Method and system for evaluating wear state of high-voltage electric mechanical part

Abstract

The invention relates to the technical field of high-voltage electric power, and particularly discloses a high-voltage electric mechanical part wear state evaluation method and system.The method comprises the steps that multi-sensor time sequence signals, operation overshoot mechanical kinetic energy and contact surface temperature are collected, a normalized wear signal sequence is obtained through kinetic energy compensation, and multi-dimensional features are extracted; a comprehensive feature vector is generated through lightweight attention mechanism fusion, a thermal coupling factor correction wear mode is calculated based on temperature, a real-time state and lubrication dielectric characteristics are fused to obtain an evaluation aggressive coefficient, levels are mapped, and the arc erosion cumulant after environment correction is combined to obtain a high-precision arc erosion model. And generating an evaluation report of the content index, the composite fault and the residual life. According to the method, accurate decoupling of a concurrent wear mode and self-adaption of an evaluation strategy working condition can be realized, evaluation precision and embedded deployment requirements are considered, the problems of rough evaluation, neglect of multi-factor coupling and poor adaptability in the prior art are effectively solved, and reliable support is provided for predictive maintenance of high-voltage electric mechanical parts.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of high-voltage power, in particular to a high-voltage power mechanical component wear state evaluation method and system. BACKGROUND

[0002] Cam, connecting rod, pin shaft and other key mechanical components in high-voltage switches, circuit breakers and other power equipment are the core support for ensuring the safe and stable operation of the power grid. Such components are prone to concurrent conditions of multiple wear modes such as abrasive wear and fatigue spalling in long-term opening and closing operations, and are subjected to mechanical impact, friction and wear, arc erosion, thermal coupling and other multiple actions. If the wear state cannot be accurately evaluated in a timely manner, it will lead to component failure, equipment malfunction or misoperation, and cause serious power grid accidents.

[0003] In the prior art, high-voltage power mechanical component wear state evaluation mainly relies on two types of methods. One is the traditional signal analysis-based diagnosis method, which extracts time-frequency domain features of vibration, acoustic emission and other signals manually, and combines machine learning algorithms such as support vector machine and decision tree to classify the state. This method is severely dependent on expert experience and feature engineering quality, and is not only inefficient, but also difficult to capture early weak wear signals, and cannot distinguish the contribution of concurrent wear modes, thus failing to provide quantitative basis for accurate maintenance. The other is an intelligent evaluation method based on deep learning, which uses one-dimensional convolutional neural network and recurrent neural network models to automatically extract features. Although this method improves the evaluation accuracy, it has a rough evaluation granularity, and most models only output a single health score or a simple state label, which cannot decouple concurrent wear modes, cannot judge the main wear type and the severity of each type, and the attention mechanism is applied shallowly, mostly global or channel-level coarse-grained attention, which is difficult to separate different wear features in the feature space, resulting in evaluation deviation and difficulty in supporting accurate maintenance decisions.

[0004] Therefore, there is an urgent need for a high-voltage power mechanical component wear state evaluation method and system to solve the above problems. SUMMARY

[0005] The application aims to provide a high-voltage power mechanical component wear state evaluation method, which comprises the following steps: Collecting multi-sensor time sequence signals, operating overshoot mechanical kinetic energy and contact surface micro-zone transient temperature during the operation of the high-voltage power mechanical component, compensating the multi-sensor time sequence signals based on the overshoot mechanical kinetic energy, and obtaining a kinetic energy normalized wear signal sequence; According to the kinetic energy normalized wear signal sequence, time domain wear characteristics, frequency domain resonance characteristics and time-frequency domain surface morphology characteristics are extracted, a material micro-ductility attenuation coefficient is obtained according to historical load cycle data and accumulated operation overshoot mechanical kinetic energy, attention weights of multi-dimensional characteristics are modulated based on a lightweight attention mechanism and according to the material micro-ductility attenuation coefficient, and a comprehensive feature vector is fused and generated; According to the comprehensive feature vector, a current dominant wear mode type is identified, a thermal mechanical fatigue coupling factor is calculated based on transient temperature of a micro area of the contact surface, wear mode probabilities are recalibrated, and a wear mode correction result after thermal force coupling correction is output; Real-time state data and dielectric property data of a lubricating medium are obtained, an instantaneous health coefficient and a lubricating film failure risk factor are calculated, a working condition adaptive evaluation aggressiveness coefficient is fused, and the evaluation aggressiveness coefficient is mapped to an evaluation grade; According to a historical breaking current sequence and a contact material ablation rate, an arc erosion cumulative amount is obtained, a multi-dimensional wear state evaluation report containing quantitative indexes of each wear mode, a composite failure mode and a residual service life interval is generated based on the wear mode correction result, the evaluation grade and the arc erosion cumulative amount.

[0006] Further, the step of compensating a multi-sensor time sequence signal based on the overshoot mechanical kinetic energy to obtain a kinetic energy normalized wear signal sequence comprises: An institution motion stationarity factor is obtained, and motion artifact elimination processing is performed on an original signal according to the institution motion stationarity factor; An electromagnetic interference characteristic spectrum is obtained, an electromagnetic interference signal-to-noise ratio is calculated based on the electromagnetic interference characteristic spectrum, and electromagnetic interference filtering is performed on the signal after motion artifact elimination according to the electromagnetic interference signal-to-noise ratio; Wavelet threshold denoising processing is performed on the filtered signal, the signal is decomposed into multiple frequency bands through discrete wavelet transform, and a threshold function based on signal sparsity adaptive adjustment is applied to remove noise components; A kinetic energy compensation coefficient is calculated based on the operation overshoot mechanical kinetic energy, the compensation coefficient is hardness corrected in combination with hardness characteristics of a contact surface material, and kinetic energy normalized compensation is performed on the denoised signal; A multi-source signal coherence index is calculated, the compensated signal is fused and enhanced based on the coherence index, a signal sequence of a key operation stage is intercepted, and a kinetic energy normalized wear signal sequence is formed.

[0007] Further, the step of modulating attention weights of multi-dimensional characteristics based on a lightweight attention mechanism and according to the material micro-ductility attenuation coefficient to fuse and generate a comprehensive feature vector comprises: Time domain statistical characteristics, frequency domain resonance peak characteristics and time-frequency domain wavelet packet energy characteristics are extracted from the kinetic energy normalized wear signal sequence; Based on historical data of the entire life cycle of mechanical components, the number of historical load cycles and the cumulative overshoot mechanical kinetic energy are obtained, and the material micro-toughness attenuation coefficient is calculated through a material fatigue damage model. Construct a multi-branch lightweight attention network, with each branch processing wear features of different dimensions; The material's micro-toughness attenuation coefficient is used as prior knowledge and input into the attention network. The attention weights of each feature dimension are dynamically adjusted to generate a comprehensive feature vector characterizing the degradation of material properties.

[0008] Furthermore, the step of recalibrating the wear mode probability and outputting the wear mode correction result after thermal coupling correction includes: The comprehensive feature vector is input into a pre-trained multi-classification model to obtain the initial probability distribution of each wear mode; Based on the transient temperature data of the micro-region of the contact surface, combined with the thermal expansion coefficient and thermal conductivity of the material, the thermomechanical fatigue coupling factor is calculated. The initial probability distribution is recalibrated based on the thermomechanical fatigue coupling factor to enhance the weight of wear modes dominated by thermal stress. The probability distribution after recalibration is normalized to determine the current dominant wear mode type and its confidence level; Output the wear mode correction results that incorporate the thermo-coupling effect.

[0009] Furthermore, the steps of acquiring real-time status data and lubricating medium dielectric property data, calculating instantaneous health coefficient and lubricating film failure risk factor, fusing them to obtain an adaptive assessment aggressive coefficient for operating conditions, and mapping the assessment aggressive coefficient to an assessment level include: Obtain real-time operating status data from the equipment monitoring system, including runtime, load current, and mechanism stroke characteristics; Obtain the dielectric constant and dielectric loss factor of the lubricating medium, and calculate the lubricating film failure risk factor based on the trend of dielectric property changes; Calculate the instantaneous health coefficient by combining real-time operational status data; Based on the instantaneous health coefficient and the lubrication film failure risk factor, the adaptive evaluation aggressive coefficient for operating conditions is obtained; Based on the preset evaluation level threshold, the evaluation radical coefficient is mapped to three evaluation levels: conservative, balanced, and radical.

[0010] Furthermore, the step of generating a multi-dimensional wear status assessment report based on the wear mode correction results, assessment level, and arc erosion accumulation, including quantitative indicators of each wear mode, composite failure mode, and remaining service life range, includes: The historical interruption current sequence is obtained. Based on the different effects of different current levels on contact erosion, the current level weighting coefficient is calculated. Combined with the contact material ablation rate model and the current level weighting coefficient, the cumulative amount of arc erosion is calculated. Data on ambient humidity and pollutant concentration are obtained, and an environmental medium insulation recovery factor is calculated based on the environmental medium insulation recovery characteristics. The cumulative amount of arc erosion is then modified according to the environmental medium insulation recovery factor to adapt to environmental conditions. Based on the wear mode correction results, the main wear mechanisms and their development rates are determined, and the wear development rate is dynamically corrected by introducing the mechanism fit clearance change rate. The assessment strategy is determined by combining the assessment level, the operation frequency acceleration factor is calculated based on the operation frequency statistics of the institution, and the life model parameters are optimized and adjusted according to the operation frequency acceleration factor. Calculate the coupling acceleration effect coefficient between mechanical wear and arc erosion, and identify mechanical-electrical composite fault modes and their risk levels based on the coupling acceleration effect coefficient; A multi-parameter fusion model for predicting remaining life is constructed, and an arc reignition risk threshold is introduced as a life termination criterion. A multi-dimensional wear status assessment report is generated, which includes wear quantification indicators, composite fault mode diagnosis, risk level assessment, and remaining service life range.

[0011] Furthermore, this application also discloses a wear condition assessment system for high-voltage power machinery components, comprising: The acquisition module is used to acquire multi-sensor time-series signals, overshoot mechanical kinetic energy, and transient temperature of micro-area of ​​contact surface during the operation of high-voltage electric mechanical components. Based on the overshoot mechanical kinetic energy, the multi-sensor time-series signals are compensated to obtain a kinetic energy normalized wear signal sequence. The fusion module is used to extract time-domain wear features, frequency-domain resonance features, and time-frequency-domain surface morphology features based on the kinetic energy normalized wear signal sequence; obtain the material micro-toughness attenuation coefficient based on historical load cycle data and accumulated operation overshoot mechanical kinetic energy; and modulate the attention weights of multi-dimensional features based on the lightweight attention mechanism and the material micro-toughness attenuation coefficient to fuse and generate a comprehensive feature vector. The output module is used to identify the current dominant wear mode type based on the comprehensive feature vector, calculate the thermomechanical fatigue coupling factor based on the transient temperature of the micro-area of ​​the contact surface, recalibrate the wear mode probability, and output the wear mode correction result after thermo-coupling correction. The evaluation module is used to acquire real-time status data and dielectric property data of lubricating medium, calculate instantaneous health coefficient and lubricating film failure risk factor, fuse them to obtain the adaptive evaluation aggressive coefficient of working condition, and map the evaluation aggressive coefficient to the evaluation level. The generation module is used to obtain the cumulative amount of arc erosion based on the historical breaking current sequence and the contact material ablation rate, and to generate a multi-dimensional wear status assessment report based on the wear mode correction results, assessment level and cumulative amount of arc erosion, which includes quantitative indicators of each wear mode, composite failure mode and remaining service life range.

[0012] Furthermore, the output module includes: The input unit is used to input the comprehensive feature vector into the pre-trained multi-classification model to obtain the initial probability distribution of each wear mode; The calculation unit is used to calculate the thermomechanical fatigue coupling factor based on the transient temperature data of the micro-area of ​​the contact surface, combined with the material's thermal expansion coefficient and thermal conductivity characteristics. A calibration unit is used to recalibrate the initial probability distribution based on the thermomechanical fatigue coupling factor, thereby enhancing the weight of wear modes dominated by thermal stress. The processing unit is used to normalize the probability distribution after recalibration and determine the current dominant wear mode type and its confidence level. An output unit is provided for outputting the wear mode correction results based on the fusion of thermo-coupling effects. This application also provides a computer device, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps of the aforementioned method for assessing the wear condition of high-voltage power machinery components.

[0013] This application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the above-described method for assessing the wear condition of high-voltage power machinery components.

[0014] This application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described method for assessing the wear condition of high-voltage power machinery components.

[0015] The beneficial effects of this application are as follows: First, this application enables precise quantitative decomposition of concurrent wear modes, solving the problems of coarse granularity and inability to distinguish the contribution of multiple modes in existing technologies. By eliminating signal interference through kinetic energy normalization, extracting multi-dimensional wear features, and modulating attention weights by combining the material's micro-toughness attenuation coefficient, and then recalibrating the wear mode probabilities through a thermo-coupling factor, it can accurately decouple concurrent modes such as abrasive wear and fatigue spalling, and output independent quantitative indicators for each mode, greatly improving the precision and interpretability of the assessment.

[0016] Secondly, this application improves the accuracy of feature extraction and wear pattern recognition, solving the problems of coarse-grained attention mechanisms and neglecting multi-factor coupling in existing technologies. By employing a multi-branch lightweight attention network, feature weights are dynamically adjusted based on material degradation characteristics, making feature extraction more adaptable to the material state. Simultaneously, it incorporates thermal coupling effects, arc erosion accumulation, and environmental correction factors, effectively compensating for the shortcomings of existing technologies in ignoring multi-physics coupling and environmental influences, and significantly improving the accuracy of wear pattern recognition and quantitative assessment. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of a method flow proposed in an embodiment of this application.

[0018] Figure 2 This is a schematic diagram of the system structure proposed in one embodiment of this application.

[0019] Figure 3 This is a schematic diagram of the internal structure of a computer device according to an embodiment of this application.

[0020] The realization of the purpose, functional features and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0021] It should be understood that the specific embodiments described herein are merely illustrative of this application and are not intended to limit this application.

[0022] like Figure 1 As shown, this application provides a method for assessing the wear condition of high-voltage power machinery components, including the following steps: S1, collect multi-sensor time-series signals, overshoot mechanical kinetic energy and transient temperature of micro-area of ​​contact surface during the operation of high-voltage electric mechanical components, and perform compensation processing on the multi-sensor time-series signals based on the overshoot mechanical kinetic energy to obtain a kinetic energy normalized wear signal sequence; S2, extract time-domain wear features, frequency-domain resonance features, and time-frequency-domain surface morphology features based on the kinetic energy normalized wear signal sequence, obtain the material micro-toughness attenuation coefficient based on historical load cycle data and accumulated operation overshoot mechanical kinetic energy, modulate the attention weights of multi-dimensional features based on the lightweight attention mechanism and the material micro-toughness attenuation coefficient, and fuse them to generate a comprehensive feature vector oriented by material properties. S3. Identify the current dominant wear mode type based on the comprehensive feature vector, calculate the thermomechanical fatigue coupling factor based on the transient temperature of the micro-area of ​​the contact surface, recalibrate the wear mode probability, and output the wear mode correction result after thermo-coupling correction. S4: Obtain real-time status data and lubrication medium dielectric property data, calculate instantaneous health coefficient and lubrication film failure risk factor, fuse them to obtain the adaptive evaluation aggressive coefficient of working condition, and map the evaluation aggressive coefficient to the evaluation level; S5. Based on the historical breaking current sequence and contact material ablation rate, the cumulative amount of arc erosion is obtained. Based on the wear mode correction results, evaluation level and cumulative amount of arc erosion, a multi-dimensional wear status assessment report is generated, which includes quantitative indicators of each wear mode, composite failure mode and remaining service life range.

[0023] As described in steps S1-S5 above, high-voltage power mechanical components, as key components of core power grid equipment, face multiple impacts during opening and closing operations, including mechanical shock, arc erosion, thermal stress, and lubrication failure. Wear modes exhibit multi-mechanism concurrent characteristics, and signals are easily distorted by operating conditions such as kinetic energy fluctuations and electromagnetic interference. The attenuation of material microstructures and the thermo-coupling effect further exacerbate the complexity of wear assessment. Existing technologies largely rely on single sensor signals, failing to consider signal deviations caused by kinetic energy fluctuations. Feature fusion does not dynamically adjust weights based on material degradation characteristics, wear mode identification ignores the influence of thermo-coupling, and assessment results are general and lack adaptability to operating conditions. They cannot accurately distinguish complex fault modes or predict remaining service life, making it difficult to meet the high-precision maintenance requirements of high-voltage equipment.

[0024] This application integrates multi-source operating condition data, material property parameters, and a lightweight attention mechanism. Through signal compensation, feature optimization and fusion, wear mode correction, adaptive grading of operating conditions, and comprehensive evaluation of multiple factors, it achieves accurate quantification, multi-mode differentiation, and full life-cycle adaptation assessment of the wear state of high-voltage power machinery components. In principle, it is based on a systematic consideration of the coupled effects of multiple factors. First, it collects multi-sensor time-series signals, operational overshoot mechanical kinetic energy, and contact surface temperature. Kinetic energy compensation signals are used to eliminate interference, resulting in a normalized wear signal sequence. Then, time-domain, frequency-domain, and time-frequency-domain features are extracted from this sequence, combined with the degradation of materials with historical load and accumulated kinetic energy. The micro-toughness attenuation coefficient is dynamically adjusted by a lightweight attention mechanism to form a comprehensive feature vector that adapts to the material degradation law. The wear mode is identified by comprehensive features, and the probability of the mode is corrected by temperature calculation of the thermo-coupling factor to solve the identification bias caused by thermo-coupling. Then, the aggressive coefficient is calculated and mapped to the level by real-time status data and lubrication dielectric properties to make the evaluation strategy adapt to the working conditions. Finally, the accumulated arc erosion, wear mode and evaluation level after environmental correction are combined to generate a report that includes quantitative indicators, composite faults and remaining life. This comprehensively considers mechanical and electrical coupling and environmental impact, forming a complete logical chain from data to evaluation.

[0025] In one embodiment, the step of acquiring the kinetic energy normalized wear signal sequence includes: S11 collects the original signals of high-voltage electric mechanical components during the opening and closing operation through vibration acceleration sensor and acoustic emission sensor, and simultaneously collects the mechanical kinetic energy of operation overshoot and the transient temperature of micro-area of ​​contact surface; S12, Based on the mechanism motion stability analysis, obtain the mechanism motion stability factor, and perform motion artifact elimination processing on the original signal according to the mechanism motion stability factor; S13, acquire the electromagnetic interference feature spectrum, calculate the electromagnetic interference signal-to-noise ratio based on the electromagnetic interference feature spectrum, and perform electromagnetic interference filtering on the signal after eliminating motion artifacts according to the electromagnetic interference signal-to-noise ratio. S14, perform wavelet threshold denoising on the filtered signal, decompose the signal into multiple frequency bands through discrete wavelet transform, and apply a threshold function based on signal sparsity adaptive adjustment to remove noise components. S15, calculate the kinetic energy compensation coefficient based on the mechanical kinetic energy of the operation overshoot, and perform hardness correction on the compensation coefficient in combination with the hardness characteristics of the contact surface material, and perform kinetic energy normalization compensation on the denoised signal. S16, calculate the coherence index of the multi-source signal, fuse and enhance the compensated signal based on the coherence index, extract the signal sequence of the key operation stage, and form the kinetic energy normalized wear signal sequence.

[0026] As described in steps S11-S16 above, by synchronously collecting multi-source data during the operation of high-voltage electric machinery components, and through motion artifact elimination, electromagnetic interference filtering, adaptive denoising, kinetic energy normalization compensation, and multi-source signal fusion enhancement, a kinetic energy normalized wear signal sequence that can truly and stably reflect the wear state of the components is obtained.

[0027] During the opening and closing operations of high-voltage power mechanical components, fluctuations in the mechanism's speed can generate motion artifacts. Electromagnetic interference from the power grid environment can superimpose onto sensor signals, and changes in the mechanical kinetic energy during operation can lead to differences in sensor signal amplitudes for the same degree of wear. All these factors cause the original signal to deviate from the true wear state. When the high-voltage switch cam is opening and closing, if the impact speed of the mechanism exceeds the rated value, even if the cam wear is the same, the vibration signal amplitude will be larger. If this deviation is not addressed, subsequent feature extraction will misjudge the degree of wear. At the same time, electromagnetic interference can mask weak wear information in the signal, making it impossible to capture early, slight wear. Therefore, these interferences and deviations must be eliminated to ensure that the signal accurately represents the wear state.

[0028] By collecting raw vibration and acoustic emission signals from the opening and closing operations of high-voltage power mechanical components, and simultaneously collecting the operational overshoot mechanical kinetic energy and transient temperature of the contact surface micro-area, the operational overshoot mechanical kinetic energy is determined. Operational overshoot mechanical kinetic energy refers to the extra kinetic energy carried and ultimately dissipated by moving parts (such as cams, connecting rods, pins, etc.) after reaching their preset mechanical stroke endpoint during circuit breaker closing or opening operations, due to inertia, as they continue to move and compress the buffer device (such as springs, hydraulic buffers, rubber pads, etc.). The operational overshoot mechanical kinetic energy is measured by the maximum impact velocity using speed sensors installed on the moving parts of the mechanism, and combined with the component mass using the formula E = The transient temperature of the micro-area at the contact surface is obtained through mv² calculation and is collected in real time by an infrared temperature sensor installed near the contact surface of the component. A motion stability factor is obtained based on the motion stability analysis of the mechanism. By analyzing the motion velocity curves of the mechanism in historical operations, the standard deviation of velocity fluctuation is calculated and normalized to a stability factor in the 0-1 range. For example, when the standard deviation of velocity fluctuation is 0.2 m / s, the stability factor is 0.8. Motion artifacts are eliminated from the original signal based on the stability factor, and the signal is weighted and corrected with the stability factor to reduce abnormal signal fluctuations caused by motion instability.

[0029] Electromagnetic interference (EMI) characteristic spectrum is obtained by collecting background signals from sensors when the equipment is shut down and analyzing the frequency distribution of the background signals. If the main interference frequencies are concentrated around 50Hz and harmonics, the EMI signal-to-noise ratio (SNR), i.e., the amplitude ratio of the useful signal to the interference signal, is calculated based on this characteristic spectrum. Based on the SNR, the artifact-free signal is then subjected to EMI filtering using an adaptive notch filter to attenuate the signal in the interference frequency band. The filtered signal is then subjected to wavelet threshold denoising. Discrete wavelet transform is used to decompose the signal into five frequency bands: below 1kHz, 1-5kHz, 5-10kHz, 10-20kHz, and above 20kHz. Each frequency band corresponds to different wear information. A threshold function adaptively adjusted based on signal sparsity is applied; a smaller threshold is used when the signal sparsity is high, and a larger threshold is used when the sparsity is low. For example, the threshold is set to 0.05 when the high-frequency band sparsity is 0.6, and the threshold is set to 0.1 when the low-frequency band sparsity is 0.3, thus removing noise while avoiding the loss of useful wear information.

[0030] The kinetic energy compensation coefficient is calculated based on the mechanical kinetic energy of the overshoot. The ratio of the actual kinetic energy to the rated kinetic energy is used as the base coefficient. When the actual kinetic energy is 1.2 times the rated kinetic energy, the base coefficient is 1.2. The hardness characteristics of the contact surface material are then corrected. For example, the correction coefficient is 1.0 when the hardness of steel is 250HB and 0.8 when the hardness of aluminum is 100HB. The final compensation coefficient is then obtained. The denoised signal is normalized and compensated using the final compensation coefficient, that is, the signal amplitude is divided by the compensation coefficient. This unifies the signal amplitude under different kinetic energy fluctuations to the rated kinetic energy benchmark. For example, if the kinetic energy causes the signal amplitude to be 20% larger, the amplitude will be restored to the true level after compensation, ensuring that the signal is comparable across operating conditions. The coherence index of multi-source signals is calculated by calculating the coherence function values ​​of vibration and acoustic emission signals at each frequency band. Based on this index, the signals are weighted and fused for compensation. The frequency bands with high coherence are given higher weights, such as 0.9 for coherence and 0.5 for coherence. After fusion, the key stages of opening and closing, such as the contact separation stage, are extracted. A signal sequence with a duration of 0.1 seconds is formed to create a kinetic energy normalized wear signal sequence. This sequence eliminates various interferences and deviations and can accurately reflect the wear state.

[0031] In one embodiment, the step of fusing and generating a comprehensive feature vector guided by material properties based on a lightweight attention mechanism and modulating the attention weights of multi-dimensional features according to the material's micro-toughness attenuation coefficient includes: S21, extract time-domain statistical features, frequency-domain resonant peak features, and time-frequency-domain wavelet packet energy features from the kinetic energy normalized wear signal sequence, respectively; S22, based on the historical data of the entire life cycle of mechanical components, obtains the number of historical load cycles and the cumulative operating overshoot mechanical kinetic energy, and calculates the material micro-toughness attenuation coefficient through the material fatigue damage model; S23, construct a multi-branch lightweight attention network, with each branch processing wear features of different dimensions; S24, the material's micro-toughness attenuation coefficient is used as prior knowledge and input into the attention network to dynamically adjust the attention weight allocation of each feature dimension, giving higher attention to feature dimensions with severe toughness attenuation; through weighted fusion and orthogonal regularization constraints, a comprehensive feature vector characterizing the degradation of material properties is generated.

[0032] As described in steps S21-S24 above, this application extracts multi-dimensional wear features from the kinetic energy normalized wear signal sequence, combines them with the micro-toughness attenuation coefficient of the mechanical component material, and uses a multi-branch lightweight attention network to dynamically adjust the attention weights of each feature dimension, ultimately generating a comprehensive feature vector that can accurately characterize the degradation of material properties and wear state.

[0033] The wear process of high-voltage power machinery components is accompanied by multi-dimensional information changes. Time-domain statistical features can reflect the impact intensity changes caused by wear, frequency-domain resonance peak features can reflect the changes in component structural stiffness caused by wear, and time-frequency domain wavelet packet energy features can capture the differences in energy distribution of local wear. These three types of features correspond to key information of different wear mechanisms. At the same time, the micro-toughness of the component material gradually decreases with the increase of historical load cycles and the cumulative consumption of mechanical kinetic energy from overshoot. After the material toughness decreases, the sensitivity of different dimensional features to wear changes. After the toughness of metallic materials decreases, the frequency-domain resonance peak features are significantly more sensitive to fatigue spalling than the time-domain features. If all features are treated equally, the features will not be able to accurately match the wear mechanism under the current material condition, affecting the accuracy of subsequent evaluations.

[0034] Three types of features were extracted from the kinetic energy normalized wear signal sequence. The time-domain statistical features include peak value and kurtosis, which were calculated using the sliding window method. In the high-voltage switch cam opening and closing signal, abrasive wear will increase the time-domain peak value from 0.5g to 0.8g. The frequency-domain resonance peak features were obtained by converting the signal to the frequency domain using fast Fourier transform. The frequency band from 1kHz to 10kHz was analyzed, and the resonance peak amplitudes at 2kHz and 5kHz were extracted. These two frequencies correspond to the structural resonance frequencies of the cam and connecting rod. Fatigue spalling will reduce the resonance peak amplitude at 2kHz. The time-frequency domain wavelet packet energy features were obtained by decomposing the signal into four layers of wavelet packets using the db4 wavelet basis. The energy values ​​of the eight nodes in the third layer were calculated. Local pitting wear will increase the energy of the fifth node in the third layer. These three types of features completely capture wear information from different dimensions.

[0035] Key parameters are obtained based on historical data throughout the entire lifecycle of mechanical components. Historical load cycle counts are retrieved from the equipment operation log database. For example, the connecting rod of a certain type of high-voltage circuit breaker's operating mechanism has completed 8000 opening and closing cycles. The cumulative overshoot mechanical kinetic energy is calculated by summing the overshoot mechanical kinetic energy of each historical operation. The kinetic energy for each operation is calculated by combining the maximum impact velocity measured by the velocity sensor with the component's mass. The cumulative overshoot mechanical kinetic energy of this connecting rod is 1200J. A linear cumulative damage model is used to calculate the material's micro-toughness attenuation coefficient. The formula for calculating the micro-toughness attenuation coefficient is as follows: ; in, This represents the coefficient of microscopic toughness attenuation. This represents the average capacity degradation rate, calculated by comparing the current usable capacity of a component with its initial rated capacity. Capacity data is obtained from the battery management system or through offline testing. This represents the average internal resistance growth rate, obtained by comparing the DC internal resistance test results with the initial internal resistance. The internal resistance is measured using a resistance meter. Indicates the capacity decay weight. The internal resistance growth weight is represented by the micro-toughness decay coefficient, which directly reflects the degree of toughness degradation of the material due to fatigue.

[0036] A multi-branch lightweight attention network is constructed, comprising three parallel branches corresponding to time-domain, frequency-domain, and time-frequency-domain feature processing, respectively. The feature extraction layer of each branch employs a one-dimensional depthwise separable convolutional structure with a kernel size of 3 and a stride of 1. The number of input channels corresponds to the number of each feature dimension: the time-domain branch has 2 input channels, corresponding to peak values ​​and kurtosis; the frequency-domain branch has 2 input channels, representing the amplitudes of the two formants; and the time-frequency-domain branch has 8 input channels. This structure reduces computational consumption while maintaining feature extraction capabilities, making it suitable for embedded deployments. Each branch is followed by an ECA lightweight attention module, which calculates the weight coefficients of each feature channel using 1×1 convolutions, avoiding the high computational complexity of global attention.

[0037] The material's micro-toughness attenuation coefficient is input into the attention network as prior knowledge. The network dynamically adjusts the attention weights of each branch based on this coefficient. When the attenuation coefficient is 0.85, the material's toughness significantly degrades, and the frequency domain resonant peak feature shows the highest sensitivity to fatigue spalling. The network increases the attention weight of the frequency domain branch from the base value of 0.3 to 0.5, while the weight of the time domain branch is reduced from 0.4 to 0.2 due to decreased sensitivity. The weight of the time-frequency domain branch remains unchanged at 0.3 to ensure that local wear information is not lost. After weight adjustment, the features of each branch are processed through feature concatenation and batch normalization to generate a comprehensive feature vector with a dimension of 12. This comprehensive feature vector highlights the key information that best reflects the wear mechanism under the current material state. Simultaneously, the lightweight network structure reduces the model parameter size, meeting the computing power requirements of embedded platforms.

[0038] In one embodiment, the step of recalibrating the wear mode probability and outputting the wear mode correction result after thermal coupling correction includes: S31, Input the comprehensive feature vector into the pre-trained multi-classification model to obtain the initial probability distribution of each wear mode; S32, Based on the transient temperature data of the micro-area of ​​the contact surface, combined with the thermal expansion coefficient and thermal conductivity of the material, calculate the thermomechanical fatigue coupling factor; S33, the initial probability distribution is recalibrated according to the thermomechanical fatigue coupling factor to enhance the weight of wear modes dominated by thermal stress; S34, the probability distribution after recalibration is normalized by the softmax function to determine the current dominant wear mode type and its confidence level; S35 outputs the wear mode correction results that incorporate the thermo-coupling effect.

[0039] As described in steps S31-S35 above, this application obtains the initial probability distribution of wear modes by inputting the comprehensive feature vector into a pre-trained multi-classification model, calculates the thermomechanical fatigue coupling factor by combining the transient temperature of the micro-area of ​​the contact surface, recalibrates and normalizes the initial probability, and finally outputs the wear mode correction result that integrates the thermo-mechanical coupling effect, thereby realizing accurate wear mode identification that takes into account the influence of thermo-mechanical coupling and providing a reliable pattern basis for subsequent wear state assessment.

[0040] During the opening and closing operations of high-voltage power machinery components, the contact surfaces experience micro-temperature increases due to mechanical friction and Joule heating from the current, generating thermal stress. The superposition of thermal and mechanical stress exacerbates thermomechanical fatigue, increasing the proportion of wear modes dominated by thermal stress, such as fatigue spalling. At the contact surface between the high-voltage switch cam and pin, the transient temperature in the micro-area can rise from room temperature to 65°C during opening and closing. This thermal stress accelerates the wear rate of fatigue spalling. If wear mode identification relies solely on comprehensive feature vectors without considering the impact of thermal coupling on the wear mechanism, the probability of thermally dominated wear modes will be underestimated. This leads to wear mode judgments deviating from actual operating conditions, thus affecting the accuracy of subsequent wear quantification and life prediction.

[0041] The comprehensive feature vector is input into a pre-trained multi-classification model, which employs a lightweight MobileNet architecture CNN. The input layer dimension is 12, consistent with the comprehensive feature vector dimension. The hidden layers consist of three 1D convolutional layers with a kernel size of 3 and a stride of 1. The output layer has two neurons, corresponding to two preset wear modes: abrasive wear and fatigue spalling. The model training data comes from accelerated wear experiments in the laboratory, containing 1000 sets of comprehensive feature vectors and corresponding labels under different wear modes. The cross-entropy loss function is used during training, with 50 iterations and a convergence accuracy of [missing information]. For example, the comprehensive feature vector of a high-voltage switch cam has a time-domain peak value of 0.7g and a kurtosis of 4.2, a frequency-domain resonant peak amplitude of 0.3V at 2kHz and 0.2V at 5kHz, and the energies of the eight wavelet packet nodes in the time and frequency domains are as follows: to J, the model outputs an initial probability distribution of 60% abrasive wear and 40% fatigue spalling.

[0042] The thermomechanical fatigue coupling factor is calculated based on transient temperature data of the contact surface micro-area. This transient temperature data is acquired by a synchronously collected infrared temperature sensor. After mean filtering, the average temperature value during the opening and closing process is taken, such as an average temperature of 65℃ within a certain operating cycle. Considering the material's coefficient of thermal expansion and thermal conductivity, steel is selected as the material, with a coefficient of thermal expansion of 11× The thermal conductivity is The unit K refers to Kelvin, and the formula for calculating the thermomechanical fatigue coupling factor is: ; in, This represents the thermomechanical fatigue coupling factor. This represents the actual micro-area temperature. Indicates the optimal operating temperature of the material. Indicates the coefficient of thermal expansion. The thermal conductivity coefficient is used in the calculation formula of the thermomechanical fatigue coupling factor. The thermal conductivity coefficient needs to be normalized first to eliminate the dimensional differences between different variables. The purpose is to ensure that all variables are on the same order of magnitude, so as to make the calculation more stable and effective. The thermomechanical fatigue coupling factor directly reflects the intensity of the influence of thermal stress on the wear mode. The larger the value, the more the weight of the thermally dominated wear mode needs to be increased.

[0043] The initial probability distribution is recalibrated based on the thermomechanical fatigue coupling factor. The recalibration rule is as follows: the probability of the thermally dominated wear mode (fatigue spalling) = initial probability × (1 + thermomechanical fatigue coupling factor), and the probability of the mechanically dominated wear mode (abrasive wear) = initial probability × (1 - thermomechanical fatigue coupling factor). Substituting the initial probability and the coupling factor 0.3, the fatigue spalling probability is calculated to be 40% × (1 + 0.3) = 52%, and the abrasive wear probability is calculated to be 60% × (1 - 0.3) = 42%. This adjustment enhances the weight of the thermally dominated wear mode, making the probability distribution more closely match the actual wear proportion under thermomechanical coupling.

[0044] The recalibrated probability distribution is normalized so that the sum of the normalized probabilities of abrasive wear and fatigue spalling is 1, ensuring the rationality of the probability distribution. For example, based on the normalization result, fatigue spalling is determined to be the current dominant wear mode, with a confidence level of 55.3%. The confidence level reflects the reliability of the mode identification result; when the confidence level is higher than 50%, it is determined to be the current dominant wear mode.

[0045] The output incorporates wear mode correction results based on the thermo-coupling effect. The results include the dominant wear mode type, namely fatigue spalling, the normalized probability of each wear mode, and the confidence level of the dominant mode. This accurately reflects the actual wear mechanism under thermo-coupling. The parameters of the fatigue spalling rate model can be adjusted to ensure that the life prediction conforms to the actual wear development trend.

[0046] In one embodiment, the steps of acquiring real-time status data and lubricating medium dielectric property data, calculating instantaneous health coefficient and lubricating film failure risk factor, fusing them to obtain an adaptive assessment aggressive coefficient for operating conditions, and mapping the assessment aggressive coefficient to an assessment level include: S41, obtain real-time operating status data from the equipment monitoring system, including running time, load current and mechanism stroke characteristics; S42, the dielectric constant and dielectric loss factor of the lubricating medium are obtained by a dielectric spectrum analyzer, and the failure risk factor of the lubricating film is calculated based on the trend of dielectric property changes; S43 calculates the instantaneous health coefficient by combining real-time operating status data, reflecting the current operating status of the equipment; S44, Based on the instantaneous health coefficient and the lubrication film failure risk factor, the adaptive evaluation aggressive coefficient for operating conditions is obtained; S45, based on a preset evaluation level threshold, maps the evaluation radical coefficient to three evaluation levels: conservative, balanced, and radical.

[0047] As described in steps S41-S45 above, this application obtains real-time operating status data from the equipment monitoring system and collects dielectric property data of the lubricating medium. It calculates the instantaneous health coefficient and the lubricating film failure risk factor respectively, and merges them to obtain the adaptive evaluation aggressive coefficient of the working condition. Then, based on a preset threshold, it maps the coefficient to three evaluation levels: conservative, balanced, and aggressive, thereby realizing the dynamic adaptation of the evaluation strategy to the real-time operating status and lubrication conditions of the high-voltage electric machinery components.

[0048] Wear assessment of high-voltage power machinery components requires a comprehensive judgment based on both the real-time equipment status and lubrication conditions. In real-time operation, the longer the operating time, the greater the deviation of the load current from the rated value, and the more significant the deviation in the mechanism's stroke characteristics, the lower the component's health level, necessitating a more cautious assessment strategy. The lubricating medium, as crucial for reducing component friction and wear, directly reflects the lubricating film's condition through changes in its dielectric constant and loss factor. The greater the deviation of the dielectric constant from its initial value and the higher the loss factor, the higher the risk of lubricating film failure. Overly aggressive assessment and use will exacerbate component wear. If a mechanism has already operated for 8000 hours, and the dielectric constant of the lubricating medium has increased from an initial 2.5 to 2.9 (its limit is 3.0), continuing with an aggressive assessment strategy will cause the connecting rod to operate under high load, accelerating fatigue spalling. Ignoring lubrication failure risk based solely on equipment status, or focusing solely on lubrication while ignoring equipment health, will both lead to inaccurate assessment strategies.

[0049] Real-time operating status data is obtained from the equipment monitoring system, which integrates monitoring modules such as speed sensors, current sensors, and displacement sensors. The runtime is the cumulative operating time of the equipment from commissioning to the present, retrieved from the equipment monitoring system's operating log database. For example, the operating mechanism of a certain model of high-voltage circuit breaker has accumulated 8000 hours of operation. The load current is the real-time current value during the opening and closing operations of the component, measured by a Hall current sensor connected in series in the circuit. The rated load current is 630A, and the current measured load current is 580A. The mechanism stroke characteristic is the actual distance the component moves during the opening and closing process, measured by a laser displacement sensor installed on the mechanism's movement trajectory. The rated stroke is 12mm, and the current measured stroke is 11.7mm. Together, these data reflect the current operating load and structural status of the equipment.

[0050] To obtain data on the dielectric properties of the lubricating medium, a dielectric constant tester is used to measure the dielectric constant of the lubricating medium, such as lithium-based grease for high-voltage equipment. A dielectric loss tester is used to measure the dielectric loss factor. The testing environment temperature is maintained at 25℃, and the sampling interval is synchronized with the real-time operating status data. For example, if the initial dielectric constant of a new lubricating medium is 2.5 and the initial dielectric loss factor is 0.005, and the currently measured dielectric constant is 2.9 and the dielectric loss factor is 0.018, the lubricating film failure risk factor is calculated based on the trend of dielectric property changes. A linear fitting method is used to analyze the rate of change of dielectric constant and dielectric loss factor over the past three months. The formula for calculating the lubricating film failure risk factor is: ; in, Indicates the risk factor for lubricant film failure. Indicates the current dielectric constant. Represents the initial dielectric constant. Denotes the limiting dielectric constant, and D represents the current dielectric loss factor. Indicates the initial dielectric loss factor. This represents the limiting dielectric loss factor, where the limiting dielectric constant of the lubricating medium is set to 3.0 and the limiting dielectric loss factor is set to 0.02. Substituting the data, we get: 0.5×(2.9-2.5) / (3.0-2.5)+0.5×(0.018-0.005) / (0.02-0.005)=0.5×0.8+0.5×0.867≈0.833. The larger the value of the lubricating film failure risk factor, the higher the risk of lubricating film failure. It should be noted that the 0.5 in the above formula is the weighting coefficient of the dielectric constant change term and the dielectric loss factor change term. It is used to balance the contribution of the two to the risk of lubricating film failure, giving them equal weights (0.5 each). This can avoid the risk assessment bias caused by a single indicator dominating, and ensure that the quantification of the risk of lubricating film failure is more in line with the actual degradation process.

[0051] The instantaneous health coefficient is calculated by combining real-time operational status data. The formula for calculating the instantaneous health coefficient is as follows: ; in, Indicates the instantaneous health coefficient. Indicates runtime. Indicates the design life. Indicates the load current. Indicates the rated load current. Indicates the characteristics of the mechanism's travel. This represents the rated stroke, where the component's design life is 10,000 hours. Substituting the data, we get: 0.4×(1-8000 / 10000)+0.3×(1-|580-630| / 630)+0.3×(11.7 / 12)=0.4×0.2+0.3×0.921+0.3×0.975≈0.08+0.276+0.293≈0.649. The closer the instantaneous health coefficient is to 1, the better the component's current health status. In the above calculation formula, "0.4, 0.3, and 0.3" in the instantaneous health coefficient calculation formula are fixed weighting coefficients, derived from the wear mechanism of high-voltage power machinery components, engineering practice data, and fault impact weight analysis. They are key parameters for balancing multi-dimensional health indicators and ensuring that the coefficients accurately reflect the actual state of the component. The core of setting these weighting coefficients is closely aligned with the failure mechanism of high-voltage power machinery components, and the degree of influence of different operating state parameters on component health varies fundamentally. Cumulative operating time is directly related to material fatigue life and is a core factor determining the remaining service capability of a component. The wear it causes is irreversible, and its impact on health status should have the highest weight. Deviations in load current from the rated value exacerbate arc erosion and mechanical friction, while deviations in mechanism stroke characteristics reflect component structural deformation. Both are dynamic influencing factors during operation, and their impact on health status is similar and less significant than that of cumulative operating time. Based on this principle, cumulative operating time should be given a higher weight, while load current and mechanism stroke characteristics should be given equal and lower weights, thus forming a weight allocation logic of "0.4, 0.3, 0.3".

[0052] The adaptive assessment aggressive coefficient for operating conditions is obtained by fusing the instantaneous health coefficient and the lubrication film failure risk factor. A linear fusion formula is then used to calculate the adaptive assessment aggressive coefficient for operating conditions. The formula for calculating the adaptive assessment aggressive coefficient for operating conditions is as follows: ; in, This represents the aggressiveness coefficient of the adaptive assessment under operating conditions. In this formula, the instantaneous health coefficient has a higher weight, prioritizing the assessment of the core health level of the equipment. At the same time, the lubrication risk suppression term is included. The calculation yields 0.6×0.649-0.4×0.833≈0.389-0.333≈0.056. The aggressiveness coefficient of the adaptive assessment under operating conditions is positively correlated with the instantaneous health coefficient and negatively correlated with the lubrication film failure risk factor, which can reflect the constraint of operating conditions on the degree of aggressiveness of the assessment.

[0053] The evaluation level is mapped based on preset evaluation level thresholds, which are calibrated using laboratory operating condition simulations and engineering application data: an aggressive coefficient < 0.3 indicates a conservative level, 0.3 ≤ aggressive coefficient < 0.7 indicates a balanced level, and an aggressive coefficient ≥ 0.7 indicates an aggressive level. The currently calculated aggressive coefficient is 0.056, which falls under the conservative level. Under this level, subsequent evaluation reports will prioritize a more conservative wear rate calculation method, such as lowering the upper limit of the wear rate estimate by 15% and raising the lower limit of the remaining service life range, to avoid overloading components due to excessively aggressive evaluation. If the coefficient is 0.5 (balanced level), the accuracy of the wear evaluation is balanced with equipment protection. If the coefficient is 0.8 (aggressive level), the wear rate estimate limit can be appropriately relaxed to fully utilize the current health and lubrication conditions of the equipment and achieve reasonable utilization of its capabilities.

[0054] In one embodiment, the step of generating a multi-dimensional wear status assessment report containing quantitative indicators of each wear mode, composite failure mode, and remaining service life range based on the wear mode correction results, assessment level, and arc erosion accumulation includes: S51. Obtain historical interruption current sequence from the power grid monitoring system. Based on the different impacts of different current levels on contact erosion, calculate the current level weighting coefficient. Combine the contact material ablation rate model with the current level weighting coefficient to calculate the cumulative amount of arc erosion. S52, acquire environmental humidity and pollutant concentration data, calculate environmental medium insulation recovery factor based on environmental medium insulation recovery characteristics, and perform environmental adaptation correction on arc erosion accumulation based on the environmental medium insulation recovery factor; S53, Based on the wear mode correction results, determine the main wear mechanism and its development rate, and introduce the mechanism fit clearance change rate to dynamically correct the wear development rate; S54, determine the assessment strategy based on the assessment level, calculate the operation frequency acceleration factor based on the operation frequency statistics of the institution, and optimize and adjust the life model parameters according to the operation frequency acceleration factor; S55, calculate the coupling acceleration effect coefficient of mechanical wear and arc erosion, and identify the mechanical-electrical composite fault mode and its risk level based on the coupling acceleration effect coefficient; S56. Construct a multi-parameter fusion model for predicting remaining lifespan, introduce an arc reignition risk threshold as a lifespan termination criterion, and generate a multi-dimensional wear status assessment report that includes wear quantification indicators, composite fault mode diagnosis, risk level assessment, and remaining service life range.

[0055] As described in steps S51-S56 above, this application calculates the cumulative amount of arc erosion after environmental adaptability correction, dynamically corrects the wear development rate by combining the wear mode correction results, optimizes the life model parameters according to the assessment level, identifies mechanical and electrical composite fault modes, and finally constructs a multi-parameter fusion remaining life prediction model and generates a multi-dimensional wear status assessment report. This enables comprehensive quantification of the wear status of high-voltage power mechanical components, composite fault diagnosis, and accurate prediction of remaining life, providing complete decision support for predictive maintenance of equipment.

[0056] High-voltage power machinery components, such as high-voltage switch contacts, not only face mechanical wear such as abrasive wear and fatigue spalling during operation, but also arc erosion caused by electric arcs generated during current interruption, leading to contact material ablation. Furthermore, mechanical wear and arc erosion exhibit a coupled accelerating effect: mechanical wear creates uneven contact surfaces, exacerbating arc generation and erosion, while arc erosion causes surface depressions, increasing mechanical friction and further accelerating mechanical wear. Simultaneously, increased ambient humidity reduces the dielectric insulation recovery capacity, and increased contaminant concentrations adhere to the contact surface, both contributing to arc erosion. Changes in the clearance of the mechanism alter the stress state of mechanical wear, and increased operating frequency shortens the wear development cycle. Ignoring these factors and solely assessing mechanical wear or arc erosion leads to inaccurate wear quantification, missed diagnosis of complex faults, and inaccurate remaining life prediction, failing to meet the maintenance requirements throughout the equipment's entire lifecycle.

[0057] Historical breaking current sequences are obtained from the equipment monitoring system's operation log database. These sequences include the breaking current value and time of each opening and closing operation. The sampling frequency is synchronized with the equipment operation. For example, the historical breaking current sequence of a 12kV high-voltage vacuum circuit breaker includes 800 operations at 630A (rated current) and 200 operations at 800A (overload current). Based on the differential impact of different current levels on contact erosion, current level weighting coefficients are calculated using a linear weighting rule. The weighting coefficient for the rated current of 630A is set to 1.0, while that for the overload current of 800A, which experiences stronger erosion, it is set to 1.5. Other current levels are adjusted proportionally to their deviation from the rated current. The cumulative arc erosion is calculated using a contact material ablation rate model. Copper alloy is selected as the contact material, and its ablation rate model is "single erosion amount = ablation rate × breaking current × operation time," where the copper alloy contact ablation rate is... The duration of a single operation is 0.02s. Substituting the data, we can calculate: the erosion amount per operation at 630A = 0.01 × 630 × 0.02 = 0.126mg, the erosion amount per operation at 800A = 0.01 × 800 × 0.02 = 0.16mg, and the total cumulative arc erosion amount = 800 × 0.126 + 200 × 0.16 = 100.8 + 32 = 132.8mg.

[0058] Ambient humidity and pollutant concentration data are acquired. Ambient humidity is collected in real-time by a temperature and humidity sensor installed inside the equipment cabinet; the current measured humidity is 65%. Pollutant concentration is measured by a dust sensor inside the cabinet, with a sampling frequency of 1Hz; the current measured concentration is 0.15 mg / m³. The environmental dielectric insulation recovery factor is calculated based on the environmental dielectric insulation recovery characteristics. The formula for calculating the environmental dielectric insulation recovery factor is: ; in, Indicates the insulation recovery factor, RH represents humidity, and PM represents the concentration of pollutants. The data represents the baseline value for pollutant concentration, with humidity expressed as a percentage. Both pollutant concentration and the baseline value are in mg / m³. 0.5 represents the baseline value for pollutant concentration. Substituting the data, we get: Insulation recovery factor = 1 - 0.2 × 0.65 - 0.3 × (0.15 / 0.5) = 1 - 0.13 - 0.09 = 0.78. Based on this factor, an environmentally adaptive correction is made to the cumulative arc erosion. The correction formula is "Corrected cumulative amount = Original cumulative amount / Insulation recovery factor". Substituting the total cumulative arc erosion data mentioned above, we get: Corrected cumulative amount = 132.8 / 0.78 ≈ 170.3 mg. This correction ensures that the calculated arc erosion amount accurately reflects the actual environmental impact.

[0059] Based on the wear mode correction results, the main wear mechanisms and their development rates are determined. For example, the correction result output above indicates that fatigue spalling is the dominant wear mode, with a confidence level of 55.3%. The basic development rate of fatigue spalling is calibrated to be 0.02 mm / thousand operations through accelerated laboratory testing. The mechanism fit clearance change rate is introduced to dynamically correct the wear development rate. The mechanism fit clearance is measured by a laser displacement sensor, with an initial clearance of 0.1 mm and a current clearance of 0.15 mm. The cumulative number of operations is 1000, and the clearance change rate is calculated as (0.15-0.1) / 1000×1000=0.05 mm / thousand operations. Using the correction formula "corrected development rate = basic rate × (1 + clearance change rate / initial clearance)", substituting the data, we get: corrected development rate = 0.02×(1+0.05 / 0.1)=0.02×1.5=0.03 mm / thousand operations. The corrected development rate can match the wear rate with the actual fit state of the mechanism.

[0060] The assessment strategy is determined by combining the assessment level. For example, the assessment level mapped above is the balanced level, and the corresponding assessment strategy is to balance the accuracy of wear quantification and equipment protection. An operation frequency acceleration factor is calculated based on the statistical data of the operation frequency of the equipment. The operation frequency is statistically analyzed from the equipment operation log. The average monthly operation frequency over the past three months is 120 times, while the rated monthly operation frequency is 100 times. The acceleration factor = 120 / 100 = 1.2. Based on this acceleration factor, the parameters of the life model are optimized and adjusted. The life model adopts the basic structure of remaining life = (limited wear amount - current wear amount) / corrected development rate. Under the balanced level, the wear amount safety factor in the model is adjusted from the conservative level of 1.2 to 1.0, and the acceleration factor is incorporated to adapt the model parameters to the current operation frequency and assessment strategy.

[0061] The coupling acceleration effect coefficient between mechanical wear and arc erosion is calculated using the following formula: ; in, Represents the coupling coefficient. This indicates the cumulative amount of arc erosion after correction. Indicates the maximum erosion amount. This indicates the rate of wear development after correction. The coefficient represents the basic development rate, where the limit of corrosion for copper alloy contacts is 500 mg. Substituting the data, we get: Coupling coefficient = 0.2 × (170.3 / 500) + 0.3 × (0.03 / 0.02) = 0.2 × 0.341 + 0.3 × 1.5 ≈ 0.068 + 0.45 = 0.518. Based on this coupling coefficient, we identify mechanical and electrical combined fault modes and their risk levels. When the coupling coefficient > 0.5, it is determined to be a combined fault mode of fatigue spalling and arc erosion. The risk level is divided according to the coupling coefficient, with 0.5-0.7 being medium risk, corresponding to a fault development rate more than 50% faster than a single mode, requiring a shorter maintenance interval.

[0062] A multi-parameter fusion model for predicting remaining lifetime was constructed. The input parameters of the model included the corrected wear development rate of 0.03 mm / thousand cycles, the corrected cumulative arc erosion of 170.3 mg, the coupling coefficient of 0.518, and the operation frequency acceleration factor of 1.2. The model was constructed using a multiple linear regression algorithm, where the limit wear was set to 0.5 mm. The current wear = corrected development rate × thousands of operations = 0.03 × 1 = 0.03 mm. Substituting the data, we can obtain: remaining lifetime = (0.5 - 0.03) / (0.03 × 1.2) × (1 - 0.518) ≈ 0.47 / (0.036) × 0.482 ≈ 13.06 × 0.482 ≈ 6.3 thousand operations. An arc reignition risk threshold is introduced as a criterion for end-of-life determination. For copper alloy contacts, the arc reignition risk threshold is a contact gap ≥ 0.8 mm. The current gap is 0.15 mm. Based on the wear rate calculation, when the remaining lifespan is exhausted, the gap will be approximately 0.15 + 0.03 × 6.3 ≈ 0.339 mm, which does not reach the threshold, verifying the rationality of the lifespan prediction. Finally, a multi-dimensional wear status assessment report is generated, including quantitative wear indicators: current fatigue spalling wear (0.03 mm), cumulative arc erosion (170.3 mg), composite failure mode diagnosis (fatigue spalling + arc erosion), risk level assessment (medium risk), and remaining service life range (6,000-6,600 operations). This provides a complete basis for maintenance personnel to develop targeted maintenance plans.

[0063] like Figure 2 As shown, this application also discloses a wear condition assessment system for high-voltage power machinery components, comprising: Acquisition module 1 is used to acquire multi-sensor time-series signals, overshoot mechanical kinetic energy, and transient temperature of micro-area of ​​contact surface during the operation of high-voltage electric machinery components. Based on the overshoot mechanical kinetic energy, the multi-sensor time-series signals are compensated to obtain a kinetic energy normalized wear signal sequence. Fusion module 2 is used to extract time-domain wear features, frequency-domain resonance features, and time-frequency-domain surface morphology features based on the kinetic energy normalized wear signal sequence; obtain the material micro-toughness attenuation coefficient based on historical load cycle data and accumulated operation overshoot mechanical kinetic energy; and modulate the attention weights of multi-dimensional features based on the lightweight attention mechanism and the material micro-toughness attenuation coefficient to generate a comprehensive feature vector. Output module 3 is used to identify the current dominant wear mode type based on the comprehensive feature vector, calculate the thermomechanical fatigue coupling factor based on the transient temperature of the micro-area of ​​the contact surface, recalibrate the wear mode probability, and output the wear mode correction result after thermo-coupling correction. Evaluation module 4 is used to acquire real-time status data and dielectric property data of lubricating medium, calculate instantaneous health coefficient and lubricating film failure risk factor, fuse them to obtain working condition adaptive evaluation aggressive coefficient, and map the evaluation aggressive coefficient to evaluation level; The generation module 5 is used to obtain the cumulative amount of arc erosion based on the historical breaking current sequence and the contact material ablation rate, and to generate a multi-dimensional wear status assessment report based on the wear mode correction results, assessment level and cumulative amount of arc erosion, which includes quantitative indicators of each wear mode, composite fault mode and remaining service life range.

[0064] In one embodiment, the output module includes: The input unit is used to input the comprehensive feature vector into the pre-trained multi-classification model to obtain the initial probability distribution of each wear mode; The calculation unit is used to calculate the thermomechanical fatigue coupling factor based on the transient temperature data of the micro-area of ​​the contact surface, combined with the material's thermal expansion coefficient and thermal conductivity characteristics. A calibration unit is used to recalibrate the initial probability distribution based on the thermomechanical fatigue coupling factor, thereby enhancing the weight of wear modes dominated by thermal stress. The processing unit is used to normalize the probability distribution after recalibration and determine the current dominant wear mode type and its confidence level. An output unit is provided for outputting the wear mode correction results based on the fusion of thermo-coupling effects. This application also provides a computer device, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps of the aforementioned method for assessing the wear condition of high-voltage power machinery components.

[0065] like Figure 3 As shown, this application also provides a computer device, which can be a server, and its internal structure can be as follows: Figure 3 As shown, the computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores all data required for the process of a method for assessing the wear condition of high-voltage power machinery components. The network interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a method for assessing the wear condition of high-voltage power machinery components.

[0066] Those skilled in the art will understand that Figure 3 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer equipment on which the present application is applied.

[0067] This application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described method for assessing the wear condition of high-voltage power machinery components.

[0068] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. 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 embodiments of the above methods. Any references to memory, storage, databases, or other media used in this application and in the embodiments can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual-speed SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0069] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, apparatus, article, or method that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, apparatus, article, or method. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, apparatus, article, or method that includes that element.

[0070] The above description is merely a preferred embodiment of this application and does not limit the scope of this application. Any equivalent results or equivalent process transformations made based on the content of this application specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the scope of protection of this application.

Claims

1. A method for evaluating a wear state of a high-voltage power machine component, characterized by, The method comprises the following steps: Collecting multi-sensor time sequence signals, operating overshoot mechanical kinetic energy and contact surface micro-zone transient temperature during operation of a high-voltage power mechanical component, performing compensation processing on the multi-sensor time sequence signals based on the operating overshoot mechanical kinetic energy, and obtaining a kinetic energy normalized wear signal sequence; Extracting time domain wear characteristics, frequency domain resonance characteristics and time-frequency domain surface morphology characteristics from the kinetic energy normalized wear signal sequence, obtaining a material micro-ductility attenuation coefficient based on historical load cycle data and cumulative operating overshoot mechanical kinetic energy, modulating attention weights of multi-dimensional features based on a lightweight attention mechanism and according to the material micro-ductility attenuation coefficient, and fusing to generate a comprehensive feature vector; Identifying a current dominant wear mode type according to the comprehensive feature vector, calculating a thermal mechanical fatigue coupling factor based on the contact surface micro-zone transient temperature, recalibrating a wear mode probability, and outputting a wear mode correction result corrected by thermal coupling; Obtaining real-time state data and lubricating medium dielectric property data, calculating an instantaneous health coefficient and a lubricating film failure risk factor, fusing to obtain a working condition adaptive evaluation aggressiveness coefficient, and mapping the evaluation aggressiveness coefficient to an evaluation grade; Obtaining an arc erosion cumulative amount based on a historical breaking current sequence and a contact material ablation rate, and generating a multi-dimensional wear state evaluation report containing quantitative indexes of each wear mode, a composite failure mode and a residual service life interval based on the wear mode correction result, the evaluation grade and the arc erosion cumulative amount.

2. The high-voltage power machine component wear state assessment method of claim 1, wherein, The step of performing compensation processing on the multi-sensor time sequence signals based on the operating overshoot mechanical kinetic energy and obtaining a kinetic energy normalized wear signal sequence comprises: Obtaining a mechanism motion smoothness factor, and performing motion artifact elimination processing on an original signal according to the mechanism motion smoothness factor; Obtaining an electromagnetic interference feature spectrum, calculating an electromagnetic interference signal-to-noise ratio based on the electromagnetic interference feature spectrum, and performing electromagnetic interference filtering on the signal after the motion artifact elimination according to the electromagnetic interference signal-to-noise ratio; Performing wavelet threshold denoising processing on the filtered signal, decomposing the signal into multiple frequency bands through discrete wavelet transform, and removing noise components by applying a threshold function based on signal sparsity adaptive adjustment; Calculating a kinetic energy compensation coefficient based on the operating overshoot mechanical kinetic energy, modifying the compensation coefficient according to contact surface material hardness characteristics, and performing kinetic energy normalization compensation on the denoised signal; Calculating a multi-source signal coherence index, fusing and enhancing the compensated signal based on the coherence index, intercepting a signal sequence of a key operation stage, and forming a kinetic energy normalized wear signal sequence.

3. The high-voltage power machine component wear state assessment method of claim 1, wherein, The step of modulating attention weights of multi-dimensional features based on a lightweight attention mechanism and according to the material micro-ductility attenuation coefficient, and fusing to generate a comprehensive feature vector comprises: Respectively extracting time domain statistical features, frequency domain resonance peak features and time-frequency domain wavelet packet energy features from the kinetic energy normalized wear signal sequence; Obtaining historical load cycle times and cumulative operating overshoot mechanical kinetic energy based on historical data of the mechanical component throughout its life cycle, and calculating a material micro-ductility attenuation coefficient through a material fatigue damage model; Constructing a multi-branch lightweight attention network, and each branch processes wear features of different dimensions; The material micro-ductility attenuation coefficient is input into the attention network as prior knowledge to dynamically adjust the attention weight distribution of each feature dimension and generate a comprehensive feature vector representing material property degradation.

4. The high-voltage power machine component wear state assessment method of claim 1, wherein, The step of recalibrating the wear mode probability and outputting the wear mode correction result after thermal coupling correction comprises: The comprehensive feature vector is input into a pre-trained multi-classification model to obtain an initial probability distribution of each wear mode; Based on the micro-area transient temperature data of the contact surface, the thermal mechanical fatigue coupling factor is calculated by combining the material thermal expansion coefficient and the thermal conductivity characteristics; The initial probability distribution is recalibrated according to the thermal mechanical fatigue coupling factor to enhance the weight of the wear mode dominated by thermal stress; The recalibrated probability distribution is normalized to determine the current dominant wear mode type and its confidence level; The wear mode correction result is output after fusing the thermal coupling effect.

5. The high-voltage power machine component wear state assessment method of claim 1, wherein, The step of obtaining real-time state data and dielectric property data of the lubricating medium, calculating the instantaneous health coefficient and the lubricating film failure risk factor, fusing to obtain the working condition adaptive evaluation aggressiveness coefficient, and mapping the evaluation aggressiveness coefficient to the evaluation level comprises: Real-time running state data is obtained from the equipment monitoring system, including running time, load current and mechanism stroke characteristics; The dielectric constant and dielectric loss factor of the lubricating medium are obtained, and the lubricating film failure risk factor is calculated based on the dielectric property change trend; The instantaneous health coefficient is calculated based on the real-time running state data; The working condition adaptive evaluation aggressiveness coefficient is obtained according to the instantaneous health coefficient and the lubricating film failure risk factor; Based on the preset evaluation level threshold, the evaluation aggressiveness coefficient is mapped to the conservative, balanced and aggressive evaluation levels.

6. The high-voltage power machine component wear state assessment method of claim 1, wherein, The step of generating a multi-dimensional wear state evaluation report containing quantitative indicators of each wear mode, composite failure modes and residual service life intervals based on the wear mode correction result, evaluation level and arc erosion cumulative amount comprises: A historical breaking current sequence is obtained, the current level weight coefficient is calculated based on the differential influence of different current levels on contact erosion, and the arc erosion cumulative amount is calculated by combining the contact material ablation rate model and the current level weight coefficient; Environmental humidity and pollutant concentration data are obtained, the environmental medium insulation recovery factor is calculated based on the environmental medium insulation recovery characteristics, and the arc erosion cumulative amount is adaptively corrected according to the environmental medium insulation recovery factor; Based on the wear mode correction result, the main wear mechanism and its development rate are determined, and the mechanism gap change rate is introduced to dynamically correct the wear development rate; The evaluation strategy is determined based on the evaluation level, the operation frequency acceleration factor is calculated based on the mechanism operation frequency statistics, and the life model parameters are optimized and adjusted according to the operation frequency acceleration factor; The coupling acceleration effect coefficient of mechanical wear and arc erosion is calculated, and the mechanical and electrical composite failure mode and its risk level are identified based on the coupling acceleration effect coefficient; A multi-parameter fused residual life prediction model is constructed, the arc reignition risk threshold is introduced as the life termination criterion, and a multi-dimensional wear state evaluation report containing wear quantitative indicators, composite failure mode diagnosis, risk level evaluation and residual service life interval is generated.

7. A high voltage power machine component wear state assessment system, characterized by, It comprises: The collection module is configured to collect multi-sensor time sequence signals, operating overshoot mechanical energy, and micro-area transient temperature of a contact surface during operation of a high-voltage power mechanical component, compensate the multi-sensor time sequence signals based on the overshoot mechanical energy, and obtain a kinetic energy normalized wear signal sequence; The fusion module is configured to extract time domain wear features, frequency domain resonance features, and time-frequency domain surface morphology features from the kinetic energy normalized wear signal sequence, obtain a material micro-elasticity attenuation coefficient based on historical load cycle data and accumulated operating overshoot mechanical energy, modulate attention weights of multi-dimensional features based on a lightweight attention mechanism and the material micro-elasticity attenuation coefficient, and generate a comprehensive feature vector through fusion; The output module is configured to identify a current dominant wear mode type based on the comprehensive feature vector, calculate a thermal-mechanical fatigue coupling factor based on the micro-area transient temperature of the contact surface, recalibrate a wear mode probability, and output a wear mode correction result after thermal coupling correction; The evaluation module is configured to obtain real-time state data and dielectric property data of a lubricating medium, calculate an instantaneous health coefficient and a lubricating film failure risk factor, obtain a working condition adaptive evaluation aggressiveness coefficient through fusion, and map the evaluation aggressiveness coefficient to an evaluation level; The generation module is configured to obtain an arc erosion cumulative amount based on a historical breaking current sequence and a contact material ablation rate, and generate a multi-dimensional wear state evaluation report including quantitative indicators of each wear mode, a composite failure mode, and a residual service life interval based on the wear mode correction result, the evaluation level, and the arc erosion cumulative amount.

8. The high-voltage power machine component wear state assessment system of claim 7, wherein, The output module includes: An input unit configured to input the comprehensive feature vector into a pre-trained multi-classification model to obtain an initial probability distribution of each wear mode; A calculation unit configured to calculate a thermal-mechanical fatigue coupling factor based on the micro-area transient temperature data of the contact surface, in combination with a material thermal expansion coefficient and thermal conduction characteristics; A calibration unit configured to recalibrate the initial probability distribution based on the thermal-mechanical fatigue coupling factor to enhance the weight of a wear mode dominated by thermal stress; A processing unit configured to normalize the recalibrated probability distribution to determine a current dominant wear mode type and a confidence thereof; An output unit configured to output a wear mode correction result after fusion of thermal coupling effects. 9.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is configured to perform the method according to any one of claims 1-8 when the computer program is executed by the processor. The processor executes the computer program to implement the steps of the method of any one of claims 1 to 6.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the method of any one of claims 1 to 6.

Images