal stress under working conditions, perform damage propagation trajectory analysis under working conditions to generate damage propagation trajectory data under working conditions.

[0050] In this embodiment of the invention, damage propagation trajectory analysis is performed based on data of abnormal stress activation regions and abnormal stress states under operating conditions. The abnormal stress activation regions serve as the spatial and temporal carriers of damage initiation, while the levels and evolution trends of abnormal stress states in each region act as driving conditions for damage propagation. Combining the damage evolution mechanisms of power devices, including electrothermal fatigue, electrical aging, and mechanical wear, a logic for inferring damage propagation under operating conditions is established. A quantitative model for damage propagation under operating conditions is constructed, with the following structure: ,in Let be the cumulative damage level at time t. This represents the quantized value of the abnormal stress state at time t. The damage propagation coefficient is determined by statistical analysis of test samples, taking into account the fatigue characteristics of power device materials and the structural tolerance characteristics. The time-series extrapolation step size is used. The model takes full-time anomaly stress distribution data as input and extrapolates the initiation, diffusion, and accumulation processes of damage in each time-series interval. Mild anomaly stress corresponds to a slow fatigue damage accumulation process, moderate anomaly stress corresponds to a continuous damage diffusion process, and severe anomaly stress corresponds to a rapid damage aggravation and local damage accumulation process. For anomaly stress intervals with multiple overlapping regions, the synergistic damage effect of multi-dimensional stress is superimposed, and the cumulative damage increment in the overlapping damage areas is statistically analyzed to characterize the damage accumulation and evolution features. The model records the damage initiation time sequence location, damage diffusion coverage area, damage accumulation gradient, and multi-regional damage convergence patterns throughout the entire time-series operation, outlining the propagation direction, diffusion range, and cumulative intensity changes of damage under all time-series conditions, forming a continuous and complete time-series damage evolution context. Finally, the damage propagation characteristics, cumulative damage data, and damage evolution trends from all time-series stages are integrated to generate damage propagation trajectory data for the operating conditions.

[0051] Furthermore, step S34 includes the following steps: Step S341: Analyze the abnormal stress diffusion process of the activated region data under abnormal working conditions to generate abnormal stress diffusion process data. In this embodiment of the invention, the abnormal stress diffusion process is analyzed using data from the activated region of abnormal stress under operating conditions. The activated region data calibrates the initial initiation interval, stress type, and basic activation intensity of abnormal stress across the entire time series. Using this as the core analysis carrier, a time-series dynamic extrapolation mechanism is employed to analyze the stress diffusion behavior. Combining the physical evolution mechanisms of electrothermal conduction, electrical stress coupling, and mechanical stress transfer in power devices, a dynamic diffusion model of abnormal stress is constructed. The model structure is as follows: ,in Let t be the spatiotemporal coverage of the abnormal stress. Let be the gradient change value of the abnormal stress state at time t. The stress diffusion coefficient is... The time-series simulation step size is defined. The diffusion coefficient is obtained by statistical fitting of an experimental sample set containing measured stress diffusion data under different abnormal stress levels and operating conditions. The model iteratively simulates each independent abnormal stress activation region, statistically analyzing the extension duration of abnormal stress in the preceding and following time intervals, the expansion range of the stress coverage area, the peak stress migration location, and the stress attenuation rate during the diffusion process. Two diffusion modes are distinguished: independent diffusion of single stress and coordinated diffusion of multiple stresses. Independent diffusion of single stress represents the autonomous extension and evolution of abnormal stress in a single dimension (thermal, electrical, or mechanical), while coordinated diffusion of multiple stresses represents the evolution of multi-dimensional abnormal stresses coupled and expanding synchronously. The stress coverage area, stress intensity distribution, diffusion start and end points, and diffusion evolution type of each activation region are recorded throughout the entire time-series simulation. Stress diffusion characteristic parameters from all time stages are collected and integrated to form a comprehensive abnormal stress diffusion process data covering all abnormal regions, with continuous time-series evolution and complete diffusion characteristics, providing dynamic evolutionary basis data for subsequent diffusion overlap characteristic analysis.

[0052] Step S342: Analyze the overlap characteristics of abnormal stress diffusion based on the abnormal stress diffusion process data under the working condition, and generate abnormal stress diffusion overlap characteristic data. In this embodiment of the invention, the overlap characteristics of abnormal stress diffusion are analyzed based on the data of the abnormal stress diffusion process under working conditions. Using all dynamically simulated abnormal stress diffusion intervals as analysis units, the entire diffusion interval is systematically aggregated according to its temporal arrangement. The temporal coverage, stress diffusion boundaries, and peak stress distribution locations of different abnormal stress diffusion intervals are compared to identify combinations of diffusion units where intervals intersect. An abnormal stress diffusion overlap calculation model is constructed, with the following structure: ,in The overlap between the i-th and j-th anomalous stress diffusion intervals, The time overlap duration of the two diffusion intervals. The total temporal coverage duration of the two diffusion intervals is calculated. Based on the model, overlap characteristics are quantified for each combination of diffusion intervals. Based on the overlap degree, three overlap states are classified: partial overlap, complete overlap, and adjacent non-overlapping. Simultaneously, two overlap types are distinguished: same-dimensional stress diffusion overlap and cross-dimensional stress diffusion overlap. Same-dimensional overlap is the temporal superposition of multiple anomalous diffusions in a single stress dimension, while cross-dimensional overlap is the temporal convergence and coupling of diffusion intervals in different dimensions of thermal stress, electrical stress, and mechanical stress. The overlap duration, overlapping stress dimension, superimposed stress intensity, and timing of overlap occurrence are statistically analyzed for each overlapping region. The intensity increase law and evolution trend change after stress superposition within the overlapping region are analyzed. Independent diffusion evolution characteristics of non-overlapping regions are collected, comprehensively distinguishing the superimposed evolution and independent evolution differences of stress diffusion across the entire domain. The quantitative parameters, feature types, and evolution laws of all overlapping units are summarized to generate anomalous stress diffusion overlap characteristic data with clear hierarchy and complete parameters, providing coupling characteristic basis for accurate prediction of damage propagation trajectories.

[0053] Step S343: Analyze the damage propagation trajectory of the working condition by using the abnormal stress diffusion overlap characteristic data and the abnormal stress state data of the working condition, and generate the working condition damage propagation trajectory data.

[0054] In this embodiment of the invention, damage propagation trajectory analysis is performed using abnormal stress diffusion overlap characteristic data and abnormal stress state data under operating conditions. The abnormal stress state data defines the basic damage induction capacity of each diffusion region, and the diffusion overlap characteristic data characterizes the damage enhancement effect after multiple stress superpositions, thus establishing a correlation and deduction system between stress state, diffusion coupling, and damage evolution. A multi-coupled operating condition damage propagation trajectory evolution model is constructed, with the following structure: ,in Let t be the total cumulative damage. The basic damage induced by a single abnormal stress is calculated from the quantified value of the abnormal stress state. Let be the stress diffusion overlap at time t. This represents the incremental damage caused by stress overlap and coupling. Model training is based on a multi-stress coupling aging test sample set for power devices to achieve parameter adaptation. The sample set covers damage data under all operating conditions, including single-stress disturbance, dual-stress superposition, and triple-stress coupling. For independent stress diffusion intervals without overlap, the accumulation process of fatigue damage is steadily deduced according to the abnormal stress state of the foundation. For diffusion intervals with overlapping coupling, an overlap parameter is introduced to amplify the superimposed damage effect and characterize the accelerated evolution characteristics of damage accumulation in multi-stress convergence areas. Damage initiation points, damage diffusion boundaries, damage accumulation gradients, and damage accumulation core areas are recorded at each time-series node, outlining the propagation direction, expansion range, and evolution rate changes of damage throughout the entire time series, forming a continuous and complete temporal damage evolution trajectory. The dominant induced stress types and coupling mechanisms at different damage stages are labeled, distinguishing three trajectory forms: steady-state slow damage evolution, superimposed accelerated damage evolution, and peak impact damage evolution. Finally, the full-domain temporal damage evolution parameters, trajectory characteristics, and damage level information are integrated to generate logically complete, mechanism-matched, and quantitatively refined operating condition damage propagation trajectory data.

[0055] Furthermore, as an embodiment of the present invention, reference is made to... Figure 2 As shown, Figure 1 A detailed flowchart illustrating the implementation steps of step S4 is provided in this embodiment. Step S4 includes: Step S41: Divide the working condition damage propagation trajectory data into working condition damage propagation units to generate working condition damage propagation unit data; In this embodiment of the invention, the damage propagation trajectory data under operating conditions is divided into damage propagation units. This data comprehensively records the initiation location, diffusion boundary, cumulative gradient, superposition and accumulation, and temporal evolution trend of abnormal stress-induced damage throughout the entire service life of the power device, including complete temporal information on independent damage evolution stages and multi-stress coupled superposition damage evolution stages. The unit division is based on the temporal continuity of damage evolution, damage growth rate, damage accumulation degree, and dominant stress dimension, employing a damage gradient hierarchical segmentation mechanism to complete the full-domain trajectory cutting and decomposition. A damage unit division judgment model is constructed, with the following model structure: ,in This represents the damage increment between adjacent time-series nodes. , This represents the cumulative damage at corresponding time-series nodes. Independent propagation units are defined based on the continuous variation range of damage increment. Continuous time-series intervals with consistent damage increment trends and no change in the dominant damage mechanism are classified as single damage propagation units. Time-series locations where the damage increment gradient deviates significantly or the dominant stress dimension changes are set as unit segmentation boundaries. Different damage propagation units correspond to differentiated damage evolution states, including three basic unit types: low-speed fatigue accumulation units, medium-speed stress superposition damage units, and high-speed impact damage agglomeration units. Each defined damage propagation unit possesses independent time-series intervals, damage increment characteristics, damage diffusion range, and damage formation mechanism attributes. Unit boundaries are clear, and evolutionary characteristics are not mixed. After completing unit segmentation and type labeling through a full domain traversal, the time-series parameters, damage quantification parameters, and evolutionary attribute parameters of all standardized damage propagation units are collected to form well-structured, hierarchical, and comprehensive damage propagation unit data covering the entire time-series damage process.

[0056] Step S42: Perform condition damage degradation mapping feature analysis based on the condition damage propagation unit data to generate condition damage degradation mapping feature data. In this embodiment of the invention, damage degradation mapping feature analysis is performed based on damage propagation unit data under operating conditions. Using each independent damage propagation unit as the basic analysis carrier, five core damage features are extracted from the unit: damage accumulation rate, damage diffusion coverage, proportion of multi-stress coupled damage, damage overlap and aggregation intensity, and damage duration. Damage degradation mapping feature parsing is then performed, including unit damage degradation mapping feature vectors, damage rate feature components, damage range feature components, proportion of coupled damage feature components, damage aggregation intensity feature components, and damage duration feature components. The simulation is completed using a power device accelerated aging test dataset, which contains device damage evolution data and corresponding lifetime decay data under different operating stress levels and different damage superposition forms. Corresponding relationships between various damage features and lifetime degradation are established. By performing structured extraction and correlation mapping of multi-dimensional damage features unit by unit, three differentiated mapping patterns are distinguished: linear fatigue damage mapping features, nonlinear accelerated degradation mapping features, and steady-state micro-damage accumulation mapping features, characterizing the effects of different damage units on device lifetime decay. We simultaneously remove weak micro-damage unit features that do not contribute effectively to degradation, retain core mapping features that have a real lifespan decay driving effect, collect multi-dimensional mapping features and correlation patterns of all damage propagation units, and integrate them to form working condition damage degradation mapping feature data that is dimensionally complete, accurately correlated, and mechanistically consistent.

[0057] Step S43: Design the contribution factor matrix of working condition life degradation based on the working condition damage degradation mapping feature data to obtain the working condition life degradation contribution matrix data. In this embodiment of the invention, a contribution factor matrix for service life degradation is designed based on the damage degradation mapping feature data. The matrix uses five core feature components from the damage degradation mapping feature data as column dimensions and full-time damage propagation units as row dimensions, thus constructing a two-dimensional contribution matrix architecture. A quantitative model for service life degradation contribution is then built, with the following structure: ,in The normalization contribution weight is assigned to the j-th type of degradation feature of the i-th damage unit. The original degradation contribution is represented by m, where m is the total number of degradation feature dimensions. The original degradation contribution is obtained statistically from device aging test samples. The basic contribution level of each dimension is determined by combining the lifetime decay amplitude under different damage characteristics, achieving an objective quantitative allocation of contribution weights. Each matrix element corresponds to the specific contribution ratio of a single damage unit and a single degradation feature, fully reflecting the impact of different time stages and different damage types on the overall lifetime degradation. The matrix dimensions are normalized to unify the feature dimension arrangement rules of damage units across the entire domain, eliminating dimension misalignment issues caused by differences in time intervals, and ensuring that the matrix has time comparability and feature superpositionability. Based on the contribution weight values, strong contribution degradation factors, medium contribution degradation factors, and weak contribution degradation factors are distinguished, and high-weight core degradation drivers in the matrix are labeled to clarify the dominant damage mechanism in the power device lifetime degradation process. Finally, a dimension-normalized, weight-quantified, mechanism-corresponding, and directly usable operating condition lifetime degradation contribution matrix data is generated, providing a quantitative matrix basis for the full-cycle lifetime degradation evolution prediction of devices.

[0058] Step S44: Perform device operating life degradation evolution analysis on the operating life degradation contribution matrix data to generate device operating life degradation evolution data. In this embodiment of the invention, device lifespan degradation evolution analysis is performed on the lifespan degradation contribution matrix data. Using the lifespan degradation contribution matrix as the basis for quantification weights, the temporal arrangement relationship, damage synergy relationship, and evolutionary constraint relationship of different damage propagation units are integrated to build a global degradation evolution inference framework. A device lifespan degradation evolution model is constructed, with the following structure: ,in This represents the cumulative lifetime degradation of the device over all timings. The overall degradation contribution weight for the i-th damaged unit is obtained by weighting the row vectors of the contribution matrix. Let i be the damage degradation mapping feature of the i-th unit. The model fits data based on long-term service tracking datasets of power devices and multi-condition accelerated aging datasets, covering various degradation scenarios such as single-stress fatigue degradation, multi-stress coupled accelerated degradation, and intermittent impact degradation, as well as the non-uniform degradation characteristics of adapter devices under complex operating conditions. The simulation process considers the synergistic degradation effect of adjacent damaged units; temporally adjacent and similarly characterized damaged units exhibit a degradation superposition effect, while temporally disjointed and contradictory units maintain an independent degradation accumulation relationship. Simultaneously, a degradation behavior constraint mechanism is introduced to limit the excessive iterative accumulation of instantaneous high-stress damage, closely reflecting the actual physical characteristics of power device material fatigue saturation. Degradation accumulation calculations are performed for each time interval, plotting the continuous evolution trend of lifespan degradation throughout the entire service life, distinguishing the evolution characteristics of slow degradation, steady degradation, and accelerated degradation stages, and comprehensively depicting the entire process of device lifespan degradation from steady-state micro-degradation to rapid decay, ultimately generating device lifespan degradation evolution data.

[0059] Step S45: Establish a prediction mapping model between device operating conditions and lifetime degradation using device operating condition lifetime degradation evolution data, and generate a device operating condition-lifetime degradation prediction model.

[0060] In this embodiment of the invention, a predictive mapping model for device operating conditions and lifetime degradation is established using device operating condition and lifetime degradation evolution data. The model input variables are full-time operating condition characteristics, multi-source stress evolution characteristics, and unit damage propagation characteristics, while the quantified lifetime degradation evolution quantity is the model output variable, constructing a multi-input, single-output nonlinear mapping prediction model. The overall model structure includes an operating condition feature input layer, a damage feature fusion layer, a degradation mapping hidden layer, and a lifetime degradation output layer. The damage feature fusion layer uses the aforementioned multi-dimensional damage degradation mapping logic to complete feature fusion, and the degradation mapping hidden layer relies on full-domain degradation evolution data to complete nonlinear relationship fitting. The core mapping relationship of the model satisfies the structural... ,in To predict the amount of lifespan degradation, This is a joint degradation mapping function of operating conditions and damage. It is a set of operating characteristics under multiple sources. This dataset represents a complete time-series set of damage propagation and evolution characteristics. The model training dataset consists of paired samples comprised of device lifecycle monitoring data, damage projection data, and lifetime measurement calibration data. These samples cover complete service scenarios under different load conditions, temperature shocks, and disturbance environments. The training process iteratively optimizes the mapping function parameters by minimizing degradation prediction bias. The converged model can directly infer the real-time lifetime degradation state and subsequent degradation trend of the device based on real-time operating data. After model construction, it possesses the ability to predict lifetime degradation under steady-state, fluctuating, and shock conditions. It can output continuous time-series quantification results of lifetime degradation and degradation stage determination results, ultimately generating a structurally complete, parameter-converged, and mapping-stable device condition-lifetime degradation prediction model, providing core model support for subsequent intelligent lifetime evaluation of target devices.

[0061] Furthermore, step S42 includes the following steps: Step S421: Perform element damage propagation dependency analysis on the element damage propagation data under working conditions to generate element damage propagation dependency data; In this embodiment of the invention, the damage propagation dependency of the working condition damage propagation units is analyzed. These units are arranged chronologically throughout the entire service life of the power device. Each unit corresponds to a different stress loading state and damage accumulation stage. Some units exhibit evolutionary correlations, such as pre-damage inducing subsequent damage and multi-unit chain-like progressive damage. Based on the chronological order as the fundamental correlation dimension, and combined with the dominant stress type, damage increment level, and damage diffusion coverage of each damage unit, a unit damage propagation dependency calculation model is constructed. The model structure is as follows: ,in Let m be the propagation dependency of the m-th damage unit on the n-th subsequent damage unit. This is the unit timing tightness coefficient. The element is the source factor of the dominant stress. This represents the damage progression coefficient between preceding and following units. The weighting coefficients are obtained through statistical fitting of a multi-stage continuous damage aging test sample set of power devices. The model quantifies the dependency of each pair of damage units across the entire domain. Unit combinations with dependencies exceeding a threshold are considered to have valid damage propagation dependencies, forming a chain-like dependency structure where the preceding unit induces the damage, and the following unit responds. Three dependency types are distinguished: strong chain dependency, weak progressive dependency, and unrelated independent evolution. Strong chain dependency indicates that the accumulation of damage in the preceding unit directly accelerates the evolution of damage in the following unit; weak progressive dependency indicates that damage in the preceding unit provides the basic fatigue conditions for the evolution of the following unit; and unrelated independent evolution indicates that unit damage is induced solely by independent operating stress. The model comprehensively records the associated objects, dependency types, dependency strengths, and temporal propagation directions of all damage units, compiling a complete set of unit damage propagation dependency data covering all damage units across the domain, with clear association logic and complete quantitative parameters.

[0062] Step S422: Perform a specificity analysis of the impact of unit damage propagation on the unit based on the unit damage propagation data, and generate specificity data of the impact of unit damage propagation on the unit. In this embodiment of the invention, based on standardized operating condition damage propagation unit data, a specific analysis is conducted on the differentiated impact characteristics of different damage units during the device lifetime degradation process, uncovering the unique effects of damage units with different time sequences and stress types on the overall degradation process. Each damage unit exhibits inherent differences in stress coupling degree, damage accumulation mechanism, operating condition duration, and location of structural damage. The contribution of different units to device aging degradation possesses irreplaceable specific characteristics. A quantitative model of the specificity of unit damage impact is constructed, with the model structure as follows: ,in Let be the influence specificity coefficient of the i-th damage unit. The intensity of the combined damage effect corresponding to the unit. This represents the average damage intensity across all damage units in the entire domain. The comprehensive damage intensity is calculated by combining the unit stress impact level, fatigue accumulation time, number of stress superpositions, and device structural loss response. The conversion benchmark is established based on a comparative test sample set of power device damage under single and combined operating conditions. The model calculates specificity coefficients for each unit, classifying them into three levels based on coefficient values: high-specificity core damage units, medium-specificity conventional damage units, and low-specificity secondary damage units. High-specificity units correspond to damage units formed by heavy-load impacts and multi-stress coupling superpositions; these units significantly accelerate device lifespan degradation and possess unique evolutionary characteristics. Medium-specificity units correspond to conventional steady-state fatigue damage units, with continuous, minor degradation accumulation as their main mode of action. Low-specificity units correspond to weak damage units formed by short-term, slight disturbances, with a relatively low impact on overall degradation. The specific causes, unique degradation modes, and degradation contribution characteristics of each level of unit are labeled. The specificity quantification parameters and attribute characteristics of all units are collected to generate unit damage propagation impact specificity data that is hierarchically clear, characteristically differentiated, and has a well-defined mechanism of action.

[0063] Step S423: Perform damage uncertainty propagation factor analysis on the specific data of damage propagation impact of unit elements, and generate damage uncertainty propagation factor data of unit elements. In this embodiment of the invention, damage uncertainty propagation factor analysis is performed on the specific data of unit damage propagation impact under operating conditions. Dynamic switching of operating conditions, random stress fluctuations, and small disturbances in environmental parameters during the actual service of power devices can cause fluctuations in the actual degradation effect of similar damaged units, forming the uncertainty characteristic of damage propagation. A solution model for the damage uncertainty propagation factor of operating conditions units is constructed, and the model structure is as follows: ,in Let be the uncertainty propagation factor of the i-th damage unit. The damage intensity of the same type of unit under different operating condition fluctuation samples. The average damage intensity of units of the same type is represented by N, which is the total number of damage samples under the same operating conditions. The sample set covers device aging test data under different operating conditions with varying fluctuation amplitudes, time-series loading intervals, and environmental disturbances under the same damage mechanism, ensuring the statistical representativeness of the uncertainty factor. The uncertainty factor value of each damage unit is calculated through the model. The higher the factor value, the more significant the impact of operating condition disturbances on the damage propagation process of the corresponding unit, and the stronger the fluctuation characteristics of damage evolution; the lower the factor value, the more stable the unit damage evolution process, and the more repeatable the degradation law. Based on the uncertainty factor, stable damage propagation units, fluctuating damage propagation units, and random disturbance damage propagation units are distinguished. The source type, fluctuation amplitude, and propagation disturbance characteristics of uncertainty for each unit are labeled, and integrated to form a unified quantitative dimension of operating condition unit damage uncertainty propagation factor data covering all damage units, providing uncertainty correction parameters for the accurate fusion of final degradation mapping characteristics.

[0064] Step S424: Perform condition damage degradation mapping feature analysis based on the unit damage propagation dependency data and the condition unit damage uncertainty propagation factor data to generate condition damage degradation mapping feature data.

[0065] In this embodiment of the invention, damage degradation mapping feature analysis is performed based on unit damage propagation dependency data and working condition unit damage uncertainty propagation factor data. The chain dependency relationship between damaged units serves as a structural constraint, clarifying the feature transmission law between preceding and subsequent units, and achieving temporal correlation fusion of discrete unit features. Unit-specific features are used as the basis for differentiated weights to distinguish the core degradation contribution features and secondary degradation contribution features of different units. The uncertainty propagation factor is used as a deviation correction parameter to compensate for the disturbance deviation caused by dynamic fluctuations in the working condition on the damage degradation mapping relationship. A multi-constraint degradation mapping feature fusion model is constructed, with the following model structure: ,in Let be the final degradation mapping feature of the i-th damaged unit. The unit specificity coefficient, This represents the propagation dependency of the preceding unit on the current unit. For the mapping feature quantity of the preceding associated unit, This represents the uncertainty propagation factor of the current unit. The model relies on the associated samples of damaged units across the entire domain and the aging samples of operating condition fluctuations to complete parameter adaptation, taking into account the progressive transmission law of damage, the unit-specific influence characteristics, and the uncertainty of operating condition disturbances. The model output features simultaneously represent the independent degradation characteristics of units, the chain-like progressive degradation characteristics between units, and the true degradation characteristics after correction for operating condition disturbances, avoiding the mapping bias caused by a single feature extraction method. The fused mapping features of all damaged units are collected and organized into operating condition damage degradation mapping feature data that is time-continuous, mechanistically complete, accurately corrected, and adapted to complex operating condition fluctuations, for use in the design and calculation of the subsequent operating condition life degradation contribution factor matrix.

[0066] Furthermore, step S44 includes the following steps: Step S441: Perform a collaborative relationship analysis of the working condition damage degradation mapping feature data to generate collaborative relationship data of working condition damage degradation. In this embodiment of the invention, a collaborative relationship analysis of condition damage degradation mapping feature data is performed. This data encompasses multi-dimensional structured features of each damage propagation unit, including damage rate characteristics, damage range characteristics, coupled damage ratio, damage aggregation intensity, and damage duration. Different feature dimensions exhibit mutually constraining and mutually promoting collaborative evolution patterns during device service. The continuous development of damage in a single dimension will drive synchronous changes in other dimensions of degradation features. A multi-dimensional feature linkage analysis mechanism is employed to construct a quantitative model for the collaborative degradation of condition damage. The model structure is as follows: ,in Let be the synergy coefficient between the x-th type of degradation feature and the y-th type of degradation feature. The covariance of the two types of degenerate feature sequences, This represents the variance of the corresponding degradation feature sequence. Model training and coefficient calibration rely on a multi-condition coupled aging test dataset for power devices. The dataset contains long-term aging samples with multiple dimensions of thermal damage, electrical damage, and mechanical damage superimposed, covering full-scenario data of single damage evolution, dual-damage coupled evolution, and multi-damage synergistic evolution. The model calculates the synergy degree of all pairwise combinations of degradation features. Based on the sign and numerical range of the synergy coefficient, three types of synergy relationships are distinguished: positive synergy, negative synergy, and weak synergy. Positive synergy represents the synchronous growth and mutual promotion of two types of degradation features; negative synergy represents the growth of one type of feature accompanied by a decline in the growth rate of another type of feature; and weak synergy represents a low correlation between the evolution of two types of features. The model fully records the synergy objects, synergy types, synergy strengths, and temporal synergy intervals for each feature dimension, sorts out the dominant damage degradation synergy patterns in different service stages, and collects data on condition damage degradation synergy relationships that cover all dimensions of degradation features, have clear temporal correlations, and have well-defined coupling rules, providing synergy mechanism support for subsequent analysis of cross-correlation factors.

[0067] Step S442: Based on the data on the synergistic relationship between working condition damage and degradation, perform cross-correlation analysis on the evolution of working condition damage and degradation factors, and generate cross-correlation data on the evolution of working condition damage and degradation factors. In this embodiment of the invention, the evolutionary cross-correlation factors of working condition damage degradation are analyzed based on the data on the synergistic relationship between working condition damage and degradation. The synergistic evolution of working condition damage degradation is not caused by a single factor, but is the result of the combined effects of working condition stress switching, damage time sequence superposition, device material fatigue accumulation, and multi-domain stress coupling. The influence weights of different cross-correlation factors on the degradation synergistic law vary significantly. A quantitative model of the cross-correlation factors of working condition damage degradation is constructed, and the model structure is as follows: ,in Let be the cross-correlation contribution of the k-th type of influencing factor. These are the weighting coefficients for the corresponding factors. The coefficients of the corresponding feature pairs are the coordination coefficients. The weighting coefficients are determined by fitting the device's full lifecycle tracking test dataset, which records device degradation evolution parameters under different operating conditions and stress superposition modes. This dataset accurately reflects the influence of various factors on the degradation synergy process. Cross-correlated factors are categorized into four core dimensions: load fluctuation factors, thermal stress cycle factors, mechanical vibration superposition factors, and multi-stress coupling lag factors. The contributing or constraining effects of each factor on the synergistic evolution of characteristics are quantified. Three types of interaction are distinguished: time-synchronous cross-correlation, time-lag cross-correlation, and indirect transmission cross-correlation. Time-synchronous cross-correlation is synergistic degradation caused by the simultaneous action of multiple factors; time-lag cross-correlation is synergistic degradation induced by the accumulation of early operating conditions; and indirect transmission cross-correlation is cross-dimensional synergistic degradation formed by a single factor through damage chain transmission. The dataset comprehensively collects the intensity, timing, scope of influence, and combination of related characteristics of various cross-correlated factors, generating data on cross-correlation factors in operating condition damage degradation evolution that are clearly dimensional, mechanistic, and quantitatively complete.

[0068] Step S443: Analyze the constraint relationship of working condition damage degradation behavior based on the working condition damage degradation mapping feature data, and generate working condition damage degradation behavior constraint relationship data. In this embodiment of the invention, the constraint relationships of damage degradation behavior under operating conditions are analyzed based on the characteristic data of damage degradation mapping under operating conditions, clarifying the boundary constraints, saturation constraints, and mechanism constraints in the device degradation evolution process. Damage degradation of power devices is not an unlimited linear accumulation; material fatigue threshold, upper limit of structural stress tolerance, operating condition stress loading limit, and weak damage self-healing effect all constrain the degradation behavior. The constraint strength and constraint dimension differ at different operating condition stages. A quantitative model of damage degradation constraints under operating conditions is constructed, with the following model structure: ,in Let be the degradation constraint coefficient of the i-th damaged element. For the degenerate mapping feature quantity of the current unit, The maximum tolerance threshold for degradation characteristics under the corresponding operating conditions of the device is defined, and the threshold parameter is obtained statistically from the accelerated aging limit test dataset of the device. The degradation constraint coefficient is calculated unit by unit through the model. A larger constraint coefficient indicates a stronger constraint on the degradation behavior at the corresponding stage, and a more gradual slowdown in the rate of damage accumulation; a smaller constraint coefficient indicates a weaker degradation constraint at the corresponding stage, and the damage can maintain a continuous accelerating accumulation trend. Three core constraint relationships are distinguished: steady-state constraint, overload constraint, and fatigue saturation constraint. Steady-state constraint corresponds to weak natural constraints under normal operating conditions, overload constraint corresponds to strong boundary constraints under high-stress conditions, and fatigue saturation constraint corresponds to the slowdown in the rate of degradation growth in the later aging stages of the device. The constraint type, constraint coefficient, constraint range, and adjustment law on the rate of degradation growth are labeled for each time-series damage unit. The temporal evolution characteristics of the degradation constraint intensity throughout the entire service life are analyzed, and a set of operating condition damage degradation behavior constraint relationship data covering all damage units, with clear constraint mechanisms and complete quantitative parameters is compiled, providing boundary constraint criteria for subsequent degradation evolution prediction.

[0069] Step S444: Analyze the device's operating life degradation evolution by using the cross-correlation factor data of operating condition damage degradation evolution and the constraint relationship data of operating condition damage degradation behavior on the operating life degradation contribution matrix data, and generate device operating life degradation evolution data.

[0070] In this embodiment of the invention, the device's lifespan degradation evolution is analyzed using data on cross-correlation factors of damage degradation evolution and data on constraint relationships of damage degradation behavior, based on the contribution matrix data of lifespan degradation. The lifespan degradation contribution matrix serves as the quantification basis, retaining the basic contribution weights of each damage unit and degradation feature. The degradation superposition increment caused by multi-feature coupling is corrected based on cross-correlation factor data, and the degradation reduction caused by device fatigue saturation and stress tolerance boundaries is corrected based on degradation behavior constraint data, achieving a three-dimensional fusion deduction of basic contribution, coupling increment, and boundary constraints. An integrated model of device lifespan degradation evolution is constructed, with the following structure: ,in This represents the cumulative lifetime degradation of the device. The degradation contribution weight of the i-th element in the contribution matrix. For unit degeneracy mapping feature quantity, The comprehensive contribution value of the corresponding cross-correlation factors of the unit. This represents the unit degradation constraint coefficient. Analyzing the actual service characteristics of the device, the constraint effect is weakened and the multi-factor coupling boosting effect is strengthened in the low-damage stage, while fatigue saturation constraint is strengthened and excessive degradation accumulation is suppressed in the high-aging stage, avoiding the evolution distortion problem caused by linear accumulation extrapolation. The model completes iterative calculation of degradation amount for each unit throughout the entire time series, integrating the degradation results of discrete units to form a continuous lifetime degradation time series curve. This distinguishes the differentiated evolution characteristics of the initial micro-degradation stable stage, the mid-term coupled accelerated degradation stage, and the late-stage constraint-type slow degradation stage, comprehensively characterizing the non-uniform degradation law of the power device throughout its entire life cycle. Finally, device operating condition lifetime degradation evolution data is generated.

[0071] Therefore, the embodiments should be considered as exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of the equivalents of the application are intended to be included within the invention.

[0072] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. A lifetime assessment method based on power device operating condition data fusion, characterized in that, Includes the following steps: Step S1: Obtain historical multi-source operation monitoring data of power devices; perform device operation condition segment analysis based on historical multi-source operation monitoring data of power devices to generate device operation condition segment data; Step S2: Analyze the migration path of stress under operating conditions based on the device operating condition segment data, and generate stress migration path data under operating conditions; Step S3: Perform damage propagation trajectory analysis on the stress migration path data under the working condition to generate damage propagation trajectory data under the working condition; Step S4: Based on the damage propagation trajectory data under operating conditions, perform device operating condition life degradation evolution analysis to generate device operating condition life degradation evolution data; establish a prediction mapping relationship model between device operating conditions and life degradation using the device operating condition life degradation evolution data to generate a device operating condition-life degradation prediction model. Step S5: Obtain the target power device operating cycle data; transmit the target power device operating cycle data to the device operating condition-lifetime degradation prediction model for intelligent lifetime evaluation of the target power device, and generate target power device lifetime evaluation data.

2. The lifetime assessment method based on power device operating condition data fusion according to claim 1, characterized in that, Step S1 includes the following steps: Step S11: Obtain historical multi-source operation monitoring data of power devices; Step S12: Perform time-series identification and monitoring data preprocessing on historical power device multi-source operation monitoring data to obtain standard device multi-source operation monitoring data; Step S13: Perform multi-source operation characteristic analysis on the multi-source operation monitoring data of standard devices to generate multi-source operation characteristic data of devices; Step S14: Perform continuous analysis of operating conditions based on multi-source operating characteristic data of the device to generate continuous operating condition data; Step S15: Divide the standard device multi-source operation monitoring data into device operation condition segments using continuous operation status data to generate device operation condition segment data.

3. The lifetime assessment method based on power device operating condition data fusion according to claim 2, characterized in that, Step S14 includes the following steps: Step S141: Based on the device multi-source operation characteristic data, perform the change boundary identification processing of the device multi-source operation characteristics to generate device multi-source operation characteristic change boundary data; Step S142: Analyze the operating status of the standard device multi-source operation monitoring data by using the boundary data of the device multi-source operation characteristic change, and generate device operating status data. Step S143: Perform a continuity analysis of the device operating condition status data to generate continuous operating condition status data.

4. The lifetime assessment method based on power device operating condition data fusion according to claim 1, characterized in that, Step S2 includes the following steps: Step S21: Perform conversion drive relationship analysis on the device operating condition segment data to generate operating condition conversion drive data; Step S22: Design the operating condition migration node based on the operating condition transition driving data to obtain the device operating condition migration node data; Step S23: Analyze the migration connection relationship of operating stress based on the device operating condition migration node data, and generate operating stress migration connection relationship data; Step S24: Perform stress migration chain combination processing on the stress migration connection relationship data under working conditions to generate stress migration chain data under working conditions; Step S25: Perform stress migration chain correlation analysis based on the stress migration chain data under the working conditions, and generate stress migration chain correlation data under the working conditions; Step S26: Reconstruct the migration path of the stress under the working condition using the associated data of the stress migration chain to generate stress migration path data.

5. The lifetime assessment method based on power device operating condition data fusion according to claim 4, characterized in that, Step S23 includes the following steps: Step S231: Perform multi-domain feature analysis on the device condition migration node data to generate multi-domain feature data of device condition migration node operation. Step S232: Calculate the stress change of time-adjacent operating condition migration nodes based on the multi-domain feature data of the device operating condition migration nodes, and generate operating condition migration node stress change data. Step S233: Analyze the migration connection relationship of working condition stress based on the stress change data of the working condition migration nodes, and generate working condition stress migration connection relationship data.

6. The lifetime assessment method based on power device operating condition data fusion according to claim 1, characterized in that, Step S3 includes the following steps: Step S31: Perform stress multi-source incremental characteristic analysis on the migration path based on the stress migration path data under the working condition, and generate stress multi-source incremental characteristic data of the migration path. Step S32: Perform an analysis of the activated region of abnormal stress under working conditions on the incremental characteristic data of multi-source stress in the migration path, and generate data of activated region of abnormal stress under working conditions. Step S33: Perform abnormal stress state analysis based on the data of the activated area of ​​abnormal stress under working conditions, and generate abnormal stress state data under working conditions. Step S34: Based on the data of the activated region of abnormal stress under working conditions and the data of the state of abnormal stress under working conditions, perform damage propagation trajectory analysis under working conditions to generate damage propagation trajectory data under working conditions.

7. The lifetime assessment method based on power device operating condition data fusion according to claim 6, characterized in that, Step S34 includes the following steps: Step S341: Analyze the abnormal stress diffusion process of the activated region data under abnormal working conditions to generate abnormal stress diffusion process data. Step S342: Analyze the overlap characteristics of abnormal stress diffusion based on the abnormal stress diffusion process data under the working condition, and generate abnormal stress diffusion overlap characteristic data. Step S343: Analyze the damage propagation trajectory of the working condition by using the abnormal stress diffusion overlap characteristic data and the abnormal stress state data of the working condition, and generate the working condition damage propagation trajectory data.

8. The lifetime assessment method based on power device operating condition data fusion according to claim 1, characterized in that, Step S4 includes the following steps: Step S41: Divide the working condition damage propagation trajectory data into working condition damage propagation units to generate working condition damage propagation unit data; Step S42: Perform condition damage degradation mapping feature analysis based on the condition damage propagation unit data to generate condition damage degradation mapping feature data. Step S43: Design the contribution factor matrix of working condition life degradation based on the working condition damage degradation mapping feature data to obtain the working condition life degradation contribution matrix data. Step S44: Perform device operating life degradation evolution analysis on the operating life degradation contribution matrix data to generate device operating life degradation evolution data. Step S45: Establish a prediction mapping model between device operating conditions and lifetime degradation using device operating condition lifetime degradation evolution data, and generate a device operating condition-lifetime degradation prediction model.

9. The lifetime assessment method based on power device operating condition data fusion according to claim 8, characterized in that, Step S42 includes the following steps: Step S421: Perform element damage propagation dependency analysis on the element damage propagation data under working conditions to generate element damage propagation dependency data; Step S422: Perform a specificity analysis of the impact of unit damage propagation on the unit based on the unit damage propagation data, and generate specificity data of the impact of unit damage propagation on the unit. Step S423: Perform damage uncertainty propagation factor analysis on the specific data of damage propagation impact of unit elements, and generate damage uncertainty propagation factor data of unit elements. Step S424: Perform condition damage degradation mapping feature analysis based on the unit damage propagation dependency data and the condition unit damage uncertainty propagation factor data to generate condition damage degradation mapping feature data.

10. The lifetime assessment method for power device operating condition data fusion according to claim 8, characterized in that, Step S44 includes the following steps: Step S441: Perform a collaborative relationship analysis of the working condition damage degradation mapping feature data to generate collaborative relationship data of working condition damage degradation. Step S442: Based on the data on the synergistic relationship between working condition damage and degradation, perform cross-correlation analysis on the evolution of working condition damage and degradation factors, and generate cross-correlation data on the evolution of working condition damage and degradation factors. Step S443: Analyze the constraint relationship of working condition damage degradation behavior based on the working condition damage degradation mapping feature data, and generate working condition damage degradation behavior constraint relationship data. Step S444: Analyze the device's operating life degradation evolution by using the cross-correlation factor data of operating condition damage degradation evolution and the constraint relationship data of operating condition damage degradation behavior on the operating life degradation contribution matrix data, and generate device operating life degradation evolution data.

Images