Discrete energy-saving test automation method, device and equipment and storage medium

Abstract

The invention relates to the technical field of automation, and discloses a discrete energy-saving test automation method and device, equipment and a storage medium. The method comprises the following steps: performing multi-source heterogeneous signal acquisition and preprocessing on a power supply aging test process of discrete energy-saving power supply aging test equipment to obtain a multi-dimensional time sequence data matrix; decomposing the multi-dimensional time sequence data matrix into a temperature subspace, a vibration subspace and an electrical parameter subspace, and performing data feature fusion to obtain a power supply aging state fusion feature vector; performing power supply health assessment based on the power supply aging state fusion feature vector to obtain a global health index and a state evolution rate index; and performing adaptive control parameter analysis on the temperature control loop and the vibration control loop under dual event triggering conditions to obtain a target control scheme. According to the invention, the characterization capability of the aging state of the power supply is enhanced, more comprehensive and accurate global health assessment is formed, the control execution frequency is reduced, and the system energy consumption is reduced.

Description

Technical Field

[0001] The present invention relates to the field of automation technology, and in particular to a discrete energy-saving test automation method, device, equipment and storage medium. Background Art

[0002] The requirements for aging test accuracy and efficiency of power supply as a core component are increasing day by day. However, the test methods in existing technologies often rely on manual experience and judgment, lacking comprehensive analysis and intelligent control of multi-source parameters. As a result, the temperature and vibration parameters during the test process cannot accurately match the aging characteristics of the power supply, which not only increases the test cost but also prolongs the product launch cycle.

[0003] Discrete energy-saving power supply burn-in testing, as a new testing method, achieves continuity and independence through independent tray support and a chain drive system. However, the unknown measurement sensitivity and non-differentiability, coupled with the inability to directly measure certain power supply parameters, pose significant challenges to the overall stability control of the test system. Existing technologies are particularly incapable of effectively assessing the probability of first failure when faced with the non-stationary fluctuations of the power supply burn-in process. This results in a lack of scientific basis for test parameter selection, which not only fails to ensure test adequacy but also leads to excessive consumption of energy resources. Summary of the Invention

[0004] The main purpose of the present invention is to provide a discrete energy-saving test automation method, device, equipment and storage medium. The present invention enhances the ability to characterize the aging status of the power supply, forms a more comprehensive and accurate global health assessment, reduces the control execution frequency, and reduces system energy consumption.

[0005] To achieve the above object, the present invention provides a discrete energy-saving test automation method, comprising the following steps: Perform multi-source heterogeneous signal acquisition and preprocessing on the power aging test process of discrete energy-saving power aging test equipment to obtain a multi-dimensional time series data matrix; Decomposing the multidimensional time series data matrix into a temperature subspace, a vibration subspace, and an electrical parameter subspace, and performing data feature fusion on the temperature subspace, the vibration subspace, and the electrical parameter subspace to obtain a power supply aging state fusion feature vector; Performing power supply health assessment based on the power supply aging state fusion feature vector to obtain a global health index and a state evolution rate index; According to the global health index and the state evolution rate index, adaptive control parameter analysis of dual event triggering conditions is performed on the temperature control loop and the vibration control loop of the discrete energy-saving power supply aging test equipment to obtain a target control solution.

[0006] The present invention also provides a discrete energy-saving test automation device, comprising: The acquisition module is used to collect and preprocess multi-source heterogeneous signals during the power aging test process of the discrete energy-saving power aging test equipment to obtain a multi-dimensional time series data matrix; a feature fusion module, configured to decompose the multidimensional time series data matrix into a temperature subspace, a vibration subspace, and an electrical parameter subspace, and perform data feature fusion on the temperature subspace, the vibration subspace, and the electrical parameter subspace to obtain a fused feature vector of the power supply aging state; An evaluation module, configured to perform a power supply health evaluation based on the power supply aging state fusion feature vector to obtain a global health index and a state evolution rate index; The analysis module is used to perform adaptive control parameter analysis of dual event triggering conditions on the temperature control loop and vibration control loop of the discrete energy-saving power supply aging test equipment according to the global health index and the state evolution rate index to obtain a target control solution.

[0007] The present invention also provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of any one of the above methods when executing the computer program.

[0008] The present invention also provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of any of the above methods are implemented.

[0009] In summary, the technical solution provided by the present invention achieves synchronous acquisition of multi-source heterogeneous signals of temperature, vibration, and electrical parameters by deploying a sensor network at key locations in a discrete energy-saving power supply aging test device, establishing temporal correlations between parameters and overcoming the incomplete testing problem caused by traditional single parameter acquisition. The multidimensional time series data matrix is ​​decomposed into three subspaces: temperature, vibration, and electrical parameters, ensuring the orthogonality and weight balance of features within the subspaces, effectively preventing certain parameters with large fluctuations from dominating the test results in the analysis. By fusing the extracted features, the interpretability of empirical features and the complex mapping capability of deep features are retained, enhancing the ability to characterize the power supply aging status. A multidimensional health index is constructed based on the subspace whitening support vector data description algorithm, with health indicators constructed for temperature, vibration, and electrical parameters respectively. The weight coefficients are determined through sensitivity analysis to form a more comprehensive and accurate global health assessment. A dual event triggering mechanism is used for adaptive parameter control, updating and executing control instructions only when both the state triggering and control triggering conditions are met, significantly reducing the control execution frequency and lowering system energy consumption. A relative sensitivity error compensation mechanism is introduced. When changes in temperature or vibration parameter measurement sensitivity are detected, the control gain coefficient is dynamically adjusted through a learning algorithm, effectively compensating for the impact of execution errors. A cubic polynomial model of a Gaussian process is used to characterize the nonlinear characteristics of the power supply response and its rate of change. A joint probability density function is constructed to estimate the probability of first failure. Multi-objective optimization is performed based on test efficiency metrics, comprehensively considering the probability of first failure, energy consumption, and test time to generate an optimal power supply test parameter combination that balances reliability assessment, energy consumption, and test time. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Figure 1 This is a schematic diagram of the steps of a discrete energy-saving test automation method according to an embodiment of the present invention; Figure 2 This is a structural block diagram of a discrete energy-saving test automation device in one embodiment of the present invention; Figure 3 It is a schematic block diagram of the structure of a computer device according to an embodiment of the present invention.

[0011] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings and in conjunction with the embodiments. DETAILED DESCRIPTION

[0012] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0013] Reference Figure 1 , this embodiment provides a discrete energy-saving test automation method, including the following steps: S1, collects and preprocesses multi-source heterogeneous signals during the power aging test process of the discrete energy-saving power aging test equipment to obtain a multi-dimensional time series data matrix; Among them, multiple temperature measurement points are arranged in the test environment of the discrete energy-saving power supply aging test equipment. The selection of points covers the key heat source areas and environmental boundary positions of the equipment operation to capture the integrity and dynamic change characteristics of the temperature distribution, and a high-precision temperature sensor is installed at each point. It is preferred to use a PT100 platinum resistance sensor, which has good linear output and high thermal response speed. In actual application, the sampling frequency is set to 1Hz to meet the collection requirements of slowly changing thermal signals during the aging process. The original temperature monitoring data is transmitted to the main processing unit in real time through the front-end acquisition module. At the same time, three-axis MEMS acceleration sensors are installed at the four corner points of the tray support frame to monitor the mechanical vibration information generated by the equipment during operation. These sensors have the advantages of small size, high sensitivity and wide response frequency. The sampling frequency is set to 200Hz to capture the vibration fluctuation characteristics from low frequency to high frequency. The original vibration monitoring data generated can reflect the impact of vibration interference on output stability during the aging process of the power supply on the back end; for the electrical signal acquisition part, high-frequency electrical raw data is obtained from the input and output ends of the power supply under test, covering core operating parameters such as voltage, current and power factor, and is collected at a frequency of 10kHz through a precision sampling card to ensure the fine reflection capability of electrical data in the case of short-term mutations. The raw data from temperature monitoring, vibration monitoring, and electrical parameter monitoring are filtered. Based on the frequency characteristics of each signal, appropriate Butterworth low-pass filters are configured to remove high-frequency noise while retaining the main effective components. A median filter is used to eliminate occasional outliers to improve data stability and anti-interference capabilities. Time-based synchronization is implemented for the three target monitoring data types—temperature, vibration, and electrical—to unify their sampling timestamps. A resampling mechanism is used to convert data at different frequencies into time series with equal step sizes. All synchronized target data are integrated into a unified multidimensional time series data matrix, where each column represents a parameter channel and each row represents a time sampling point.

[0014] S2, decompose the multidimensional time series data matrix into temperature subspace, vibration subspace and electrical parameter subspace, and perform data feature fusion on the temperature subspace, vibration subspace and electrical parameter subspace to obtain the fused feature vector of the power supply aging state; Specifically, the multidimensional time series data matrix is ​​structurally split according to parameter attributes and divided into temperature subspace, vibration subspace and electrical parameter subspace, so that each type of physical parameter can be centrally expressed and processed in an independent subspace. For the temperature subspace, based on its slow-varying response characteristics, six statistical features, including mean, maximum, minimum, standard deviation, rate of rise and rate of fall, are calculated through a sliding time window to form a temperature feature vector, which reflects the temperature stability and fluctuation trend during the aging process. For the vibration subspace, time domain statistical extraction and frequency domain energy analysis are performed on the three-axis acceleration signal to obtain its classic time domain indicators such as root mean square, peak value, peak-to-peak value, form factor, and pulse factor. The frequency domain energy characteristics of each frequency band from 0 to 50 Hz, 50 to 100 Hz, and 100 to 200 Hz are calculated through fast Fourier transform to characterize the performance of the vibration state in both time and frequency dimensions and generate a vibration feature vector. For the electrical parameter subspace, five characterization indicators, including input voltage stability, output voltage deviation rate, output current ripple factor, power conversion efficiency, and temperature-efficiency sensitivity coefficient, are extracted. These indicators together constitute the electrical parameter feature vector, covering aspects such as power supply operation efficiency, output fluctuation, and thermal coupling effects. After obtaining the initial eigenvectors for each of the three subspaces, feature whitening and orthogonality transformations are performed to address issues such as inconsistent scales, large amplitude differences, and strong correlations among features across different dimensions. Each eigenvector is then Z-score normalized and its covariance matrix is ​​calculated. A whitening transformation matrix is ​​constructed through eigenvalue decomposition, yielding temperature, vibration, and electrical parameter decomposition features with zero mean, unit variance, and approximately orthogonality. Based on these decomposition features, cross-feature analysis is performed to model the coupling relationships between the different subspaces and convert them into a set of composite indices. These include the output voltage stability index (measured by the ratio of voltage standard deviation to mean), the temperature-efficiency response index (reflecting the response ratio of efficiency changes to temperature changes), the vibration immunity index (reflecting the ability to suppress the effects of vibration changes on output voltage), the power conversion consistency index (characterizing the level of average power fluctuations), and the startup response time index (characterizing the response speed from power-on to stable output). These cross-feature sets, through well-defined and physically meaningful indices, model the interactions between multidimensional features. The temperature decomposition features, vibration decomposition features, electrical parameter decomposition features and cross feature sets are subjected to attention weighted fusion. According to the contribution of the features to the accuracy of aging status prediction during training, the weight distribution of empirical features, statistical features and composite features is dynamically adjusted to construct a fusion weight matrix. The temperature decomposition features, vibration decomposition features, electrical parameter decomposition features and cross feature sets are uniformly mapped to a feature space with comprehensive characterization capabilities through weighted superposition to obtain the fused feature vector of power supply aging status.

[0015] S3, performs power health assessment based on the fusion feature vector of the power aging state to obtain the global health index and state evolution rate index; It should be noted that for each power supply under test, N sample points in normal operation are selected from the initial stage of the aging test. The corresponding fused feature vectors are input into the system as training samples. These fused feature samples are used to construct a high-dimensional normal operation feature space for the power supply. This space characterizes the combined distribution of all key features of the power supply in its non-aging state. To effectively model this normal state feature space, a subspace whitening support vector data description algorithm is introduced. Based on the known normal state samples, a minimum closed hypersphere containing all training samples is constructed in the high-dimensional feature space. To enhance the modeling capability of nonlinear distribution features, the algorithm introduces a radial basis kernel function to nonlinearly map the fused feature vectors into a high-dimensional space. This linearly represents the nonlinear relationships between samples, facilitating the standard quadratic programming solution for the optimization problem of the minimum hypersphere. This process outputs a set of support vectors and their corresponding Lagrange multipliers, which are used to calculate the discriminant distance of the samples relative to the center point in the dual problem. Based on support vectors and Lagrange multipliers, the relative distance between the fused eigenvector of the power supply aging state and the center of the normal region is calculated to obtain a health index value. This value is considered a quantitative expression of whether the sample deviates from the normal state; a larger distance indicates a higher degree of aging. This health index value is further mapped to various physical subspaces. Subspace-related health indices are constructed for the temperature subspace, vibration subspace, and electrical parameter subspace, resulting in a multidimensional health index vector consisting of the temperature-related health index, the vibration-related health index, and the electrical parameter-related health index. A sensitivity analysis of the multidimensional health index is performed to analyze the impact of each subspace on the overall aging trend and determine its weighting coefficient. Under the condition that the weight sums to 1, a unified global health index with comprehensive discriminative capabilities is obtained by summing the products of each dimensional health index and the corresponding weight. The global health index is continuously recorded at multiple moments in a time window of length τ, and its rate of change over time is calculated to obtain a state evolution rate index.

[0016] S4, based on the global health index and state evolution rate index, the temperature control loop and vibration control loop of the discrete energy-saving power supply aging test equipment are adaptively analyzed under dual event triggering conditions to obtain the target control scheme.

[0017] Specifically, dynamic response models for temperature and vibration are constructed for a discrete energy-saving power supply aging test device. For temperature control, pulse width modulation (PWM) is used to adjust the power output of the heating wire. Temperature control is achieved by adjusting the duty cycle of the PWM signal, establishing a control model that shows how temperature responds to PWM changes. For vibration control, a frequency converter (VFD) adjusts the speed of the drive motor to control the amplitude of the mechanical vibration generated by the device. Using the motor frequency as an input parameter, a dynamic model is established that shows how vibration intensity responds to frequency changes. Based on these two control channel models, a dual-event trigger mechanism is constructed to minimize unnecessary control operations while ensuring response accuracy, thereby optimizing energy consumption and extending device life. This mechanism consists of two components: a state trigger condition, which determines whether the current operating state has significantly changed compared to the previous control; and a control trigger condition, which evaluates whether the currently calculated control instruction is sufficiently different from the previously executed instruction. Only when both conditions are met will the system trigger a new control operation. On this basis, to ensure the control system dynamically adjusts its control strategy based on the current health status and aging rate of the equipment, an adaptive control method was introduced. By analyzing the current global health indicator value and the state evolution rate reflecting health trends, the amplitude and direction of the temperature and vibration control inputs were dynamically adjusted, forming a preliminary control strategy with real-time responsiveness. Due to factors such as sensor sensitivity variations, load disturbances, and external interference in real-world environments, a relative sensitivity error compensation mechanism was introduced. This mechanism continuously monitors the deviation between the control results and the desired state to determine the actual effectiveness of the current control strategy. Based on this, the gain and response amplitude of the control input are fine-tuned to enhance interference resistance and improve control accuracy and stability. This interference-resistant control input, combined with a dual-event triggering mechanism, forms a control execution scheme that combines energy-saving characteristics with intelligent adjustment capabilities. A global stability analysis of the energy-optimized control execution scheme was conducted. Through theoretical modeling and experimental verification, it was confirmed that the control system can quickly converge to the target state under different operating conditions, avoiding control failure or excessive energy consumption caused by frequent adjustments or feedback oscillation. The target control scheme was ultimately obtained.

[0018] In this embodiment, two independent nominal control inputs are calculated based on the temperature control mathematical model and the vibration control mathematical model. The first nominal control input is used to adjust the output power of the heating wire to control the temperature of the test environment, and the second nominal control input is used to adjust the motor frequency to regulate the overall vibration intensity of the device. These two nominal control inputs are generated solely based on the inherent dynamic relationship between the current set target and the control model, without any health feedback information. They serve as the initial control signals for the system under an ideal, undisturbed state. To adapt the control strategy to the current power supply aging state, a global health indicator and a state evolution rate indicator are introduced into the control process as key feedback signals reflecting the actual operating state and potential trends of the system. A control compensation calculation is then performed to obtain a parameter control compensation amount. This compensation amount reflects the degree of correction to the nominal control amount due to the current system health. Specifically, when the system is severely aged or aging is accelerated, the compensation amount is appropriately increased to strengthen the control response, while otherwise it is weakened. The parameter control compensation amount is superimposed on the first and second nominal control inputs to obtain a complete control law. This control law comprehensively considers the three factors of target drive, model prediction, and health feedback, and has stronger dynamic response capabilities and personalized adjustment characteristics. To prevent problems such as increased energy consumption or execution jitter caused by overly frequent control system adjustments, a pre-set trigger judgment function logically evaluates the current system state and control input changes to determine whether the conditions for triggering a control update are met. Only when the magnitude of the state change or control intent exceeds a preset threshold will the control signal update be activated, generating a control input with event-triggered characteristics. Dynamic gain tracking error feedback adjustment is performed on event-triggered control inputs. Based on the tracking error generated during actual operation, the control signal gain factor is fine-tuned to ensure that the system maintains good control accuracy and responsiveness despite nonlinear fluctuations, external disturbances, or sensor noise. Through continuous tracking error monitoring and feedback correction, a preliminary control strategy with high adaptability, dynamic responsiveness, and energy optimization capabilities is formed.

[0019] In this embodiment, the device is driven to perform a test task according to the target control scheme. During operation, key response data of the power supply, including the time-varying processes of output voltage, current, and power, is collected in real time. Using this time-series data as input samples, a Gaussian process cubic polynomial model is fitted with time as the independent variable and output response as the dependent variable. This model effectively captures the slow-changing trends and local fluctuations in the power supply response process by constructing a cubic function with continuous derivative properties. This cubic polynomial model is then differentiated to obtain the first-order derivative of the response function with respect to time, i.e., the rate of change of the power supply response, thereby characterizing its speed of change and dynamic response characteristics. After obtaining the response function and rate of change function, linear moment analysis is performed on each function to quantify their statistical distribution characteristics. Moments including mean, variance, skewness, and kurtosis are calculated to construct a set of power supply response characteristics, which reflects the stability, symmetry, and extreme fluctuations of the power supply output process. Based on the characteristic attributes characterized by the linear moments, marginal probability density functions and cumulative distribution functions are constructed for the power supply response value and its rate of change, respectively, to obtain a univariate probability model reflecting the individual distribution characteristics of the two variables. Considering the highly nonlinear and asymmetric correlation between the power supply response and its rate of change, a set of candidate dependency structure models, including Gaussian Copula, t Copula, Clayton, Gumbel, and Frank, are introduced. Based on information criteria such as the AIC value, the dependency structure with the smallest error and the best fit is selected as the optimal dependency model between the power supply response and its rate of change. A joint probability density function is then constructed that accurately reflects their joint distribution behavior. Based on this joint model, two threshold conditions are set: the power supply response exceeds the fault limit or its rate of change exceeds the dynamic threshold. The probability of falling into the failure region in the joint distribution is calculated, and an index of the power supply's first failure probability is obtained. This index serves as a basis for evaluating the reliability of the power supply under specific test environments and control parameters. Finally, a multi-objective optimization problem is constructed with the first failure probability as one of the objective functions, combined with factors such as test energy consumption and test duration. Intelligent search methods such as the particle swarm optimization algorithm are used to solve the problem, resulting in a set of optimal test parameter combinations that achieve high reliability, low energy consumption, and reasonable test duration. These parameters include the optimal temperature range, optimal vibration intensity setting, and reasonable test duration.

[0020] In one example, multi-source heterogeneous signal acquisition and preprocessing are performed on the power aging test process of a discrete energy-saving power aging test device to obtain a multi-dimensional time series data matrix, including: Set up multiple temperature measurement points in the test environment of the discrete energy-saving power supply aging test equipment, and install a temperature sensor at each temperature measurement point to collect raw temperature monitoring data; Install triaxial MEMS accelerometers at the four corners of the tray support frame of the discrete energy-saving power supply aging test equipment and collect raw vibration monitoring data; Collect raw electrical parameter data including voltage, current, and power factor from the input and output ports of the power supply under test in a discrete energy-saving power supply aging test device; Filtering the temperature monitoring raw data, vibration monitoring raw data and electrical parameter raw data to obtain target temperature monitoring data, target vibration monitoring data and target electrical monitoring data; The target temperature monitoring data, target vibration monitoring data and target electrical monitoring data are time synchronized and matrix converted to obtain a multi-dimensional time series data matrix.

[0021] In this example, multiple temperature measurement points are arranged within the test environment of a discrete, energy-efficient power supply burn-in tester based on heat source distribution, air flow paths, and temperature gradient characteristics. Each point reflects distinct heat conduction and convection zones within and outside the device. These points are strategically located within key areas such as the test sample concentration area, near the electrical heat source, and at the top and bottom of the test chamber. This ensures that the collected temperature data is locally sensitive while also reflecting the dynamic changes in the overall thermal environment. At each temperature point, a high-precision, responsive temperature sensor is installed. Industrial-grade PT100 platinum resistance sensors are used, offering excellent linear output characteristics, low error drift, and high temperature resolution, enabling accurate measurement of thermal environment changes within a range of -50°C to 150°C. Each sensor is connected to the front-end acquisition unit via an independent signal channel, with a sampling frequency of 1Hz. This not only accommodates the slow temperature changes characteristic of the burn-in process, but also helps reduce data bandwidth and storage requirements, thereby lowering system power consumption. Furthermore, triaxial MEMS accelerometers are installed within the structural support areas of the burn-in tester, specifically at the four corners of the tray supporting the power supply sample under test. This selected point is mechanically stable and fully detects the mechanical vibration effects caused by power supply operation. Triaxial MEMS sensors, due to their high integration, strong sensitivity, fast response, and low power consumption, are suitable for continuous vibration detection and dynamic feature extraction. Each sensor has a sampling frequency of 200Hz, covering micro-vibration signals from low-frequency mechanical vibration to high-frequency electromagnetic interference, effectively recording vibration signals of different frequency bands and their mutation characteristics generated during the aging process. Raw electrical parameter data, including voltage, current, and power factor, is collected from the input and output ports of the power supply under test in a discrete, energy-saving power supply aging tester. A high-precision data acquisition card is directly connected to the input and output ports of the power supply under test in analog or digital form, enabling high-frequency and high-fidelity electrical signal capture. A sampling rate of 10kHz ensures accurate capture of high-frequency fluctuations, ripple, voltage drops, current spikes, and other phenomena in the power supply response, meeting the requirements for identifying microsecond-level dynamic events. Filtering is performed on the raw temperature, vibration, and electrical parameter data. The filtering operation configures parameters according to the frequency characteristics of the data type. The temperature data is processed using a first-order Butterworth low-pass filter with a cutoff frequency set to 0.5Hz to eliminate short-term abnormal fluctuations caused by external wind disturbances, power supply fluctuations, or analog circuit noise. The vibration data is processed using a second-order Butterworth filter combined with a median filter. The former is used to filter out high-frequency sensor noise, and the latter is used to eliminate abnormal peaks caused by mechanical structure resonance or impact. The electrical parameter data is processed using a high-order low-pass filter with a cutoff frequency of 3kHz based on the 10kHz sampling characteristic. A sliding average window smoothing algorithm is introduced to address high-frequency spikes to ensure that the final extracted signal retains important features while removing high-frequency interference.After filtering, the target temperature monitoring data, target vibration monitoring data, and target electrical monitoring data are obtained. These data are then time-synchronized and matrix-converted. Through a resampling mechanism, data streams with different sampling frequencies are uniformly converted into time series with the same time step. At the same time, all data are timestamped using the master clock as a reference, ensuring that data from different channels has consistent time semantics at the same moment. After data synchronization is complete, all target data is organized into a standard matrix structure, arranged in time-order as a set of two-dimensional matrices. Each column represents a specific physical parameter, such as a temperature sensor output, a vibration signal in a certain direction, or a sampled value of an electrical parameter, while each row corresponds to a data snapshot at a discrete sampling moment. Ultimately, a multidimensional time series data matrix is ​​formed.

[0022] In one example, a multidimensional time series data matrix is ​​decomposed into a temperature subspace, a vibration subspace, and an electrical parameter subspace. Data features of the temperature subspace, the vibration subspace, and the electrical parameter subspace are fused to obtain a fused feature vector of the power supply aging state, including: Decompose the multidimensional time series data matrix into temperature subspace, vibration subspace and electrical parameter subspace; Statistical feature extraction is performed on the target temperature monitoring data in the temperature subspace to obtain a temperature feature vector; time domain feature and energy feature extraction is performed on the target vibration monitoring data in the vibration subspace to obtain a vibration feature vector; and feature extraction is performed on the target electrical monitoring data in the electrical parameter subspace to obtain an electrical parameter feature vector; Performing feature decomposition and orthogonality transformation on the temperature feature vector, vibration feature vector and electrical parameter feature vector respectively to obtain temperature decomposition feature, vibration decomposition feature and electrical parameter decomposition feature; Based on the temperature decomposition characteristics, vibration decomposition characteristics and electrical parameter decomposition characteristics, a cross-feature analysis is performed to obtain a cross-feature set, which includes the power supply output voltage stability index, temperature-efficiency response index, vibration immunity index, power conversion consistency index and startup response time index; The temperature decomposition features, vibration decomposition features, electrical parameter decomposition features and cross feature sets are integrated into attention weighted fusion to obtain the power supply aging status fusion feature vector.

[0023] In this example, multidimensional time series data is structurally classified at the data structure level. The data matrix is ​​logically divided according to the source and functional attributes of the physical parameters, and three independent data sub-matrices are extracted, namely the temperature subspace, the vibration subspace, and the electrical parameter subspace. For the temperature subspace, considering that the temperature signal is a slowly changing process, a sliding time window is used to calculate its statistical characteristics, including six dimensions of indicators such as mean, maximum, minimum, standard deviation, rising slope and falling slope, to construct a temperature feature vector to reflect the temperature stability, thermal shock response capability and local variation trend during the test; for the vibration subspace, the three-axis vibration signal is analyzed in the time domain, and its root mean square value, peak value, peak-to-peak value, pulse factor, waveform factor and other indicators are extracted to reflect the vibration intensity and impact characteristics, and then the fast Fourier transform algorithm is used to analyze its spectral distribution. The power supply is partitioned and the frequency domain energy distribution is calculated for the three frequency bands (0-50Hz, 50-100Hz, and 100-200Hz) to capture the energy characteristics between periodic vibration and non-stationary shocks, constructing a multidimensional vibration eigenvector. For the electrical parameter subspace, multiple indicators that directly characterize power supply performance fluctuations, electrical output stability, and thermoelectric coupling effects are extracted from the perspectives of power transmission efficiency, voltage output deviation, input voltage stability, current ripple coefficient, and temperature-efficiency sensitivity. This constructs an electrical parameter eigenvector that accurately describes the aging response of the power supply in the electrical dimension. To eliminate interference caused by different units, inconsistent orders of magnitude, and strong correlations between the different feature dimensions, a unified eigendecomposition and orthogonality transformation are performed on each of the three eigenvectors. The temperature, vibration, and electrical parameter eigenvectors are normalized to a uniform scale. Eigenvalue decomposition is then performed based on the covariance matrix. The original feature space is mapped to a new orthogonal space by constructing an eigentransformation matrix, resulting in independent temperature, vibration, and electrical parameter decomposition features with zero mean and unit variance, respectively. Cross-feature analysis is performed based on temperature decomposition features, vibration decomposition features, and electrical parameter decomposition features to reveal the coupling relationship between different physical mechanisms and the comprehensive characteristics of aging behavior. The power supply output voltage stability index is calculated, which measures its stability by analyzing the ratio of the standard deviation to the mean of the output voltage during the test period. The temperature-efficiency response index is defined, which characterizes the linkage response between thermal load and electrical performance by statistically analyzing the amplitude of efficiency changes caused by temperature changes. The vibration immunity index is calculated, which models the amplitude of output voltage interference based on changes in vibration acceleration to evaluate the structure's anti-interference ability. The power conversion consistency index evaluates the stability of the conversion process based on the variance level of the power output, and the startup response time index is used to characterize the time span from the power supply startup to the stabilization of electrical parameters. Together, these indicators constitute a cross-feature set.The temperature, vibration, and electrical parameter decomposition features are combined with a cross-feature set through an attention-weighted fusion process. This automatically learns the importance of different features in aging characterization and adjusts their fusion weights accordingly. A trainable attention weight matrix is ​​constructed, assigning different weight coefficients to the temperature, vibration, and electrical parameter decomposition features, as well as the cross-feature set. The weighting process prioritizes the features that are most sensitive to aging state changes and have the highest discriminative power. All features are fused through a weighted sum to form a fused feature vector for the power supply aging state.

[0024] In one example, a power supply health assessment is performed based on a fusion feature vector of the power supply aging state to obtain a global health indicator and a state evolution rate indicator, including: For each power supply under test, N sample points in the initial stage of normal operation are selected as a training set. The aging state fusion feature vector of the power supply is input into the training set to perform normal state modeling and obtain the normal working feature space of the power supply. Based on the normal working characteristic space of the power supply, a subspace whitening support vector data description algorithm is executed. By solving the minimum closed hypersphere optimization problem and introducing the radial basis kernel function for calculation, the support vector and its corresponding Lagrange multiplier are obtained. Based on the support vector and Lagrange multiplier, the relative distance between the fusion feature vector of the power supply aging state and the center of the normal area is calculated to obtain the health index value; Based on the health index values, a multidimensional health index including temperature-related health indicators, vibration-related health indicators and electrical parameter-related health indicators is constructed for the temperature subspace, vibration subspace and electrical parameter subspace respectively; Conduct sensitivity analysis on multidimensional health indicators, determine weight coefficients, and perform weighted summation of health indicator values ​​and multidimensional health indicators to obtain a global health indicator; By calculating the changing trend of the global health index in consecutive τ time windows, the state evolution rate index is obtained.

[0025] In this example, at the beginning of the burn-in test—the initial time interval when the power supply under test is functionally stable and in normal condition—the system automatically selects N consecutive time windows of fused feature vectors representing the power supply's aging state. These samples are defined as the healthy state training set. The selection of this training set ensures that the collected samples are within standard ranges with minimal fluctuation across multiple metrics, including performance, stability, temperature, current, voltage, and vibration, representing the "optimal operating state" of the power supply when unaffected by aging. After the training set is constructed, all fused feature vectors are input into the healthy modeling module, and healthy state modeling is performed based on the support vector data description (SVM) principle. To improve the feature discrimination capability and numerical stability of the modeling process, subspace whitening is performed on the training samples before SVM modeling. This ensures that the samples exhibit approximately orthogonal distributions, consistent variance, and zero mean in the feature space, effectively avoiding model bias caused by differences in feature scales. Entering the core modeling phase, the system uses the fused feature vectors after subspace whitening as input and executes the support vector data description algorithm. It constructs a minimal closed hypersphere in the high-dimensional feature space that encloses all normal samples. By solving a convex optimization problem with a minimum radius, it determines the boundary of the normal feature space and selects the support vectors and their corresponding Lagrange multiplier values ​​that constitute the boundary. Considering the complex distribution of power supply operating conditions across nonlinear dimensions, to enhance the model's adaptability to nonlinear characteristics, a radial basis function (RBF) is introduced as a kernel function to map the original features into a higher-dimensional kernel space, where linear discrimination is then performed. This results in a stronger hypersphere envelopment and more flexible classification boundaries, effectively addressing the high coupling between feature dimensions. After the model is completed, the distance between the fused feature vector of the power supply aging status acquired at any time and the center of the normal feature space is determined. The greater the distance, the greater the deviation of the feature vector from the normal range, indicating significant aging or performance degradation of the tested power supply. This distance is defined as the health index value at the current moment. By continuously tracking this value, the health fluctuation trajectory of the power supply throughout the test cycle can be effectively quantified. Considering the inconsistent contributions of different physical subspaces to the overall aging performance, the subfeatures in the fused feature vector are projected back into the original three subspaces: the temperature subspace, the vibration subspace, and the electrical parameter subspace. The degree of health deviation in each dimension is calculated, and temperature-related health indicators, vibration-related health indicators, and electrical parameter-related health indicators are constructed. These three health indicators together form a multidimensional health indicator system. A sensitivity analysis of the multidimensional health indicators is conducted. Through historical sample analysis and model perturbation experiments, the degree of influence of each dimensional health indicator on the overall aging risk assessment results is identified, and a reasonable weight coefficient is assigned to each sub-indicator. Assuming that the total weight is one, the health sub-indicators are weighted and summed according to the weight ratio to obtain a global health indicator that uniformly expresses the overall health status of the system.The larger the value of this indicator, the further the system's operating status deviates from the normal range and the more severe the aging. The timeline is divided into sliding windows of fixed length, each covering τ consecutive time steps. The rate of change of the global health indicator is calculated in each time window. This is calculated by subtracting the health indicator at the beginning of the window from the health indicator at the end of the current window and dividing the result by the window length. This indicator reflects the speed of power supply aging and reveals whether there are nonlinear characteristics such as staged mutations, stable plateaus, or degradation rebound during the aging process. In particular, when the state evolution rate shows a continuous upward trend, it indicates that the power supply has entered a rapid aging stage, and the testing strategy should be adjusted or the early warning mechanism should be triggered in a timely manner. If the state evolution rate approaches zero or decreases, it indicates that the system operation is becoming stable.

[0026] In one example, based on global health indicators and state evolution rate indicators, adaptive control parameter analysis was performed on the temperature control loop and vibration control loop of a discrete energy-saving power supply aging test equipment under dual event triggering conditions. The target control solution was obtained, including: PWM modulation is used to control the heating wire power output in the temperature control loop of the discrete energy-saving power supply aging test equipment, and a temperature control mathematical model is established. The frequency converter is used to adjust the motor speed in the vibration control loop of the discrete energy-saving power supply aging test equipment, and a vibration control mathematical model is established. Construct dual event trigger conditions, which include state trigger conditions and control trigger conditions. The state trigger condition is set as the deviation between the current state and the state at the last trigger exceeds the trigger threshold. The control trigger condition is set as the deviation between the currently calculated control input and the last executed control input exceeds the control threshold. Based on the temperature control mathematical model and the vibration control mathematical model, dynamic gain adaptive control analysis is performed on the global health index and the state evolution rate index to obtain a preliminary control strategy; The preliminary control strategy is compensated for relative sensitivity errors to obtain a control input with enhanced anti-interference capability. This control input is then combined with a dual-event trigger condition to obtain an energy-optimized control execution scheme. A global stability analysis is performed on the energy consumption optimized control execution scheme to obtain the target control scheme.

[0027] In this example, a heating wire heating control method based on pulse-width modulation (PWM) technology is employed. PWM signals serve as control commands to adjust the heating wire's power output, achieving continuous and stable control of the ambient temperature. This method offers the advantages of fast response and high control precision. By varying the PWM signal's duty cycle, the heating power can be flexibly adjusted, adapting to various test scenarios. Based on this, a temperature control mathematical model is developed based on the experimental environment's thermal inertia, heat capacity distribution, and environmental disturbances. This model establishes a quantitative functional relationship between the dynamic changes in ambient temperature and the PWM signal. A thermal disturbance term is introduced to simulate the uncertain influence of the external environment, enabling the model to describe the thermal response of a real-world system. Corresponding to the temperature control loop, the vibration intensity of the entire test platform is adjusted by adjusting the speed of the drive motor when controlling vibration parameters. This control method uses a frequency converter as the actuator. The input frequency signal controls the motor's operating speed, thereby regulating the vibration output level. To accurately describe the response of the vibration system, a mathematical vibration control model is constructed by combining the motor-load coupling characteristics, the structural transfer function, and the damping characteristics of the vibration platform. This model expresses the nonlinear dynamic relationship between vibration intensity and motor frequency input as a discrete-time recursive system model. A disturbance term is introduced to simulate vibration fluctuations caused by power supply load variations or small displacements of the platform's mechanical structure. Building on the two aforementioned control models, a dual-event triggering mechanism is introduced to address the high energy consumption, resource waste, and system over-response issues associated with high-frequency execution of traditional periodic control. The dual-event triggering mechanism consists of a state trigger condition and a control trigger condition. The state trigger condition determines whether the deviation between the current actual operating state and the state at the time of the previous control exceeds a set threshold. This means that a control update is triggered only when the system state change is significant enough to indicate a control demand has occurred. The control trigger condition detects the deviation between the currently calculated control input and the last actual control input. If this deviation exceeds the threshold, it indicates that the new control instruction has significant regulatory value for the current system and should trigger control execution immediately. These two conditions together form a "and logic" judgment mechanism. Control operations are only executed when the state change is significant enough and the control input update is meaningful. This reduces unnecessary control frequency, lowers overall energy consumption, and improves the targetedness of system response. Based on the above model and triggering logic, dynamic gain adaptive control analysis is performed in conjunction with the current global health indicator and state evolution rate indicator. The health indicator is used as quantitative feedback on system performance deviations, and the state evolution rate is used as dynamic feedback on performance change trends. Both factors contribute to the control gain adjustment logic.During the implementation, the severity of the current deviation from the normal state is assessed based on a global health indicator. A high indicator indicates deterioration in system health, and the control response should be appropriately strengthened. The state evolution rate is also used to determine the rate of change in the current health level. A significant trend indicates accelerated degradation, and the control strategy requires stronger feedforward response capabilities. Based on this, adaptive control gains are generated by adjusting the gain parameters of the control inputs in the temperature and vibration control models, forming a preliminary control strategy that precisely intervenes in the current test state. To enhance the robustness and interference tolerance of the control system, a relative sensitivity error compensation mechanism is introduced to modify the control strategy. This mechanism uses feedback calculations based on the tracking error between the control output and the target state. When the system response resulting from the control input deviates from the expected state, the amplitude and direction of the current control command are corrected by adjusting the gain or compensation amount, thereby mitigating deviations caused by non-ideal factors such as environmental fluctuations, equipment aging, and signal hysteresis. This compensation process is a real-time, dynamic, and self-correcting process, ensuring that the control system maintains response stability and control accuracy over long-term operation. By combining the compensated control inputs with a dual-event triggering mechanism, an energy-optimized control execution scheme with the triple capabilities of execution judgment, precision enhancement, and energy consumption control is constructed. To verify the effectiveness and reliability of this control strategy, a global stability analysis method is introduced. By establishing a stability function for the control system in the discrete time domain or introducing a Lyapunov-like function, it is evaluated whether the control system can always remain within the bounded operating range under arbitrary disturbances and arbitrary initial states, and whether oscillations, divergence, or hysteresis will occur, thereby verifying whether the system has global convergence and asymptotic stability. If the stability conditions are met, the control scheme is confirmed as the target control scheme for the current operating conditions.

[0028] In one example, based on the temperature control mathematical model and the vibration control mathematical model, a dynamic gain adaptive control analysis is performed on the global health index and the state evolution rate index to obtain a preliminary control strategy, including: calculating a first nominal control input for temperature control and a second nominal control input for vibration control according to a temperature control mathematical model and a vibration control mathematical model; Compensation control is performed based on the global health index and the state evolution rate index to obtain the parameter control compensation amount; The first nominal control input and the second nominal control input are respectively added to the parameter control compensation to obtain a complete control law, and then a trigger judgment function is used to determine whether to execute a control update of the complete control law to obtain an event-triggered control input; Dynamic gain tracking error feedback adjustment is performed on the event-triggered control input to obtain a preliminary control strategy.

[0029] In this example, two independent sets of nominal control inputs are calculated based on the temperature control and vibration control mathematical models, respectively, according to the current system operating objectives and the physical characteristics of the equipment. The temperature control mathematical model reflects the dynamic mapping between the PWM duty cycle and the actual test environment temperature, taking into account factors such as the heating wire's response inertia, thermal diffusion hysteresis, and ambient thermal disturbances. Based on this model, the system infers a theoretical control input based on the current desired temperature curve. This first nominal control input represents the reference signal for maintaining the desired temperature under the absence of any abnormalities. Similarly, the vibration control mathematical model describes the functional relationship between the inverter frequency adjustment and the mechanical vibration intensity. This model integrates the motor response curve, structural transfer characteristics, and load disturbances. Based on this, the second nominal control input is derived: a control frequency input calculated based on the desired vibration intensity, which is used to maintain the desired vibration amplitude under standard conditions. These two nominal control inputs do not account for changes in the equipment's aging state and are initial commands in the context of ideal static control. To enable the control system to respond to changes in aging conditions, a global health indicator and a state evolution rate indicator, updated in real time during the test, are incorporated into the control logic as key feedback parameters for the current power supply operating status and its dynamic trends. The global health indicator measures the degree to which the current characteristic state deviates from the normal model, while the state evolution rate indicator characterizes whether this deviation is accelerating or stabilizing. Based on these two indicators, a compensation control algorithm is executed to construct a parameterized control compensation, which is used to adjust the response amplitude and direction of the nominal control input. When the health indicator increases significantly, indicating that the power supply's operating state is deviating from the ideal state, control intervention is strengthened, and the system increases the compensation. When the state evolution rate changes negatively, indicating that the device's state is recovering, the system appropriately reduces the control amplitude, thus achieving dynamic regulation. This compensation applies simultaneously to the temperature and vibration control paths. The first nominal control input is added to the compensation to obtain the adjusted temperature control signal, while the second nominal control input is added to the compensation to obtain the adjusted vibration control signal. Together, these two constitute the complete control law. Considering that frequent control updates not only increase system load and energy consumption but also induce oscillation or instability in the execution structure, an event trigger mechanism is introduced to determine whether the control law should be executed. The trigger judgment function simultaneously evaluates two aspects: first, whether the deviation between the current state and the last triggered state exceeds the set state trigger threshold; second, whether the difference between the current calculated control input and the last actual control input exceeds the control trigger threshold. Only when both conditions are met does the system consider the current control update necessary, allowing the full control law to be executed, and defining this as an event-triggered control input. Dynamic gain tracking error feedback adjustment is performed on this event-triggered control input. This process uses the tracking error between the control output and the desired response as the core basis to determine the effectiveness of the current control signal and adjust the control gain accordingly.If the current control output fails to effectively approach the target state, it indicates insufficient control input response. The system will then increase the gain coefficient to strengthen control. If the control response is ahead of schedule or even overshoots, the system will reduce the gain to prevent overcontrol. This adjustment process is based on continuous online monitoring, combining historical control trajectories with current error trends. A learning rate control strategy is employed to prevent oscillations caused by overly rapid gain adjustments, ensuring that the control system maintains stable and responsive control behavior under complex, dynamic, and multi-disturbance conditions. The control signal after dynamic gain adjustment constitutes the initial control strategy for the current test cycle.

[0030] In one example, the discrete energy-saving test automation method further includes: Control the operation of discrete energy-saving power supply aging test equipment according to the target control scheme, collect power supply response data, and establish a cubic polynomial model of Gaussian process based on the power supply response data; Perform differential operation on the cubic polynomial model to obtain the rate of change, and calculate the power supply response data and the linear moment of the rate of change respectively to obtain the power supply response characteristic set; Based on the power response characteristic set, the marginal probability density function and cumulative distribution function of the power response and change rate are created to obtain the univariate probability distribution characteristics; Calculate the optimal dependency structure model between power supply response and its rate of change based on the univariate probability distribution characteristics; The joint probability density function is constructed using the optimal dependency structure model. The first failure probability of the joint probability density function is then calculated based on the failure threshold and the change rate threshold to obtain the reliability evaluation index of the power supply parameters. A multi-objective optimization solution is performed based on the reliability evaluation indicators of power supply parameters to obtain the optimal power supply test parameter combination.

[0031] In this example, while the system operates according to the target control scheme, response data from the power supply under test is continuously collected. This response data includes time-varying physical quantities such as temperature, voltage, current, and power, covering state changes throughout the entire burn-in test cycle. Data collection maintains the same time base and sampling frequency as the previous control data to ensure data consistency and comparability. Based on sufficient data accumulation, a cubic polynomial model is constructed based on the Gaussian process concept to reflect the evolution of the power supply response. By fitting the collected time series response data, a cubic time function model with smooth curvature is represented. This transforms the originally highly volatile response data into a unified and well-structured time-continuous function, making subsequent analytical operations mathematically feasible and derivable. The cubic polynomial model is then differentiated to obtain its first-order derivative, yielding the corresponding rate of change at each moment. This reflects the speed and direction of the power supply output response and reveals the system's state regulation capability and operational stability under controlled conditions. Linear moments of the power supply response data and rate of change are calculated separately. By extracting statistical quantities such as expectation, variance, skewness, and kurtosis, a feature set is formed, including the response mean, response variance, average amplitude of the rate of change, degree of response asymmetry, and sharpness of dynamic changes. This is known as the power supply response characteristic set. Based on this power supply response characteristic set, marginal probability density modeling is performed for the response value and its rate of change. At the implementation level, nonparametric kernel density estimation or parametric fitting methods are used to construct the marginal probability density function for each variable, and the corresponding cumulative distribution function is derived to statistically characterize the univariate distribution characteristics of the response and its dynamic changes. These distribution functions reflect the probability distribution pattern of the power supply output characteristic values ​​and are key tools for describing the stability and uncertainty of power supply performance. To characterize the interdependence between the power supply response and its rate of change, a joint modeling strategy is introduced based on two univariate distributions. Copula functions are introduced to construct a dependency structure model. By capturing the nonlinear correlation between marginal variables in a high-dimensional space, a joint distribution framework for the power supply response and its rate of change is constructed. To ensure the optimal fit of the selected copula structure, candidate function sets such as Gaussian, t-type, Clayton, Gumbel, and Frank were evaluated one by one. The copula type that best suited the current data distribution structure was selected based on the Akaike Information Criterion to establish the optimal dependency structure model of the power supply response system. This optimal dependency structure model was combined with the marginal probability density function to construct a joint probability density function of the power supply response and its rate of change. This function reflects the probabilistic characteristics of the power supply system in the multidimensional response space and is used to further evaluate its risk characteristics during actual operation.Based on this, if a response threshold is set as the benchmark for determining potential system failures, and a rate-of-change threshold is set to identify sudden abnormal behavior, the probability of the system's first failure is calculated by integrating the cumulative distribution corresponding to the region of the joint probability density function outside the threshold boundary. This reflects the likelihood that the power supply will first breach the stable operating boundary within a certain period under the current test conditions, and thus provides a reliability assessment index for the power supply parameters. The reliability assessment results are then incorporated into a multi-objective optimization mechanism to construct an optimization framework that integrates test safety, energy consumption, and time efficiency. The optimization objectives include minimizing test energy consumption and shortening the test cycle while ensuring that the probability of first failure does not exceed a set tolerance. To achieve this balance, multiple weight coefficients are set to construct a comprehensive objective function. The test temperature range, vibration amplitude range, and test duration are included as optimization variables, and corresponding constraints are set. During the solution process, a particle swarm optimization algorithm, genetic algorithm, or other heuristic search strategies are used to explore the global solution space. Through multiple rounds of iteration, the optimal power supply test parameter combination is converged. This parameter combination meets system reliability requirements and achieves an optimal balance between test efficiency, energy saving, and device load. The optimal test parameters obtained will be fed back to the automated control system to form a closed-loop iterative mechanism, improving the intelligence, precision and adaptability of discrete energy-saving power supply aging tests.

[0032] Reference Figure 2 , this embodiment provides a discrete energy-saving test automation device, including: Acquisition module 1 is used to collect and preprocess multi-source heterogeneous signals during the power aging test process of the discrete energy-saving power aging test equipment to obtain a multi-dimensional time series data matrix; Feature fusion module 2 is used to decompose the multidimensional time series data matrix into temperature subspace, vibration subspace and electrical parameter subspace, and perform data feature fusion on the temperature subspace, vibration subspace and electrical parameter subspace to obtain the power supply aging state fusion feature vector; Evaluation module 3, used to perform power supply health assessment based on the power supply aging state fusion feature vector to obtain global health indicators and state evolution rate indicators; Analysis module 4 is used to perform adaptive control parameter analysis of dual event triggering conditions on the temperature control loop and vibration control loop of the discrete energy-saving power supply aging test equipment based on the global health index and the state evolution rate index to obtain a target control solution.

[0033] In this embodiment, for the specific implementation of each unit in the above device embodiment, please refer to the above method embodiment, which will not be repeated here.

[0034] Reference Figure 3In an embodiment of the present invention, a computer device is also provided. The computer device may be a server, and its internal structure may be as follows: Figure 3 As shown. The computer device includes a processor, memory, display screen, input device, network interface and database connected via a system bus. The processor of the computer design is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The database of the computer device is used to store the corresponding data in this embodiment. The network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, the above method is implemented.

[0035] Those skilled in the art will understand that Figure 3 The structure shown in the figure is merely a block diagram of a portion of the structure related to the solution of the present invention and does not constitute a limitation on the computer device to which the solution of the present invention is applied.

[0036] An embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon, which implements the above-described method when executed by a processor. It is understood that the computer-readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0037] Those skilled in the art will appreciate that all or part of the processes in the above-described method embodiments can be implemented by instructing the relevant hardware using a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the above-described method embodiments. Any reference to memory, storage, database, or other media provided herein and used in the embodiments may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double-speed SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct RAMbus dynamic RAM (DRDRAM), and RAMbus dynamic RAM.

[0038] It should be noted that, in this document, the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, apparatus, article, or method comprising a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, apparatus, article, or method. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, apparatus, article, or method comprising the element.

[0039] The above description is only a preferred embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the present invention description and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A discrete energy-saving test automation method, characterized in that: include: Perform multi-source heterogeneous signal acquisition and preprocessing on the power aging test process of discrete energy-saving power aging test equipment to obtain a multi-dimensional time series data matrix; Decomposing the multidimensional time series data matrix into a temperature subspace, a vibration subspace, and an electrical parameter subspace, and performing data feature fusion on the temperature subspace, the vibration subspace, and the electrical parameter subspace to obtain a power supply aging state fusion feature vector; Performing power supply health assessment based on the power supply aging state fusion feature vector to obtain a global health index and a state evolution rate index; According to the global health index and the state evolution rate index, adaptive control parameter analysis of dual event triggering conditions is performed on the temperature control loop and the vibration control loop of the discrete energy-saving power supply aging test equipment to obtain a target control solution.

2. The discrete energy-saving test automation method according to claim 1, characterized in that: The power aging test process of the discrete energy-saving power supply aging test equipment is subjected to multi-source heterogeneous signal acquisition and preprocessing to obtain a multi-dimensional time series data matrix, including: Set up multiple temperature measurement points in the test environment of the discrete energy-saving power supply aging test equipment, and install a temperature sensor at each temperature measurement point to collect raw temperature monitoring data; Installing triaxial MEMS acceleration sensors at the four corner points of the tray support frame of the discrete energy-saving power supply aging test equipment and collecting raw vibration monitoring data; Collecting raw electrical parameter data including voltage, current, and power factor from the input port and output port of the power supply under test in the discrete energy-saving power supply aging test equipment; Filtering the temperature monitoring raw data, the vibration monitoring raw data, and the electrical parameter raw data to obtain target temperature monitoring data, target vibration monitoring data, and target electrical monitoring data; Time synchronization and matrix conversion are performed on the target temperature monitoring data, the target vibration monitoring data, and the target electrical monitoring data to obtain a multi-dimensional time series data matrix.

3. The discrete energy-saving test automation method according to claim 2, characterized in that: The step of decomposing the multidimensional time series data matrix into a temperature subspace, a vibration subspace, and an electrical parameter subspace, and performing data feature fusion on the temperature subspace, the vibration subspace, and the electrical parameter subspace to obtain a power supply aging state fusion feature vector includes: Decomposing the multidimensional time series data matrix into a temperature subspace, a vibration subspace, and an electrical parameter subspace; Performing statistical feature extraction on the target temperature monitoring data of the temperature subspace to obtain a temperature feature vector; performing time domain feature and energy feature extraction on the target vibration monitoring data of the vibration subspace to obtain a vibration feature vector; and performing feature extraction on the target electrical monitoring data of the electrical parameter subspace to obtain an electrical parameter feature vector; Performing feature decomposition and orthogonality conversion on the temperature feature vector, the vibration feature vector, and the electrical parameter feature vector respectively to obtain temperature decomposition features, vibration decomposition features, and electrical parameter decomposition features; Performing a cross-feature analysis based on the temperature decomposition feature, the vibration decomposition feature, and the electrical parameter decomposition feature to obtain a cross-feature set, wherein the cross-feature set includes a power supply output voltage stability index, a temperature-efficiency response index, a vibration immunity index, a power conversion consistency index, and a startup response time index; The temperature decomposition feature, the vibration decomposition feature, the electrical parameter decomposition feature and the cross feature set are subjected to attention weighted fusion to obtain a power supply aging state fusion feature vector.

4. The discrete energy-saving test automation method according to claim 1, characterized in that: The power supply health assessment is performed based on the power supply aging state fusion feature vector to obtain a global health index and a state evolution rate index, including: For each power supply under test, N sample points in the initial stage of normal operation are selected as a training set, and the power supply aging state fusion feature vector is input into the training set to perform normal state modeling, thereby obtaining a normal operating feature space of the power supply; A subspace whitening support vector data description algorithm is executed based on the normal working characteristic space of the power supply, and the support vector and its corresponding Lagrange multiplier are obtained by solving the minimum closed hypersphere optimization problem and introducing the radial basis kernel function for calculation; Calculating the relative distance between the power supply aging state fusion feature vector and the center of the normal area according to the support vector and the Lagrange multiplier to obtain a health index value; Based on the health indicator values, constructing a multidimensional health indicator including temperature-related health indicators, vibration-related health indicators and electrical parameter-related health indicators for the temperature subspace, vibration subspace and electrical parameter subspace respectively; Performing a sensitivity analysis on the multidimensional health index to determine a weight coefficient, and performing a weighted sum operation on the health index value and the multidimensional health index to obtain a global health index; By calculating the changing trend of the global health index in consecutive τ time windows, the state evolution rate index is obtained.

5. The discrete energy-saving test automation method according to claim 1, characterized in that: The method of performing adaptive control parameter analysis of dual event triggering conditions on the temperature control loop and the vibration control loop of the discrete energy-saving power supply aging test equipment based on the global health indicator and the state evolution rate indicator to obtain a target control solution includes: Using PWM modulation to control the power output of the heating wire in the temperature control loop of the discrete energy-saving power supply aging test equipment, establishing a temperature control mathematical model, and using a frequency converter to adjust the motor speed in the vibration control loop of the discrete energy-saving power supply aging test equipment, establishing a vibration control mathematical model; Constructing a dual event trigger condition, wherein the dual event trigger condition includes a state trigger condition and a control trigger condition, wherein the state trigger condition is set to that the deviation between the current state and the state at the time of the last trigger exceeds a trigger threshold, and the control trigger condition is set to that the deviation between the currently calculated control input and the last executed control input exceeds a control threshold; Based on the temperature control mathematical model and the vibration control mathematical model, performing dynamic gain adaptive control analysis on the global health index and the state evolution rate index to obtain a preliminary control strategy; performing relative sensitivity error compensation on the preliminary control strategy to obtain a control input with enhanced anti-interference performance, and combining the control input with the dual event triggering condition to obtain a control execution plan with optimized energy consumption; A global stability analysis is performed on the energy consumption optimized control execution scheme to obtain a target control scheme.

6. The discrete energy-saving test automation method according to claim 5, characterized in that: The performing of dynamic gain adaptive control analysis on the global health index and the state evolution rate index based on the temperature control mathematical model and the vibration control mathematical model to obtain a preliminary control strategy includes: Calculating a first nominal control input for temperature control and a second nominal control input for vibration control according to the temperature control mathematical model and the vibration control mathematical model respectively; Performing compensation control based on the global health indicator and the state evolution rate indicator to obtain a parameter control compensation amount; Adding the first nominal control input and the second nominal control input to the parameter control compensation to obtain a complete control law, and then determining whether to execute a control update of the complete control law according to a trigger judgment function to obtain an event-triggered control input; Dynamic gain tracking error feedback adjustment is performed on the event-triggered control input to obtain a preliminary control strategy.

7. The discrete energy-saving test automation method according to claim 1, characterized in that: The discrete energy-saving test automation method further includes: Controlling the operation of the discrete energy-saving power supply aging test device according to the target control scheme, collecting power supply response data, and establishing a cubic polynomial model of the Gaussian process based on the power supply response data; Performing a differential operation on the cubic polynomial model to obtain a rate of change, and respectively calculating the power supply response data and a linear moment of the rate of change to obtain a power supply response characteristic set; Creating a marginal probability density function and a cumulative distribution function of the power supply response and the rate of change based on the power supply response characteristic set to obtain a univariate probability distribution characteristic; Calculating an optimal dependency structure model between power supply response and its rate of change based on the univariate probability distribution characteristics; Constructing a joint probability density function using the optimal dependency structure model, and then calculating the first failure probability of the joint probability density function based on a failure threshold and a change rate threshold to obtain a reliability evaluation index of the power supply parameters; A multi-objective optimization solution is performed based on the reliability evaluation index of the power supply parameters to obtain the optimal power supply test parameter combination.

8. A discrete energy-saving test automation device, characterized in that: For implementing the steps of the discrete energy-saving test automation method according to any one of claims 1 to 7, the discrete energy-saving test automation device comprises: The acquisition module is used to collect and preprocess multi-source heterogeneous signals during the power aging test process of the discrete energy-saving power aging test equipment to obtain a multi-dimensional time series data matrix; a feature fusion module, configured to decompose the multidimensional time series data matrix into a temperature subspace, a vibration subspace, and an electrical parameter subspace, and perform data feature fusion on the temperature subspace, the vibration subspace, and the electrical parameter subspace to obtain a fused feature vector of the power supply aging state; An evaluation module, configured to perform a power supply health evaluation based on the power supply aging state fusion feature vector to obtain a global health index and a state evolution rate index; The analysis module is used to perform adaptive control parameter analysis of dual event triggering conditions on the temperature control loop and vibration control loop of the discrete energy-saving power supply aging test equipment according to the global health index and the state evolution rate index to obtain a target control solution.

9. A computer device comprising a memory and a processor, wherein a computer program is stored in the memory, wherein: When the processor executes the computer program, the steps of the discrete energy-saving test automation method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the discrete energy-saving test automation method according to any one of claims 1 to 7 are implemented.

Images