An automated full performance evaluation method for industrial control computing devices in a scenario simulation

Abstract

The present application relates to a kind of industrial control computing device full performance evaluation method of automation scene simulation, belong to automation measurement and control and equipment performance evaluation technical field.The method includes: parsing design specification document, extracting performance benchmark parameter and boundary examination condition, and constructing equipment full performance cross-domain correlation graph;Extract the node with physical field coupling failure relationship and carry out multi-stress superposition, generate simulation scene containing multi-domain synchronous loading and compile as multiple test units collaborative execution script;Collect state and performance parameters under the condition of multi-domain superposition scene, extract degradation characteristic quantity and carry out cross-domain coupling degradation modeling, and construct cross-domain performance degradation response surface model;Based on the model, trace analysis is carried out along the direction of graph edge to lock root cause node, and full performance grade is evaluated, and the evaluation result is fed back to graph to complete graph weight updating.Full performance evaluation dynamic closed loop is realized, and the precision of equipment degradation traceability and performance evaluation under the condition of multi-physical field coupling is improved.

Description

Technical Field

[0001] This invention belongs to the field of automated measurement and control and equipment performance evaluation technology, specifically relating to a method for full performance evaluation of industrial control computing equipment using automated scenario simulation. Background Technology

[0002] With the rapid development of complex industrial systems and highly integrated electronic devices, these devices often face the combined effects of multiple physical fields, including electrical stress, transient overload, fluctuating ambient temperature, and electromagnetic interference, during actual operation. To accurately assess the reliability and degradation characteristics of such devices under complex operating conditions, existing automated measurement and control systems typically employ scenario simulation methods involving sequential loading of single stresses or simple superposition of discrete stresses. In existing technologies, different test execution units are usually driven sequentially to obtain the device's performance response under a specific single environment, and phased reliability tests and evaluations are conducted based on discrete test sequences and static models.

[0003] However, existing technologies suffer from the following serious problems and shortcomings when facing multi-physics cross-domain coupling degradation assessment: First, existing test scenario simulations lack time axis overlap and follow-up control mechanisms during multi-stress synchronous loading, failing to realistically reproduce induced secondary failures (such as sudden mechanical thermal stress changes caused by electrical heating and electromagnetic leakage caused by deformation) generated under multi-physics interaction, leading to distorted scenario simulations. Second, at the data processing level, existing technologies typically average or isolate the collected multi-domain degradation parameters, making it difficult to effectively isolate strong collinearity between parameters and accurately construct a hyperspace response surface that maps complex nonlinear characteristics. Third, existing performance evaluation methods usually employ static, isolated, hard-coded index weighted models, unable to dynamically adjust evaluation weights according to actual test scenarios; simultaneously, there is a lack of dynamic feedback and update mechanisms between the evaluation results and early design evaluation networks, preventing the test system from adaptively evolving topological relationships as actual failure data accumulates, thus limiting the overall accuracy of performance evaluation and degradation tracing. Summary of the Invention

[0004] To address the aforementioned problems in the existing technology, this invention provides a method for full performance evaluation of industrial control computing equipment for automated scenario simulation. The objective of this invention can be achieved through the following technical solutions: S1: Analyze the design specification document of the industrial control computing device under test, extract the performance benchmark parameters and boundary assessment conditions of the industrial control computing device under test in different functional domains, and construct a cross-domain correlation graph of the device's full performance with the performance benchmark parameters as nodes and the coupling failure relationship between different functional domains as edges; S2: Extract nodes with physical field coupling failure relationships from the cross-domain correlation graph of the full performance of the equipment, perform multi-stress superposition on the corresponding boundary assessment conditions, generate a multi-domain superposition simulation scenario that includes synchronous loading of electrical, environmental and electromagnetic stresses, and compile the loading sequence and duration of the test conditions in the multi-domain superposition simulation scenario into a multi-test unit collaborative execution script. S3: Real-time acquisition of the operating status parameters and performance response parameters of the industrial control computing device under test in a multi-domain superimposed simulation scenario; time-frequency domain separation of the performance response parameters; extraction of degradation feature quantities of each performance index dimension from the separated performance response parameters; cross-domain coupled degradation modeling; construction of a cross-domain performance degradation response surface model. S4: Based on the cross-domain performance degradation response surface model, perform coupling degradation source analysis along the direction of each physical field coupling failure relationship in the cross-domain correlation graph of the equipment's full performance to obtain the root cause node of performance degradation. Based on the degradation amount of the root cause node, evaluate the overall performance level of the equipment and feed the evaluation results back to the cross-domain correlation graph of the equipment's full performance to update the physical field coupling failure weights between each index.

[0005] Specifically, the extraction process of the performance benchmark parameters and boundary assessment conditions includes: The design specification document is semantically segmented at the chapter level using a semantic parsing model. The segmented text paragraphs are categorized into electrical functional domain, physical environment tolerance domain, and electromagnetic compatibility domain. Performance index names and corresponding benchmark value ranges are extracted from the text paragraphs corresponding to each functional domain. The test stress types and corresponding stress intensity ranges are also extracted from the text paragraphs corresponding to each functional domain. The performance index names, benchmark value ranges, test stress types, and stress intensity ranges extracted within the same functional domain are then structurally associated and stored to form boundary test conditions.

[0006] Specifically, the process of constructing the cross-domain correlation graph of the device's full performance includes: Using the extracted performance benchmark parameters as topology skeleton nodes, multidimensional degradation correlation data of performance indicators of different functional domains under joint test conditions are extracted in the multi-stress coupling test channel. The multidimensional degradation correlation data is input into a preset Bayesian network topology operator to carry out cross-domain causal inference analysis, and the causal transmission paths of the sudden change in mechanical thermal stress of the pin induced by electrical overload heating and the electromagnetic leakage of the gap induced by mechanical vibration deformation are identified. Asymmetric unidirectional directed edges are connected between each topological skeleton node along the evolution direction of the causal transmission path, and the failure propagation slope of the directed edges is calculated as the coupling failure weight to construct the cross-domain correlation graph of the device's full performance.

[0007] Specifically, the multi-physics coupling superposition includes: Specific nodes marked with physical field coupling failure relationships in the cross-domain correlation graph of the full performance of the equipment are retrieved, and the boundary assessment conditions of the corresponding nodes are extracted through the independent hardware channel of the unified scheduling controller. The stress-sensitive timing sequence of the industrial control computing equipment under test in typical failure modes is extracted, and the loading order of electrical stress, environmental stress and electromagnetic stress is dynamically adjusted and switched according to the stress-sensitive timing sequence. The unified scheduling controller is driven to issue follow-up control commands to multiple groups of external test execution units, and the superposition of overload voltage, temperature alternation and spatial multi-band electromagnetic interference is completed on a unified spatiotemporal axis to form a multi-domain superposition simulation scenario.

[0008] Specifically, the multi-test-unit collaborative execution script includes: An object-oriented modular compilation approach is adopted to encapsulate the loading timing and duration parameters of each test condition in the multi-domain superimposed simulation scenario. Within the compilation architecture, a test triggering module for triggering hardware channels, a timing control module for controlling the slope of multiple stress fields, and an exception rollback module for sudden transient short circuit protection are independently constructed. The compiled control fields are distributed to multiple sets of test execution units arranged in parallel, and each test execution unit is bound to a unified reference clock synchronization mechanism within the local area network to maintain coordinated operation, generating a multi-test unit coordinated execution script containing multi-level overload strategies.

[0009] Specifically, the execution of the preset security policy by the exception rollback module in the multi-test unit collaborative execution script includes: When the multi-test unit detects a transient overcurrent or waveform distortion anomaly in the industrial control computing device under test during the synchronous execution of the script, the anomaly rollback module forcibly suspends the current hardware test channels through the hardware interrupt controller. The pre-stored stepped load reduction control algorithm is retrieved, and the unified clock synchronization mechanism is driven to send a synchronization peak reduction command to each test execution unit. Within a preset microsecond time window, the electrical overload amplitude and temperature alternation slope are proportionally reduced to the safety margin range. Initiate a closed-loop stress retry scheduling mechanism to drive each hardware test channel to restart and execute unfinished test fields within the safety margin range, and write the interrupt protection trace back to the policy log in the local memory.

[0010] Specifically, the time-frequency domain separation includes: The real-time acquired performance response parameters are input into the continuous wavelet transform and short-time Fourier transform modules in the preset computing environment. The continuous wavelet transform module is called to perform multi-scale decomposition on the non-stationary transient response components in the performance response parameters to obtain time-domain nonlinear abrupt change characteristics. Simultaneously, the short-time Fourier transform module is used to perform sliding window spectrum analysis on the performance response parameters to lock the steady-state periodic response components. The energy distribution and variance change rate of the corresponding components are calculated, and the corresponding degradation features are extracted to complete the time-frequency domain separation, providing input for subsequent cross-domain coupled degradation modeling. The non-stationary transient response components include transient current and impact mechanical strain, and the steady-state periodic response components include periodic electromagnetic interference.

[0011] Specifically, the construction process of the cross-domain performance degradation response surface model is as follows: A hierarchical topology architecture comprising a feature fusion layer, an interaction fitting layer, and a boundary convergence layer is constructed. The feature fusion layer imports the extracted degraded feature sequence into a regression matrix constructed by fusing partial least squares regression and gradient boosting decision trees, performing cross-domain recombination to eliminate strong collinearity and output cross-domain multiphysics parameters. The interaction fitting layer performs multivariate nonlinear fitting on the cross-domain multiphysics parameters to solve the response surface equation, calculating the interaction term coefficients characterizing the cross-interference between stresses, and generating a hyperspace response surface. The boundary convergence layer extracts the system's dynamic load noise and injects it as a boundary constraint into the hyperspace response surface, constructing a cross-domain performance degradation response surface model.

[0012] Specifically, the process of constructing the regression matrix includes: The partial least squares regression is used to project the input degenerate feature sequence into the principal component space to remove collinearity interference and output a decoupled feature matrix. The decoupled feature matrix is ​​used as a training sample and input into the gradient boosting decision tree. Multi-level cascaded weak learners are used to fit the residuals of multi-field coupling degenerate along the negative gradient direction. The output weights of the leaf nodes of the gradient boosting decision tree are fused with the mapping vector of the principal component space projection to construct the regression matrix.

[0013] Specifically, the coupling degradation source analysis includes: Using the performance degradation characteristic calculated by the cross-domain performance degradation response surface model as the starting point for tracing, a path backtracking algorithm based on graph search is invoked to retrieve the historical signal disturbance characteristics of each topological node in reverse cross-domain tracing along the direction of the coupling failure edge with the largest topological weight value in the cross-domain correlation graph of the device's full performance. Simultaneously, the local disturbance sensitivity matrix is ​​calculated for the electrical, environmental, and electromagnetic parameters of the industrial control computing device under test. The local sensitivity gradient of each topological node is compared along the backtracking path to lock the topological source node with the highest amplification coefficient and the largest deviation as the root cause node, thus completing the coupling degradation source tracing analysis.

[0014] Specifically, the evaluation process for the full performance level of the equipment includes: The degradation amount of the root cause node and the degradation feature amount of each performance index dimension are obtained, and the coupling sensitivity of the root cause node in the cross-domain correlation graph of the overall performance of the device is calculated. The degradation amount and the coupling sensitivity are correlated and mapped to determine the performance degradation level of the industrial control computing device under test in the corresponding functional domain. The current actual application scenario of the industrial control computing device under test is obtained, and the weight coefficient ratio of the corresponding functional domain is dynamically adjusted according to the actual application scenario. The performance degradation level of each functional domain and the adjusted weight coefficient are weighted and comprehensively calculated to evaluate the overall performance level.

[0015] Specifically, the update of the device full-performance cross-domain correlation graph includes: The evaluation results of the overall performance level assessment of the equipment are converted into the actual degradation amount of each performance index. The actual degradation amount is compared with the current physical field coupling failure weight of each coupling failure relationship in the cross-domain correlation map of the overall performance of the equipment to obtain the mismatch path between the actual degradation amount and the weight. Reverse topological mapping is performed along the direction of the coupling failure relationship corresponding to the mismatch path. The physical field coupling failure weight of each coupling failure relationship is corrected based on the mapping result.

[0016] The beneficial effects of this invention are as follows: (1) This invention breaks through the limitations of single stress discrete loading in traditional testing by constructing a full-performance cross-domain correlation graph and introducing a unified scheduling controller and independent hardware channels. Based on the stress-sensitive timing sequence of the typical failure modes of the industrial control computing equipment under test, the system dynamically adjusts the loading sequence of electrical, environmental and electromagnetic stresses, and completes the synchronous superposition of overload voltage, temperature alternation and multi-band electromagnetic interference on a unified spatiotemporal axis. It can realistically induce and reproduce cross-domain secondary failures such as sudden changes in pin thermal stress caused by electrical heating, thus improving the realism of the scenario simulation. At the same time, through the abnormal rollback module in the script executed by multiple test units, when transient overcurrent or waveform distortion abnormalities are captured, the hardware interrupt controller is used to forcibly suspend the test channel and perform microsecond-level synchronous peak clipping. While ensuring the safety of high-risk multi-stress joint testing, it greatly improves the dynamic defense capability of complex automated measurement and control processes.

[0017] (2) This invention utilizes partial least squares regression and gradient boosting decision tree fusion to construct a regression matrix. Principal component spatial projection and residual fitting are performed on the degraded feature sequence after time-frequency domain separation, effectively eliminating strong collinearity among multi-physics parameters and achieving high-precision decoupling construction of the hyperspace response surface model. Based on this, the system, through path backtracking algorithm and local perturbation sensitivity matrix calculation, can accurately reverse-locate the topological source node with the highest amplification coefficient, significantly improving the root cause tracing accuracy. More importantly, this invention overcomes the limitations of traditional static weighted rating, converting the rating results into actual degradation amounts and comparing them with predicted weights. It extracts mismatched paths and performs reverse topological mapping, dynamically correcting the failure weights in the graph. This allows the entire full-performance evaluation system to adaptively evolve its topological relationships as actual failure data samples accumulate, achieving a dynamic feedback loop for full-performance evaluation. Attached Figure Description

[0018] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings.

[0019] Figure 1 This is a flowchart illustrating a method for full performance evaluation of industrial control computing equipment using automated scenario simulation, as described in this invention. Figure 2 This is a structural block diagram of a method for full performance evaluation of industrial control computing equipment for automated scenario simulation, as described in this invention. Detailed Implementation

[0020] To further illustrate the technical means and effects of the present invention in achieving the intended purpose, the following detailed description of the specific implementation methods, structures, features and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided.

[0021] Please see Figures 1-2A method for full performance evaluation of industrial control computing equipment for automated scenario simulation, comprising: S1: Analyze the design specification document of the industrial control computing device under test, extract the performance benchmark parameters and boundary assessment conditions of the industrial control computing device under test in different functional domains, and construct a cross-domain correlation graph of the device's full performance with the performance benchmark parameters as nodes and the coupling failure relationship between different functional domains as edges; S2: Extract nodes with physical field coupling failure relationships from the cross-domain correlation graph of the full performance of the equipment, perform multi-stress superposition on the corresponding boundary assessment conditions, generate a multi-domain superposition simulation scenario that includes synchronous loading of electrical, environmental and electromagnetic stresses, and compile the loading sequence and duration of the test conditions in the multi-domain superposition simulation scenario into a multi-test unit collaborative execution script. S3: Real-time acquisition of the operating status parameters and performance response parameters of the industrial control computing device under test in a multi-domain superimposed simulation scenario; time-frequency domain separation of the performance response parameters; extraction of degradation feature quantities of each performance index dimension from the separated performance response parameters; cross-domain coupled degradation modeling; construction of a cross-domain performance degradation response surface model. S4: Based on the cross-domain performance degradation response surface model, perform coupling degradation source analysis along the direction of each physical field coupling failure relationship in the cross-domain correlation graph of the equipment's full performance to obtain the root cause node of performance degradation. Based on the degradation amount of the root cause node, evaluate the overall performance level of the equipment and feed the evaluation results back to the cross-domain correlation graph of the equipment's full performance to update the physical field coupling failure weights between each index.

[0022] Specifically, the extraction process of the performance benchmark parameters and boundary assessment conditions includes: The design specification document is semantically segmented at the chapter level using a semantic parsing model. The segmented text paragraphs are categorized into electrical functional domain, physical environment tolerance domain, and electromagnetic compatibility domain. Performance index names and corresponding benchmark value ranges are extracted from the text paragraphs corresponding to each functional domain. The test stress types and corresponding stress intensity ranges are also extracted from the text paragraphs corresponding to each functional domain. The performance index names, benchmark value ranges, test stress types, and stress intensity ranges extracted within the same functional domain are then structurally associated and stored to form boundary test conditions.

[0023] Specifically, the process of constructing the cross-domain correlation graph of the device's full performance includes: Using the extracted performance benchmark parameters as topology skeleton nodes, multidimensional degradation correlation data of performance indicators of different functional domains under joint test conditions are extracted in the multi-stress coupling test channel. The multidimensional degradation correlation data is input into a preset Bayesian network topology operator to carry out cross-domain causal inference analysis, and the causal transmission paths of the sudden change in mechanical thermal stress of the pin induced by electrical overload heating and the electromagnetic leakage of the gap induced by mechanical vibration deformation are identified. Asymmetric unidirectional directed edges are connected between each topological skeleton node along the evolution direction of the causal transmission path, and the failure propagation slope of the directed edges is calculated as the coupling failure weight to construct the cross-domain correlation graph of the device's full performance.

[0024] Specifically, the multi-physics coupling superposition includes: Specific nodes marked with physical field coupling failure relationships in the cross-domain correlation graph of the full performance of the equipment are retrieved, and the boundary assessment conditions of the corresponding nodes are extracted through the independent hardware channel of the unified scheduling controller. The stress-sensitive timing sequence of the industrial control computing equipment under test in typical failure modes is extracted, and the loading order of electrical stress, environmental stress and electromagnetic stress is dynamically adjusted and switched according to the stress-sensitive timing sequence. The unified scheduling controller is driven to issue follow-up control commands to multiple groups of external test execution units, and the superposition of overload voltage, temperature alternation and spatial multi-band electromagnetic interference is completed on a unified spatiotemporal axis to form a multi-domain superposition simulation scenario.

[0025] Specifically, the multi-test-unit collaborative execution script includes: An object-oriented modular compilation approach is adopted to encapsulate the loading timing and duration parameters of each test condition in the multi-domain superimposed simulation scenario. Within the compilation architecture, a test triggering module for triggering hardware channels, a timing control module for controlling the slope of multiple stress fields, and an exception rollback module for sudden transient short circuit protection are independently constructed. The compiled control fields are distributed to multiple sets of test execution units arranged in parallel, and each test execution unit is bound to a unified reference clock synchronization mechanism within the local area network to maintain coordinated operation, generating a multi-test unit coordinated execution script containing multi-level overload strategies.

[0026] Specifically, the execution of the preset security policy by the exception rollback module in the multi-test unit collaborative execution script includes: When the multi-test unit detects a transient overcurrent or waveform distortion anomaly in the industrial control computing device under test during the synchronous execution of the script, the anomaly rollback module forcibly suspends the current hardware test channels through the hardware interrupt controller. The pre-stored stepped load reduction control algorithm is retrieved, and the unified clock synchronization mechanism is driven to send a synchronization peak reduction command to each test execution unit. Within a preset microsecond time window, the electrical overload amplitude and temperature alternation slope are proportionally reduced to the safety margin range. Initiate a closed-loop stress retry scheduling mechanism to drive each hardware test channel to restart and execute unfinished test fields within the safety margin range, and write the interrupt protection trace back to the policy log in the local memory.

[0027] Specifically, the time-frequency domain separation includes: The real-time acquired performance response parameters are input into the continuous wavelet transform and short-time Fourier transform modules in the preset computing environment. The continuous wavelet transform module is called to perform multi-scale decomposition on the non-stationary transient response components in the performance response parameters to obtain time-domain nonlinear abrupt change characteristics. Simultaneously, the short-time Fourier transform module is used to perform sliding window spectrum analysis on the performance response parameters to lock the steady-state periodic response components. The energy distribution and variance change rate of the corresponding components are calculated, and the corresponding degradation features are extracted to complete the time-frequency domain separation, providing input for subsequent cross-domain coupled degradation modeling. The non-stationary transient response components include transient current and impact mechanical strain, and the steady-state periodic response components include periodic electromagnetic interference.

[0028] Specifically, the construction process of the cross-domain performance degradation response surface model is as follows: A hierarchical topology architecture comprising a feature fusion layer, an interaction fitting layer, and a boundary convergence layer is constructed. The feature fusion layer imports the extracted degraded feature sequence into a regression matrix constructed by fusing partial least squares regression and gradient boosting decision trees, performing cross-domain recombination to eliminate strong collinearity and output cross-domain multiphysics parameters. The interaction fitting layer performs multivariate nonlinear fitting on the cross-domain multiphysics parameters to solve the response surface equation, calculating the interaction term coefficients characterizing the cross-interference between stresses, and generating a hyperspace response surface. The boundary convergence layer extracts the system's dynamic load noise and injects it as a boundary constraint into the hyperspace response surface, constructing a cross-domain performance degradation response surface model.

[0029] Specifically, the process of constructing the regression matrix includes: The partial least squares regression is used to project the input degenerate feature sequence into the principal component space to remove collinearity interference and output a decoupled feature matrix. The decoupled feature matrix is ​​used as a training sample and input into the gradient boosting decision tree. Multi-level cascaded weak learners are used to fit the residuals of multi-field coupling degenerate along the negative gradient direction. The output weights of the leaf nodes of the gradient boosting decision tree are fused with the mapping vector of the principal component space projection to construct the regression matrix.

[0030] Specifically, the coupling degradation source analysis includes: Using the performance degradation characteristic calculated by the cross-domain performance degradation response surface model as the starting point for tracing, a path backtracking algorithm based on graph search is invoked to retrieve the historical signal disturbance characteristics of each topological node in reverse cross-domain tracing along the direction of the coupling failure edge with the largest topological weight value in the cross-domain correlation graph of the device's full performance. Simultaneously, the local disturbance sensitivity matrix is ​​calculated for the electrical, environmental, and electromagnetic parameters of the industrial control computing device under test. The local sensitivity gradient of each topological node is compared along the backtracking path to lock the topological source node with the highest amplification coefficient and the largest deviation as the root cause node, thus completing the coupling degradation source tracing analysis.

[0031] Specifically, the evaluation process for the full performance level of the equipment includes: The degradation amount of the root cause node and the degradation feature amount of each performance index dimension are obtained, and the coupling sensitivity of the root cause node in the cross-domain correlation graph of the overall performance of the device is calculated. The degradation amount and the coupling sensitivity are correlated and mapped to determine the performance degradation level of the industrial control computing device under test in the corresponding functional domain. The current actual application scenario of the industrial control computing device under test is obtained, and the weight coefficient ratio of the corresponding functional domain is dynamically adjusted according to the actual application scenario. The performance degradation level of each functional domain and the adjusted weight coefficient are weighted and comprehensively calculated to evaluate the overall performance level.

[0032] Specifically, the update of the device full-performance cross-domain correlation graph includes: The evaluation results of the overall performance level assessment of the equipment are converted into the actual degradation amount of each performance index. The actual degradation amount is compared with the current physical field coupling failure weight of each coupling failure relationship in the cross-domain correlation map of the overall performance of the equipment to obtain the mismatch path between the actual degradation amount and the weight. Reverse topological mapping is performed along the direction of the coupling failure relationship corresponding to the mismatch path. The physical field coupling failure weight of each coupling failure relationship is corrected based on the mapping result.

[0033] Example 1: Construction of a Cross-Domain Association Map of Full-Performance Equipment This embodiment uses a certain type of industrial control equipment as a scenario to specifically demonstrate how to extract performance benchmark parameters and boundary assessment conditions from design specification documents, and construct a cross-domain correlation graph of the equipment's full performance: Step 1: Semantic parsing and parameter extraction of design specification documents The design specification document of the industrial control computing device under test is obtained. This document contains detailed design specifications for the device in the electrical functional domain, physical environment tolerance domain, and electromagnetic compatibility domain. A semantic parsing model is used to perform document-level semantic segmentation on the design specification document, automatically dividing it into electrical functional domain paragraphs, physical environment tolerance domain paragraphs, and electromagnetic compatibility domain paragraphs based on chapter titles and keywords. In a specific embodiment of this invention, the industrial control computing device does not specifically refer to a single general-purpose computer, but rather to an ecosystem of computing and display / control peripherals designed for harsh industrial environments and collaboratively executing automated measurement and control scenarios. Specifically, the industrial control computing device comprises two main parts: First, there is the industrial control computer host, which serves as the core of computing and control. Its specific forms include, but are not limited to, original industrial machines, embedded industrial control computers, ruggedized computers, ruggedized servers, ruggedized laptops, open pluggable computers, simulation computer terminals, network security terminals, serial port servers, display computing all-in-one machines, and industrial panel PCs. Secondly, as peripheral display and communication components for information transmission and human-computer interaction, these include, but are not limited to, industrial switches, multi-computer switches, industrial displays, ruggedized display systems, multi-screen splicing terminals, commercial display systems, industrial control consoles, industrial keyboards, and industrial computer mice.

[0034] When the aforementioned host and peripheral components are actually performing automated scenario simulation tests, they are all bound to a unified reference clock synchronization mechanism to maintain coordinated operation and jointly face the cross-domain coupling degradation test of multiple physical fields such as electrical, environmental and electromagnetic stress.

[0035] Extract the performance indicator names and corresponding baseline value ranges from the electrical functional domain section: Rated operating voltage V rated and allowable fluctuation range [V] min V max Rated operating current I rated and overload protection threshold I protect Eye opening threshold E, a signal integrity parameter min Extracted from the physical environment tolerance domain section: operating temperature range [T] min ,T max Storage temperature range [T] st_min ,T st_max Vibration tolerance spectrum [f] vib_min ,f vib_max and the corresponding acceleration amplitude A vib Drop test height H drop Extracted from the electromagnetic compatibility domain section: Radiated emission limit E rad_limit In frequency band [f emc_min ,f emc_max Envelope curve within ], conducted emission limit C cond_limitElectrostatic discharge withstand level V esd .

[0036] Extract the stress types and corresponding stress intensity ranges from the text paragraphs corresponding to each functional domain: The stress types for the electrical functional domain include overload voltage V. over With transient current I trans The stress types tested in the physical environment tolerance domain include the temperature alternation amplitude ΔT and vibration acceleration A. vib The stress types tested in the electromagnetic compatibility domain include radiation field strength E rad With electrostatic discharge voltage V esd The performance index names, benchmark value ranges, stress types, and stress intensity ranges extracted from the same functional domain are structured and stored together to form boundary assessment conditions.

[0037] Step 2: Causal Inference and Graph Construction of Bayesian Network Topological Operators The extracted performance baseline parameters are used as topology skeleton nodes. Multidimensional degradation correlation data of performance indicators in different functional domains under joint test conditions are extracted from the multi-stress coupling test channel among the three sets of nodes in the electrical functional domain, physical environmental tolerance domain, and electromagnetic compatibility domain. The joint test condition refers to simultaneously applying at least two of the following stresses to the equipment: electrical stress, environmental stress, and electromagnetic stress, and recording the degradation of each performance indicator under the coupling condition.

[0038] Multidimensional degradation correlation data are input into a pre-defined Bayesian network topology operator to conduct cross-domain causal inference analysis. The structure of the Bayesian network topology operator is as follows: each performance index node is a random variable, and directed edges between nodes represent causal relationship assumptions, with each directed edge associated with a conditional probability distribution. The causal inference analysis process is as follows: for each pair of cross-domain nodes, the conditional probability P(Di) of the subsequent node degradation given the degradation amount of the preceding node is calculated. j >δ j |D i >δ i ), where D i D represents the degradation amount of the preceding node i. j δ represents the degradation of the subsequent node j. i With δ j These represent the degradation thresholds for both nodes. The higher the conditional probability value, the greater the likelihood of a causal coupling failure relationship between the two nodes.

[0039] Identify the causal transmission path of the sudden change in pin mechanical thermal stress induced by electrical overload heating: when the overload current in the electrical functional domain causes the component temperature rise to exceed the threshold T. thAt this time, mechanical stress concentration occurs at the pin solder joints due to differences in the coefficients of thermal expansion of the materials, leading to degradation of the mechanical reliability of the physical environment tolerance zone. The causal transmission path of electromagnetic leakage in the gap induced by mechanical vibration deformation is identified: when the vibration acceleration in the physical environment tolerance zone exceeds the threshold A... th At that time, the gaps in the equipment casing increased due to deformation, causing the radiated emissions in the electromagnetic compatibility domain to exceed the limit.

[0040] Asymmetric unidirectional directed edges are connected between the nodes of each topological skeleton along the evolution direction of the causal propagation path. Asymmetry means that the direction of the causal relationship is irreversible—electrical degradation can induce mechanical degradation, but mechanical degradation usually does not in turn induce electrical degradation. The failure propagation slope k of each directed edge is calculated. ij As a weight for coupling failure: k ij =ΔD j / ΔD i , Where ΔD i Let ΔD be the degradation increment of the preceding node i. j The response degradation increment of the subsequent node j due to the degradation of the preceding node. Failure propagation slope k. ij The larger the value, the more significant the amplification effect of the degradation of the preceding node on the degradation of the following node. The coupling failure weight is written into the edge attribute of the corresponding edge to construct a cross-domain correlation graph of the device's overall performance.

[0041] Example 2: Stress-Sensitive Timing Extraction and Multi-Stress Loading Mechanism This embodiment follows Embodiment 1, using the same model of industrial control equipment as a scenario, and specifically describes the entire process of extracting coupled failure nodes from the cross-domain correlation graph of the equipment's full performance and generating a multi-domain superimposed simulation scenario: Step 1: Extraction of stress-sensitive time series Specific node pairs marked with physical field coupling failure relationships are extracted from the cross-domain correlation map of the overall performance of the equipment. Taking the overload current node in the electrical functional domain and the pin thermal stress node in the physical environment tolerance domain as an example, there is a coupling failure relationship between the two nodes.

[0042] Stress-sensitive time series data of the industrial control computing equipment under test under typical failure modes were extracted. Typical failure modes were obtained from a historical fault database, which recorded stress change data at various time points during the overload-thermal stress coupled failure process. The stress-sensitive time series is a multi-dimensional time series. S(t) = {V} over (t),T junction (t),ε stress (t)}, where V over (t) is the curve of overload voltage versus time, T junction (t) is the curve showing the change of the junction temperature of the component with time, εstress (t) represents the curve of the pin's mechanical strain over time. The stress-sensitive timing reflects the time sequence in a typical failure process: electrical stress rises first, followed by thermal stress, and then mechanical stress manifests last.

[0043] The loading order of electrical stress, environmental stress, and electromagnetic stress is dynamically adjusted and switched based on the stress-sensitive timing sequence. The adjustment logic is as follows: the onset time of each stress component in the stress-sensitive timing sequence is sorted in order, and the stress type with the earliest onset is loaded first. In the overload-thermal stress coupling scenario, the onset time of electrical overload stress is the earliest, so it is loaded first; the onset time of temperature alternating stress is the second earliest, and loading begins after the electrical stress is loaded to the preset amplitude; vibration stress begins to be loaded after the temperature reaches the preset threshold.

[0044] Step 2: Execution of Multi-Stress Loading and Compilation of Co-scripts The unified scheduling controller sends follow-up control commands to multiple external test execution units. The unified scheduling controller is connected to the electrical stress loading unit, environmental stress loading unit, and electromagnetic stress loading unit through independent hardware channels.

[0045] Overload voltage, temperature alternation, and spatial multi-band electromagnetic interference are superimposed on a unified spatiotemporal axis. The superposition method is as follows: the amplitudes of each stress component are aligned along the time axis, and stress of corresponding amplitudes is applied synchronously to each loading unit at the same time. Synchronization and alignment are achieved through a unified reference clock. The clock signal of each loading unit is obtained by frequency division from the same clock source, and the clock deviation is controlled within a preset tolerance range.

[0046] The loading timing and duration of each test condition in a multi-domain overlay simulation scenario are compiled into a multi-test-unit collaborative execution script. The compilation method employs object-oriented modular compilation: encapsulating the loading timing and duration parameters of each test condition, constructing a test trigger module for triggering hardware channels, a timing control module for controlling the slope of multiple stress fields, and an exception fallback module for sudden transient short-circuit protection. The test trigger module stores the loading start and end times for each stress type, the timing control module stores the loading slope for each stress type, and the exception fallback module stores preset safety policies. The compiled control fields are distributed to multiple sets of test execution units arranged in parallel, and each test execution unit maintains collaborative operation under a unified reference clock synchronization mechanism.

[0047] Example 3: Construction of a Cross-Domain Performance Degradation Response Surface Model This embodiment follows from Embodiment 2, and the specific process of performing time-frequency domain separation on the collected performance response parameters, extracting degradation feature quantities, and constructing a cross-domain performance degradation response surface model is as follows: Step 1: Time-frequency domain separation of performance response parameters and extraction of degradation features During the execution of a multi-domain overlay simulation scenario by the industrial control computing device under test, operational status parameters and performance response parameters are collected in real time. Operational status parameters include the amplitude data of each stress type currently applied, and performance response parameters include the response signals of the device under various stresses.

[0048] The performance response parameters are separated into time and frequency domains. The real-time acquired performance response parameters are input into the continuous wavelet transform and short-time Fourier transform modules. The continuous wavelet transform module is used to perform multi-scale decomposition of the non-stationary transient response components in the performance response parameters: Morlet wavelet is selected as the mother wavelet function, and wavelet transforms are performed on the transient current signal and the impact mechanical strain signal at different scales to generate wavelet coefficient matrices. The energy distribution at each scale is extracted from the wavelet coefficient matrix, and the scales with concentrated energy are identified as the characteristic time-frequency regions of transient degradation.

[0049] Simultaneously utilize the short-time Fourier transform function module to perform sliding window spectrum analysis on the performance response parameters: set the window width W. stft With sliding step size S stft Fourier transforms are performed on the signal segments within each time window to obtain the relationship between the spectrum and time. Frequency components with stable amplitudes are extracted from the spectrum, and the frequency position and amplitude of these components are used as characteristics of the steady-state periodic response components.

[0050] Calculate the energy distribution and variance rate of change of the corresponding components. The energy distribution is the sum of the squares of the signal amplitudes within each frequency band, and the variance rate of change is the change in the standard deviation of the energy distribution between two consecutive frames. Components whose energy distribution and variance rate of change exceed a preset threshold are marked as degradation-sensitive components, and their corresponding degradation features are extracted.

[0051] Step 2: Modeling the fusion of partial least squares regression and gradient boosting decision trees A hierarchical topology architecture is constructed, comprising a feature fusion layer, an interactive fitting layer, and a boundary convergence layer. In this architecture, the layers are connected sequentially, with the output of the feature fusion layer serving as the input to the interactive fitting layer, and the output of the interactive fitting layer serving as the input to the boundary convergence layer.

[0052] The degraded features are imported into the regression matrix constructed by fusing partial least squares regression (PLL) and gradient boosting decision trees through a feature fusion layer. PLL projects the input degraded features into the principal component space: Let the degraded feature matrix be X, and the performance degradation vector be y. PLL searches for a set of orthogonal principal components T = {t1, t2, ..., t...} in the column space of X. h}, where h is the number of principal components. Each principal component t k It is a linear combination of the original features, and its covariance with y is maximized: t k =Xwk , where w k The weight vector of the k-th principal component is obtained by maximizing the covariance Cov(t). k The principal component space projection (PSP) is obtained by solving for y. The PSP removes the collinearity interference from the original degenerate features, outputting the decoupled feature matrix T.

[0053] The decoupled feature matrix T is used as training samples and input into the gradient boosting decision tree. The gradient boosting decision tree consists of M cascaded weak learners, each of which is a depth-constrained regression decision tree. The training process involves layer-by-layer residual fitting along the negative gradient direction to the multi-field coupling degradation residuals: the m-th weak learner f... m (T) The cumulative prediction residuals r of the m-1 weak learners before fitting m-1 The residual is the difference between the actual degradation and the current cumulative prediction. After M iterations, the cumulative prediction model is: , where ν is the learning rate, which controls the contribution of each weak learner to the final model.

[0054] The nonlinear fitting residuals of the gradient boosting decision tree for each degenerate feature are element-wise weighted and fused with the principal component score vectors of partial least squares regression to construct a regression matrix. The fusion weights are determined based on the signal-to-noise ratio (SNR) of each feature—features with higher SNRs are assigned higher weights in the fusion. The constructed regression matrix possesses both the collinearity elimination capability of partial least squares regression and the nonlinear fitting capability of gradient boosting decision trees. After cross-domain recombination, cross-domain degenerate correlation features are output.

[0055] Step 3: Response Surface Generation and Boundary Convergence The response surface equation is solved by performing multivariate nonlinear fitting on the cross-domain degradation correlation features through an interactive fitting layer. Let the multi-domain stress input variable be x=(x elec ,x env ,x emc ), corresponding to electrical stress intensity, environmental stress intensity, and electromagnetic stress intensity, respectively, with the degradation output variable being y. The response surface equation adopts a quadratic polynomial form: Where β0 is a constant term, β i β is the coefficient of the linear term. ii β is the coefficient of the quadratic term. ij This represents the interaction term coefficient. The interaction term coefficient β ij The cross-interference effect between different stresses was characterized—for example, β. elec,env This represents the additional contribution to degradation when electrical stress and environmental stress act simultaneously; this contribution is not simply the sum of the degradation amounts when the two stresses act alone. The coefficients are obtained by fitting the cross-domain degradation correlation characteristics using the least squares method.

[0056] The obtained response surface equation is used to generate a hyperspace response surface. The hyperspace response surface is represented as a surface in the three-dimensional input space, and the height of the surface represents the degradation prediction under various stress combinations.

[0057] Dynamic load noise of the system is extracted through a boundary convergence layer. This dynamic load noise originates from uncontrollable random disturbances in the test environment, such as power fluctuations, minor temperature drifts, and electromagnetic background noise. The load noise is then injected as a boundary constraint into the hyperspace response surface: let the standard deviation of the load noise be σ. noise The hyperspace response surface is shifted upwards by +kσ along the degradation direction. noise and downward offset -kσ noise , where k is the confidence coefficient, generates the confidence boundary of the response surface. The hyperspace response surface after injecting boundary constraints is used as the cross-domain performance degradation response surface model.

[0058] Example 4: Coupled Degradation Source Analysis and Full Performance Level Assessment This embodiment follows from Embodiment 3. The specific process of performing coupled degradation source analysis, evaluating the overall performance level, and updating the graph weights based on the cross-domain performance degradation response surface model is as follows: Step 1: Path backtracking and root cause localization based on graph search The performance degradation features of each performance index dimension are extracted from the cross-domain performance degradation response surface model. The performance degradation feature is the extent to which the degradation exceeds a preset benchmark value, and is the difference between the degradation of each performance index and the corresponding benchmark threshold.

[0059] Using performance degradation features as the starting point for tracing, a graph search-based path backtracking algorithm is invoked. The path backtracking algorithm employs an improved Dijkstra's algorithm: using the failure propagation slope k of each coupled failure edge in the device's full-performance cross-domain correlation graph. ij The reciprocal of the value is used as the edge weight w. ij =1 / k ij The smaller the edge weight, the more likely the path is to be the main channel for degradation propagation. Taking the performance index node with the largest current degradation as the source node, the shortest path from the source node to each upstream node is calculated by reverse retrieval along the direction of the coupling failure edge. The reverse retrieval process is to trace back along the direction of the physical field coupling failure relationship—from the affected downstream node to its upstream node, rather than in the forward propagation direction.

[0060] Simultaneously, the local disturbance sensitivity matrix is ​​calculated for the electrical, environmental, and electromagnetic parameters of the industrial control computing equipment under test. Local disturbance sensitivity matrix S local element s pq Defined as the degenerate response of node q when a unit perturbation is applied to node p: s pq =ΔD q / ΔD p , Where ΔD p Let ΔD be the small perturbation degradation of node p. q Let be the change in response degradation of node q. The sensitivity matrix is ​​calculated using the finite difference method: small perturbations are applied to each node sequentially, and the changes in degradation of other nodes are recorded. The sensitivity matrix is ​​then constructed column by column.

[0061] Compare the local sensitivity gradients of each node along the backtracking path. (Sensitivity gradient) Let be the magnitude of the row vector corresponding to node p in the sensitivity matrix, reflecting the overall influence of that node in the coupled network. The node with the largest sensitivity gradient and the largest deviation in degradation is searched along the backtracking path and identified as the root cause node. Degradation deviation is the multiple of the difference between the node's current degradation and its historical mean relative to the standard deviation. The root cause node is the earliest source node to degrade in the coupling failure chain; its degradation propagates downstream along the coupling path, causing other nodes to degrade.

[0062] Step 2: Coupling Sensitivity Calculation and Full Performance Level Assessment Obtain the degradation amount of the root cause node and the degradation feature quantity of each performance index dimension. Calculate the coupling sensitivity C of the root cause node in the cross-domain correlation graph of the device's overall performance. root : , Where out(root) represents the set of all outgoing edges pointing to the subsequent nodes of the root node, k root,j Let C be the failure propagation slope from the root cause node to the subsequent node j. Coupling sensitivity C root This reflects the impact of root cause node degradation on overall performance—the more outgoing edges and the greater the failure propagation slope, the higher the sensitivity.

[0063] A correlation mapping is performed between the degradation level of the root cause node and its coupling sensitivity to determine the performance degradation level of the industrial control computing device under test in the corresponding functional domain. The correlation mapping is performed using predefined multi-level thresholds: the product of degradation level and coupling sensitivity is used as a comprehensive degradation index, which is then matched against preset multi-level threshold ranges. A comprehensive degradation index below the first threshold indicates a slight degradation level, between the first and second thresholds indicates a moderate degradation level, between the second and third thresholds indicates a severe degradation level, and above the third threshold indicates a failure level.

[0064] The system obtains the current actual application scenario of the industrial control computing device under test and dynamically adjusts the weight coefficient ratio of the corresponding functional domains according to the actual application scenario. Application scenarios include industrial control, communication base stations, and automotive electronics, etc. Different scenarios have different rigid requirements for electrical safety and signal interference resistance. The weight coefficients are obtained by querying the preset scenario-weight mapping table.

[0065] The performance degradation level of each functional domain is weighted and calculated together with the adjusted weighting coefficients to obtain the overall performance level. The overall performance level is a quantitative evaluation result of the overall performance of the machine, used to determine whether the equipment meets the factory quality requirements.

[0066] Step 3: Graph Weight Update The evaluation results of the full performance level assessment are converted into the actual degradation amount of each performance index. The actual degradation amount is the actual measured degradation value of each performance index during the assessment process. The actual degradation amount is compared with the current physical field coupling failure weight of each coupling failure relationship in the cross-domain correlation map of the equipment's full performance. The comparison method is to calculate the ratio between the actual degradation amounts. , and the failure propagation slope k currently stored in the graph ij Compare the two and calculate the deviation rate between them.

[0067] Coupled failure relationships with deviation rates exceeding a preset deviation threshold are marked as mismatched paths. A reverse topology mapping is performed along the direction of the coupled failure relationship corresponding to the mismatched path: starting from the subsequent node of the mismatched path, the actual degradation ratio is mapped and updated along the graph topology along the direction of the coupled failure edge. The physical field coupled failure weights of each coupled failure relationship are corrected based on the mapping results. α is the smoothing coefficient, with a preset value between [0,1]. The smoothing coefficient controls the update speed of the old and new weights—the closer α is to 1, the more conservative the weight update, maintaining the stability of historical data; the closer α is to 0, the more aggressive the weight update, quickly responding to changes in new data. This completes the closed-loop update of the cross-domain correlation map of the entire device performance.

[0068] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for full performance evaluation of industrial control computing equipment for automated scenario simulation, characterized in that, include: S1: Analyze the design specification document of the industrial control computing device under test, extract the performance benchmark parameters and boundary assessment conditions of the industrial control device under test in different functional domains, and construct a cross-domain correlation graph of the device's full performance with the performance benchmark parameters as nodes and the coupling failure relationship between different functional domains as edges; S2: Extract nodes with physical field coupling failure relationships from the cross-domain correlation graph of the full performance of the equipment, perform multi-stress superposition on the corresponding boundary assessment conditions, generate a multi-domain superposition simulation scenario that includes synchronous loading of electrical, environmental and electromagnetic stresses, and compile the loading sequence and duration of the test conditions in the multi-domain superposition simulation scenario into a multi-test unit collaborative execution script. S3: Real-time acquisition of the operating status parameters and performance response parameters of the industrial control computing device under test in a multi-domain superimposed simulation scenario; time-frequency domain separation of the performance response parameters; extraction of degradation feature quantities of each performance index dimension from the separated performance response parameters; cross-domain coupled degradation modeling; construction of a cross-domain performance degradation response surface model. S4: Based on the cross-domain performance degradation response surface model, perform coupling degradation source analysis along the direction of each physical field coupling failure relationship in the cross-domain correlation graph of the equipment's full performance to obtain the root cause node of performance degradation. Based on the degradation amount of the root cause node, evaluate the overall performance level of the equipment and feed the evaluation results back to the cross-domain correlation graph of the equipment's full performance to update the physical field coupling failure weights between each index.

2. The method according to claim 1, characterized in that, The extraction process of the performance benchmark parameters and boundary assessment conditions specifically includes: The design specification document is semantically segmented at the chapter level using a semantic parsing model. The segmented text paragraphs are categorized into electrical functional domain, physical environment tolerance domain, and electromagnetic compatibility domain. Performance index names and corresponding benchmark value ranges are extracted from the text paragraphs corresponding to each functional domain. The test stress types and corresponding stress intensity ranges are also extracted from the text paragraphs corresponding to each functional domain. The performance index names, benchmark value ranges, test stress types, and stress intensity ranges extracted within the same functional domain are then structurally associated and stored to form boundary test conditions.

3. The method according to claim 1, characterized in that, The process of constructing the cross-domain correlation map of the device's full performance specifically includes: Using the extracted performance benchmark parameters as topology skeleton nodes, multidimensional degradation correlation data of performance indicators of different functional domains under joint test conditions are extracted in the multi-stress coupling test channel. The multidimensional degradation correlation data is input into a preset Bayesian network topology operator to carry out cross-domain causal inference analysis, and the causal transmission paths of the sudden change in mechanical thermal stress of the pin induced by electrical overload heating and the electromagnetic leakage of the gap induced by mechanical vibration deformation are identified. Asymmetric unidirectional directed edges are connected between each topological skeleton node along the evolution direction of the causal transmission path, and the failure propagation slope of the directed edges is calculated as the coupling failure weight to construct the cross-domain correlation graph of the device's full performance.

4. The method according to claim 1, characterized in that, The multiphysics field coupling superposition specifically includes: Specific nodes marked with physical field coupling failure relationships in the cross-domain correlation graph of the full performance of the equipment are retrieved, and the boundary assessment conditions of the corresponding nodes are extracted through the independent hardware channel of the unified scheduling controller. The stress-sensitive timing sequence of the industrial control computing equipment under test in typical failure modes is extracted, and the loading order of electrical stress, environmental stress and electromagnetic stress is dynamically adjusted and switched according to the stress-sensitive timing sequence. The unified scheduling controller is driven to issue follow-up control commands to multiple groups of external test execution units, and the superposition of overload voltage, temperature alternation and spatial multi-band electromagnetic interference is completed on a unified spatiotemporal axis to form a multi-domain superposition simulation scenario.

5. The method according to claim 1, characterized in that, The multi-test-unit collaborative execution script specifically includes: An object-oriented modular compilation approach is adopted to encapsulate the loading timing and duration parameters of each test condition in the multi-domain superimposed simulation scenario. Within the compilation architecture, a test triggering module for triggering hardware channels, a timing control module for controlling the slope of multiple stress fields, and an exception rollback module for sudden transient short circuit protection are built. The compiled control fields are distributed to multiple sets of test execution units arranged in parallel, and each test execution unit is bound to a unified reference clock synchronization mechanism within the local area network to maintain coordinated operation, generating a multi-test unit coordinated execution script containing multi-level overload strategies.

6. The method according to claim 1, characterized in that, The execution of the preset security policy by the exception rollback module in the multi-test unit collaborative execution script specifically includes: When the multi-test unit detects a transient overcurrent or waveform distortion anomaly in the industrial control computing device under test during the synchronous execution of the script, the anomaly rollback module forcibly suspends the current hardware test channels through the hardware interrupt controller. The pre-stored stepped load reduction control algorithm is retrieved, and the unified clock synchronization mechanism is driven to send a synchronization peak reduction command to each test execution unit. Within a preset microsecond time window, the electrical overload amplitude and temperature alternation slope are proportionally reduced to the safety margin range. Initiate a closed-loop stress retry scheduling mechanism to drive each hardware test channel to restart and execute unfinished test fields within the safety margin range, and write the interrupt protection trace back to the policy log in the local memory.

7. The method according to claim 1, characterized in that, The time-frequency domain separation specifically includes: The real-time acquired performance response parameters are input into the continuous wavelet transform and short-time Fourier transform modules in the preset computing environment. The continuous wavelet transform module is called to perform multi-scale decomposition on the non-stationary transient response components in the performance response parameters to obtain time-domain nonlinear abrupt change characteristics. Simultaneously, the short-time Fourier transform module is used to perform sliding window spectrum analysis on the performance response parameters to lock the steady-state periodic response components. The energy distribution and variance change rate of the corresponding components are calculated, and the corresponding degradation features are extracted to complete the time-frequency domain separation.

8. The method according to claim 1, characterized in that, The construction process of the cross-domain performance degradation response surface model is as follows: A hierarchical topology architecture comprising a feature fusion layer, an interaction fitting layer, and a boundary convergence layer is constructed. The feature fusion layer imports the extracted degenerate feature sequence into a regression matrix constructed by fusing partial least squares regression and gradient boosting decision trees, performing cross-domain recombination to eliminate strong collinearity and output cross-domain multiphysics parameters. The interaction fitting layer performs multivariate nonlinear fitting on the cross-domain multiphysics parameters, calculating the interaction coefficients characterizing the cross-interference between stresses and generating a hyperspace response surface. The boundary convergence layer extracts the system's dynamic load noise and injects it as a boundary constraint into the hyperspace response surface, constructing a cross-domain performance degradation response surface model.

9. The method according to claim 8, characterized in that, The process of constructing the regression matrix specifically includes: The partial least squares regression is used to project the input degenerate feature sequence into the principal component space to remove collinearity interference and output a decoupled feature matrix. The decoupled feature matrix is ​​used as a training sample and input into the gradient boosting decision tree. Multi-level cascaded weak learners are used to fit the residuals of multi-field coupling degenerate along the negative gradient direction. The output weights of the leaf nodes of the gradient boosting decision tree are fused with the mapping vector of the principal component space projection to construct the regression matrix.

10. The method according to claim 1, characterized in that, The coupling degradation source analysis specifically includes: Using the performance degradation feature calculated by the cross-domain performance degradation response surface model as the starting point for tracing, a path backtracking algorithm based on graph search is invoked to retrieve the historical signal disturbance features of each topological node in reverse cross-domain tracing along the direction of the coupling failure edge with the largest topological weight value in the cross-domain correlation graph of the device's full performance. Simultaneously, the local disturbance sensitivity matrix is ​​calculated for the electrical, environmental, and electromagnetic parameters of the industrial control computing device under test. The local sensitivity gradient of each topological node is compared along the backtracking path to lock the topological source node as the root cause node, thus completing the coupling degradation source tracing analysis.

11. The method according to claim 1, characterized in that, The evaluation process for the full performance level of the equipment specifically includes: The degradation amount of the root cause node and the degradation feature amount of each performance index dimension are obtained, and the coupling sensitivity of the root cause node in the cross-domain correlation graph of the overall performance of the device is calculated. The degradation amount and the coupling sensitivity are correlated and mapped to determine the performance degradation level of the industrial control computing device under test in the corresponding functional domain. The current actual application scenario of the industrial control computing device under test is obtained, and the weight coefficient ratio of the corresponding functional domain is dynamically adjusted according to the actual application scenario. The performance degradation level of each functional domain and the adjusted weight coefficient are weighted and comprehensively calculated to evaluate the overall performance level.

12. The method according to claim 1, characterized in that, The update of the device full-performance cross-domain correlation graph specifically includes: The evaluation results of the overall performance level assessment of the equipment are converted into the actual degradation amount of each performance index. The actual degradation amount is compared with the current physical field coupling failure weight of each coupling failure relationship in the cross-domain correlation map of the overall performance of the equipment to obtain the mismatch path between the actual degradation amount and the weight. Reverse topological mapping is performed along the direction of the coupling failure relationship corresponding to the mismatch path. The physical field coupling failure weight of each coupling failure relationship is corrected based on the mapping result.

Images