Multi-sensor fusion intelligent actuator fault self-diagnosis method and system

Abstract

The invention relates to a multi-sensor fusion intelligent actuator fault self-diagnosis method and system, and belongs to the technical field of actuator fault diagnosis. The method comprises the following steps: realizing multi-sensor clock synchronization through a unified clock source, generating a synchronous time sequence for asynchronous data such as Hall, pressure and temperature by adopting an interpolation method, and intensively acquiring signals such as rotating speed, torque and vibration; based on a current sensor and temperature data, primary fault identification is carried out through a current overload protection model, and an adaptive processing mechanism is triggered; demodulating the noise-reduced vibration signal, extracting high-frequency energy and detecting the deviation degree between the fault frequency of the bearing and a base line; bearing energy consumption abnormity is detected, and a fusion weight is dynamically set by combining the deviation degree, the energy consumption coefficient and an overload result; and carrying out weighted fusion on the multi-modal features, inputting an intelligent diagnosis model to carry out fault mode identification, and finally outputting a fault type / position and triggering a processing strategy. Accurate and rapid self-diagnosis of the actuator fault is realized.

Description

Technical Field

[0001] The present invention belongs to the technical field of actuator fault diagnosis, and in particular relates to a multi-sensor fusion intelligent actuator fault self-diagnosis method and system. Background Art

[0002] With the continuous advancement of industrial automation, actuators, as key components in automation systems, are crucial for their reliability and stability to ensure the proper functioning of the entire system. However, in practical applications, actuators often fail due to various factors, leading to reduced production efficiency and even system downtime. Traditional fault diagnosis methods often rely on manual experience and periodic testing, which is not only inefficient but also difficult to accurately determine the fault type and location. Therefore, it is particularly important to develop a method and system that can automatically and accurately diagnose actuator faults.

[0003] Existing technologies have made progress in multi-sensor synchronization, fault feature extraction, and multi-modal fusion, but the following core issues still exist: Insufficient adaptability to dynamic environments: Synchronization accuracy is limited by hardware, and the fault model has poor robustness under changing working conditions.

[0004] Feature extraction and noise reduction bottlenecks: Low-frequency fault signals are easily masked by noise, and the demodulation algorithm relies on manual experience.

[0005] The multimodal fusion strategy is rigid: the weight distribution lacks dynamic optimization and the cross-modal data alignment is highly complex.

[0006] Intelligent diagnosis has weak generalization capabilities: small samples and data imbalance restrict model performance, and online diagnosis lacks real-time performance. Summary of the Invention

[0007] In order to solve the above problems existing in the prior art, the present invention provides a multi-sensor fusion intelligent actuator fault self-diagnosis method and system. The purpose of the present invention can be achieved through the following technical solutions: A multi-sensor fusion intelligent actuator fault self-diagnosis method: S1: Synchronizes the clocks of multiple sensors, uses Hall sensors to provide real-time feedback on motor speed and position information, uses pressure sensors to dynamically monitor output shaft torque, uses temperature sensors to detect motor winding and controller temperatures, uses current sensors to monitor motor current in real time, and uses vibration sensors to monitor mechanical vibration signals in real time. Sensor data is centrally collected, and asynchronous data at different sampling rates from multiple sensors is interpolated to generate a synchronized time series. S2: extracting motor current data and motor winding and controller temperatures from the sensor data based on the synchronized time series, and performing preliminary fault identification using a current overload protection model; triggering a corresponding adaptive processing mechanism in a self-diagnosis linkage unit based on the result of the preliminary fault identification; S3: demodulating the amplitude-frequency modulation signal of the vibration sensor after noise reduction in the sensor data using the bearing fault early warning model, extracting the high-frequency energy of the demodulated signal and detecting the deviation between the amplitude of the characteristic frequency of the rolling bearing fault and a preset baseline; S4: Detecting the energy loss coefficient caused by bearing abnormality through correlation analysis of the torque signal of the pressure sensor and the speed fluctuation signal of the Hall sensor; constructing a unified spatiotemporal coordinate system based on the synchronized time series, aligning the characteristic delays of different modal data, and setting a fusion weight based on the deviation, energy loss coefficient, and the overload identification result output by the current overload protection model; S5: Perform weighted fusion on the features of different modal data according to the fusion weight to generate a comprehensive fault feature vector; input the comprehensive fault feature vector into the intelligent diagnosis model, and the intelligent diagnosis model performs fault mode recognition and outputs fault type and fault location information; trigger a fault handling strategy based on the fault type and fault location information.

[0008] Specifically, the current overload protection model adopts a two-level heat accumulation method. The first level calculates the instantaneous temperature rise of the winding based on the current square time integral principle, and predicts the thermal equilibrium state in combination with the initial thermal state of the motor and the cooling time constant. The calculation formula is: , Among them, K is the negative sequence current heating coefficient, T is the overload time constant, A0 is the initial temperature, t is the current time, I rms is the effective value of the current at time τ, θ is the estimated temperature of the winding over time, and τ is the integral variable; The second level introduces adaptive threshold adjustment, which fuses current sensor data and temperature sensor data through the Kalman filter and dynamically corrects the overload trigger threshold according to ambient temperature and load fluctuations.

[0009] Specifically, the self-diagnosis linkage unit constructs a fault level-response action mapping table, and generates a comprehensive risk index based on the fault confidence output by the current overload protection model and the bearing fault feature deviation: , Among them, R is the comprehensive risk index, which is used to evaluate the overall risk level of the system; α is the weight coefficient of the current overload protection model, which is used to adjust the influence of current overload on the comprehensive risk index; β is the weight coefficient of the bearing fault characteristic deviation, which is used to adjust the influence of the bearing fault characteristic on the comprehensive risk index; Iover is the actual current overload value, indicating the extent to which the current in the system exceeds the normal range, I th is the current threshold, which indicates the maximum normal current value allowed by the system. F is the bearing fault characteristic deviation, which indicates the difference between the current bearing state and the normal state. base is the baseline fault characteristic value, which represents the characteristic value of the bearing under normal conditions.

[0010] Specifically, the adaptive processing mechanism is a three-level response mode, including predictive load reduction, active thermal balancing, and hardware protection; The predictive load reduction dynamically adjusts the target torque of the pressure sensor through a PID algorithm based on the estimated temperature of the winding over time. The active thermal balance triggers the variable frequency operation of the cooling system, and optimizes the cooling fan speed distribution in real time through finite element simulation based on the temperature gradient distribution model. The hardware protection is used to directly cut off the IGBT drive signal when the high-frequency energy of the vibration signal and the current harmonic distortion rate exceed the specified threshold.

[0011] Specifically, the bearing fault warning model uses improved variational mode decomposition for noise reduction, optimizes variational mode decomposition parameters using an energy difference grid search method to suppress endpoint effects, and uses a support vector regression machine to correct and reconstruct the signal. The amplitude-frequency modulated signal is then demodulated using Hilbert envelope demodulation to extract the modulated high-frequency resonance components.

[0012] Specifically, the deviation fuzzy comprehensive evaluation method quantifies the degree to which the characteristic frequency amplitude deviates from a preset baseline.

[0013] Specifically, the collaborative allocation method of fusion weight information entropy-time decay quantifies the deviation, energy loss coefficient, and overload identification result feature credibility, and sets the feature effectiveness coefficient according to the time decay factor; combines the modal contribution, the feature effectiveness coefficient and the quantized information entropy weight to generate the final fusion weight.

[0014] Specifically, the intelligent diagnosis model receives the comprehensive fault feature vector and processes the heterogeneous data of vibration, current, temperature and torque through parallel channels. Among them, the vibration signal uses one-dimensional depth-wise separable convolution to extract multi-scale impact features, the current and temperature data capture temporal dependencies through bidirectional gated recurrent units, and the torque signal models spatial correlation through a self-attention mechanism.

[0015] Specifically, the training process of the intelligent diagnosis model adopts a collaborative optimization strategy of transfer learning and adversarial domain adaptation.

[0016] A multi-sensor fusion intelligent actuator fault self-diagnosis system, including a multi-sensor data synchronization acquisition module, an overload initial judgment and adaptive response module, a bearing fault feature demodulation module, a multi-modal feature fusion weight module, and an intelligent diagnosis and strategy triggering module; The multi-sensor data synchronization acquisition module synchronizes the clocks of multiple sensors, uses Hall sensors to provide real-time feedback on motor speed and position information, uses pressure sensors to dynamically monitor output shaft torque, uses temperature sensors to detect motor winding and controller temperatures, uses current sensors to monitor motor current in real time, uses vibration sensors to monitor mechanical vibration signals in real time, centrally collects sensor data, and generates a synchronized time series by interpolating asynchronous data at different sampling rates from multiple sensors. The overload preliminary judgment and adaptive response module extracts motor current data and motor winding and controller temperature from the sensor data based on the synchronous time series and performs preliminary fault identification through the current overload protection model; based on the result of the preliminary fault identification, triggers the corresponding adaptive processing mechanism in the self-diagnosis linkage unit; The bearing fault feature demodulation module demodulates the amplitude-frequency modulation signal of the vibration sensor after noise reduction in the sensor data through the bearing fault early warning model, extracts the high-frequency energy of the demodulated signal and detects the deviation of the rolling bearing fault feature frequency amplitude from the preset baseline; The multimodal feature fusion weight module detects the energy loss coefficient caused by bearing abnormality by performing correlation analysis on the torque signal of the pressure sensor and the speed fluctuation signal of the Hall sensor; constructs a unified spatiotemporal coordinate system based on the synchronized time series, aligns the feature delays of different modal data, and sets the fusion weight based on the deviation, energy loss coefficient, and the overload identification result output by the current overload protection model; The intelligent diagnosis and strategy triggering module performs weighted fusion of different modal data features according to the fusion weight to generate a comprehensive fault feature vector; the comprehensive fault feature vector is input into the intelligent diagnosis model, and the intelligent diagnosis model performs fault mode recognition and outputs fault type and fault location information; and the fault handling strategy is triggered according to the fault type and fault location information.

[0017] The beneficial effects of the present invention are: The multi-sensor fusion intelligent actuator fault self-diagnosis method and system provided by the present invention can realize comprehensive, accurate and timely diagnosis of actuator faults. First, the multi-sensor data synchronization acquisition module ensures the synchronization and accuracy of the data, providing a reliable basis for subsequent fault identification. The overload initial judgment and adaptive response module can quickly identify preliminary faults such as current overload, and trigger the corresponding adaptive processing mechanism according to the fault level, effectively avoiding further deterioration of the fault. The bearing fault feature demodulation module focuses on the fault feature extraction of rolling bearings, and can accurately capture the fault signal of the bearing through noise reduction processing and demodulation technology. The multimodal feature fusion weight module cleverly fuses the data features of different modes, and improves the accuracy and robustness of fault identification through scientifically set fusion weights. Finally, the intelligent diagnosis and strategy triggering module uses advanced intelligent diagnosis models to identify fault patterns and output detailed fault type and location information, providing valuable reference for maintenance personnel. BRIEF DESCRIPTION OF THE DRAWINGS

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

[0019] Figure 1 Schematic diagram of a flow chart of a multi-sensor fusion intelligent actuator fault self-diagnosis method of the present invention; Figure 2 This is a timing diagram of data collection from each sensor in the fault diagnosis process. DETAILED DESCRIPTION

[0020] In order to further illustrate the technical means and effects adopted by the present invention to achieve the predetermined purpose of the invention, the specific implementation methods, structures, features and effects of the present invention are described in detail below in conjunction with the accompanying drawings and preferred embodiments.

[0021] See also Figure 1 , a multi-sensor fusion intelligent actuator fault self-diagnosis method: S1: Synchronizes the clocks of multiple sensors, uses Hall sensors to provide real-time feedback on motor speed and position information, uses pressure sensors to dynamically monitor output shaft torque, uses temperature sensors to detect motor winding and controller temperatures, uses current sensors to monitor motor current in real time, and uses vibration sensors to monitor mechanical vibration signals in real time. Sensor data is centrally collected, and asynchronous data at different sampling rates from multiple sensors is interpolated to generate a synchronized time series. S2: extracting motor current data and motor winding and controller temperatures from the sensor data based on the synchronized time series, and performing preliminary fault identification using a current overload protection model; triggering a corresponding adaptive processing mechanism in a self-diagnosis linkage unit based on the result of the preliminary fault identification; S3: demodulating the amplitude-frequency modulation signal of the vibration sensor after noise reduction in the sensor data using the bearing fault early warning model, extracting the high-frequency energy of the demodulated signal and detecting the deviation between the amplitude of the characteristic frequency of the rolling bearing fault and a preset baseline; S4: Detecting the energy loss coefficient caused by bearing abnormality through correlation analysis of the torque signal of the pressure sensor and the speed fluctuation signal of the Hall sensor; constructing a unified spatiotemporal coordinate system based on the synchronized time series, aligning the characteristic delays of different modal data, and setting a fusion weight based on the deviation, energy loss coefficient, and the overload identification result output by the current overload protection model; S5: Perform weighted fusion on the features of different modal data according to the fusion weight to generate a comprehensive fault feature vector; input the comprehensive fault feature vector into the intelligent diagnosis model, and the intelligent diagnosis model performs fault mode recognition and outputs fault type and fault location information; trigger a fault handling strategy based on the fault type and fault location information.

[0022] Specifically, the current overload protection model adopts a two-level heat accumulation method. The first level calculates the instantaneous temperature rise of the winding based on the current square time integral principle, and predicts the thermal equilibrium state in combination with the initial thermal state of the motor and the cooling time constant. The calculation formula is: , Among them, K is the negative sequence current heating coefficient, T is the overload time constant, A0 is the initial temperature, t is the current time, I rms is the effective value of the current at time τ, θ is the estimated temperature of the winding over time, and τ is the integral variable; The second level introduces adaptive threshold adjustment, which fuses current sensor data and temperature sensor data through the Kalman filter and dynamically corrects the overload trigger threshold according to ambient temperature and load fluctuations.

[0023] Specifically, the self-diagnosis linkage unit constructs a fault level-response action mapping table, and generates a comprehensive risk index based on the fault confidence output by the current overload protection model and the bearing fault feature deviation: , Among them, R is the comprehensive risk index, which is used to evaluate the overall risk level of the system; α is the weight coefficient of the current overload protection model, which is used to adjust the influence of current overload on the comprehensive risk index; β is the weight coefficient of the bearing fault characteristic deviation, which is used to adjust the influence of the bearing fault characteristic on the comprehensive risk index; I over is the actual current overload value, indicating the extent to which the current in the system exceeds the normal range, I th is the current threshold, which indicates the maximum normal current value allowed by the system. F is the bearing fault characteristic deviation, which indicates the difference between the current bearing state and the normal state.base is the baseline fault characteristic value, which represents the characteristic value of the bearing under normal conditions.

[0024] Specifically, the adaptive processing mechanism is a three-level response mode, including predictive load reduction, active thermal balancing, and hardware protection; The predictive load reduction dynamically adjusts the target torque of the pressure sensor through a PID algorithm based on the estimated temperature of the winding over time. The active thermal balance triggers the variable frequency operation of the cooling system, and optimizes the cooling fan speed distribution in real time through finite element simulation based on the temperature gradient distribution model. The hardware protection is used to directly cut off the IGBT drive signal when the high-frequency energy of the vibration signal and the current harmonic distortion rate exceed the specified threshold.

[0025] Specifically, the bearing fault warning model uses improved variational mode decomposition for noise reduction, optimizes variational mode decomposition parameters using an energy difference grid search method to suppress endpoint effects, and uses a support vector regression machine to correct and reconstruct the signal. The amplitude-frequency modulated signal is then demodulated using Hilbert envelope demodulation to extract the modulated high-frequency resonance components.

[0026] In this embodiment, the vibration signal is multiplied by the local carrier, and the high-frequency components are filtered out to obtain the envelope signal. Wavelet packet decomposition is used to obtain the energy distribution in different frequency bands and locate the impulse response frequency band. The demodulated signal is transformed by FFT to highlight the amplitude information of the characteristic frequencies of the bearing fault (such as the outer race / inner race fault frequency). High-frequency energy calculation: Multi-scale weighted permutation entropy (MWPE) is used to quantify the signal randomness, combined with the improved variational mode decomposition (VMD) reconstruction capability to capture the high-frequency energy mutation after the fault onset. The formula is:

[0027] Among them, E HF is the high frequency energy, f c is the cutoff frequency, X(k) is the frequency amplitude; Eigenfrequency identification is based on bearing geometric parameters (number of balls, pitch diameter, contact angle, etc.) and calculates the theoretical eigenfrequency: , , Among them, f o is the characteristic frequency of the outer ring, f i is the characteristic frequency of the inner ring, f r is the rotation frequency related to the mechanical rotation, n is the number of balls, d is the ball diameter, D is the average diameter of the bearing raceway, and α is the contact angle; Analyze the actual signal by power spectral density (PSD), match the theoretical frequency and record the amplitude A actual .

[0028] Specifically, the deviation fuzzy comprehensive evaluation method quantifies the degree to which the characteristic frequency amplitude deviates from a preset baseline.

[0029] In this embodiment, the mean μ of the characteristic frequency amplitude is calculated in a healthy state. base and standard deviation σ base , set the baseline interval [μ base -3σ base , μ base +3σ base ], use the Gaussian function to convert the actual amplitude A actual , mapped to the baseline interval, generating distance membership μ d and peak membership μ p ; Dynamically adjust the weights of the distance and peak evaluation matrix through the coefficient of variation method to calculate the comprehensive deviation: .

[0030] Specifically, the collaborative allocation method of fusion weight information entropy-time decay quantifies the deviation, energy loss coefficient, and overload identification result feature credibility, and sets the feature effectiveness coefficient according to the time decay factor; combines the modal contribution, the feature effectiveness coefficient and the quantized information entropy weight to generate the final fusion weight.

[0031] In this embodiment, the information entropy weights are calculated for the deviation (F), energy loss coefficient (C), and overload identification result (I): , Among them, H j is the information entropy of the j-th feature, p ij is the probability of eigenvalue distribution within the sliding window; Set the feature validity coefficient: , Where λ is the time decay factor and t is the feature generation time delay; Construct the Gram angular field mapping space-time correlation matrix, extract the main coupling mode through singular value decomposition (SVD), and calculate the modal contribution: , Among them, σ is a singular value, u is a left singular vector; Generate the final fusion weight based on two-layer fuzzy inference: , The partial derivative term represents the sensitivity of the comprehensive risk index R to the feature xj; The final fusion weight is W=[W F ,W C ,W I ].

[0032] Specifically, the intelligent diagnosis model receives the comprehensive fault feature vector and processes the heterogeneous data of vibration, current, temperature and torque through parallel channels. Among them, the vibration signal uses one-dimensional depth-wise separable convolution to extract multi-scale impact features, the current and temperature data capture temporal dependencies through bidirectional gated recurrent units, and the torque signal models spatial correlation through a self-attention mechanism.

[0033] In this embodiment, the intelligent diagnosis model adopts a hybrid architecture of a multimodal residual attention network and a dynamic knowledge graph, specifically including: 1. Multimodal feature fusion layer: Receives the comprehensive fault feature vector from the S5 and processes the heterogeneous data of vibration, current, temperature, and torque separately through parallel channels. A one-dimensional depthwise separable convolution (CNN) is used to extract multi-scale impact features for vibration signals. A bidirectional gated recurrent unit (BiGRU) is used to capture temporal dependencies in current and temperature data. A self-attention mechanism is used to model spatial correlation in torque signals. 2. Residual Cross-Attention Module: Projects the high-dimensional features of each modality into a unified semantic space and calculates the weights using the cross-modal attention matrix. The formula is: , Among them, Q is the query vector (current modality feature), K and V are key-value pairs (other modality features), and is the dimension scaling factor; 3. Dynamic Knowledge Reasoning Engine: This engine integrates the equipment failure knowledge graph, where nodes represent entities such as failure modes, component associations, and maintenance strategies, while edges represent fault transmission paths and conditional probabilities. It uses a graph neural network (GNN) for real-time subgraph retrieval. When the deviation of a bearing's characteristic frequency exceeds a threshold, it activates the reasoning path from "bearing wear → abnormal vibration → torque fluctuation." 4. Online incremental learning mechanism: Deploy the Elastic Weight Consolidation algorithm. When new failure modes emerge, the model parameters are dynamically updated based on historical diagnosis results and maintenance records, preserving important weights while adapting to the new data distribution.

[0034] Specifically, the training process of the intelligent diagnosis model adopts a collaborative optimization strategy of transfer learning and adversarial domain adaptation.

[0035] In this embodiment, the intelligent diagnosis model pre-training stage uses a masked autoencoder to reconstruct multimodal signals on public fault datasets (such as the CWRU bearing dataset and the MIT motor current library) and learn universal fault representations. In the domain adaptation stage, a gradient reversal layer (GRL) is introduced to minimize the difference in feature distribution between the source domain (simulated data) and the target domain (real device data). The objective function is: , in, is the fault classification loss, is the domain classification loss, and λ is the adaptive coefficient.

[0036] A multi-sensor fusion intelligent actuator fault self-diagnosis system, including a multi-sensor data synchronization acquisition module, an overload initial judgment and adaptive response module, a bearing fault feature demodulation module, a multi-modal feature fusion weight module, and an intelligent diagnosis and strategy triggering module; The multi-sensor data synchronization acquisition module synchronizes the clocks of multiple sensors, uses Hall sensors to provide real-time feedback on motor speed and position information, uses pressure sensors to dynamically monitor output shaft torque, uses temperature sensors to detect motor winding and controller temperatures, uses current sensors to monitor motor current in real time, uses vibration sensors to monitor mechanical vibration signals in real time, centrally collects sensor data, and generates a synchronized time series by interpolating asynchronous data at different sampling rates from multiple sensors. The overload preliminary judgment and adaptive response module extracts motor current data and motor winding and controller temperature from the sensor data based on the synchronous time series and performs preliminary fault identification through the current overload protection model; based on the result of the preliminary fault identification, triggers the corresponding adaptive processing mechanism in the self-diagnosis linkage unit; The bearing fault feature demodulation module demodulates the amplitude-frequency modulation signal of the vibration sensor after noise reduction in the sensor data through the bearing fault early warning model, extracts the high-frequency energy of the demodulated signal and detects the deviation of the rolling bearing fault feature frequency amplitude from the preset baseline; The multimodal feature fusion weight module detects the energy loss coefficient caused by bearing abnormality by performing correlation analysis on the torque signal of the pressure sensor and the speed fluctuation signal of the Hall sensor; constructs a unified spatiotemporal coordinate system based on the synchronized time series, aligns the feature delays of different modal data, and sets the fusion weight based on the deviation, energy loss coefficient, and the overload identification result output by the current overload protection model; The intelligent diagnosis and strategy triggering module performs weighted fusion of different modal data features according to the fusion weight to generate a comprehensive fault feature vector; the comprehensive fault feature vector is input into the intelligent diagnosis model, and the intelligent diagnosis model performs fault mode recognition and outputs fault type and fault location information; and the fault handling strategy is triggered according to the fault type and fault location information.

[0037] The computer storage medium of the embodiment of the present invention may adopt any combination of one or more computer-readable media. The computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or component, or any combination thereof. More specific examples (a non-exhaustive list) of computer-readable storage media include: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In this document, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, device or device.

[0038] A computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, which carries computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium that can transmit, propagate, or transport a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0039] The program code included in the computer-readable medium can be transmitted with any appropriate medium, including but not limited to wireless, electric wire, optical cable, RF or the like, or any suitable combination thereof. The computer program code for performing the operation of the present invention can be written in one or more programming languages ​​or a combination thereof, and the programming language includes an object-oriented programming language such as Java, Smalltalk, C++, and also includes a conventional procedural programming language such as "C" language or similar programming language. The program code can be executed completely on the user's computer, partially on the user's computer, as an independent software package, partially on the user's computer and partially on a remote computer, or completely on a remote computer or server. In the case of a remote computer, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computer (for example, utilizing an Internet service provider to connect through the Internet).

[0040] The above description is merely a preferred embodiment of the present invention and does not constitute any form of limitation to the present invention. Although the present invention has been disclosed as a preferred embodiment as above, it is not intended to limit the present invention. Any person skilled in the art can make some changes or modifications to equivalent embodiments using the technical contents disclosed above without departing from the scope of the technical solution of the present invention. However, any simple modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention are still within the scope of the technical solution of the present invention.

Claims

1. A multi-sensor fusion intelligent actuator fault self-diagnosis method, characterized in that: include: S1: Synchronizes the clocks of multiple sensors, uses Hall sensors to provide real-time feedback on motor speed and position information, uses pressure sensors to dynamically monitor output shaft torque, uses temperature sensors to detect motor winding and controller temperatures, uses current sensors to monitor motor current in real time, and uses vibration sensors to monitor mechanical vibration signals in real time. Sensor data is centrally collected, and asynchronous data at different sampling rates from multiple sensors is interpolated to generate a synchronized time series. S2: extracting motor current data and motor winding and controller temperatures from the sensor data based on the synchronized time series, and performing preliminary fault identification using a current overload protection model; triggering a corresponding adaptive processing mechanism in a self-diagnosis linkage unit based on the result of the preliminary fault identification; S3: demodulating the amplitude-frequency modulation signal of the vibration sensor after noise reduction in the sensor data using the bearing fault early warning model, extracting the high-frequency energy of the demodulated signal and detecting the deviation between the amplitude of the characteristic frequency of the rolling bearing fault and a preset baseline; S4: Detecting the energy loss coefficient caused by bearing abnormality through correlation analysis of the torque signal of the pressure sensor and the speed fluctuation signal of the Hall sensor; constructing a unified spatiotemporal coordinate system based on the synchronized time series, aligning the characteristic delays of different modal data, and setting a fusion weight based on the deviation, energy loss coefficient, and the overload identification result output by the current overload protection model; S5: performing weighted fusion on the features of different modal data according to the fusion weight to generate a comprehensive fault feature vector; The comprehensive fault feature vector is input into an intelligent diagnosis model, and the intelligent diagnosis model performs fault pattern recognition and outputs fault type and fault location information; and a fault handling strategy is triggered according to the fault type and fault location information.

2. The method according to claim 1, characterized in that The current overload protection model adopts a two-level heat accumulation method. The first level calculates the instantaneous temperature rise of the winding based on the current square time integral principle, and predicts the thermal equilibrium state by combining the initial thermal state of the motor and the cooling time constant. The calculation formula is: ， Among them, K is the negative sequence current heating coefficient, T is the overload time constant, A0 is the initial temperature, t is the current time, I rms is the effective value of the current at time τ, θ is the estimated temperature of the winding over time, and τ is the integral variable; The second level introduces adaptive threshold adjustment, which fuses current sensor data and temperature sensor data through the Kalman filter and dynamically corrects the overload trigger threshold according to ambient temperature and load fluctuations.

3. The method according to claim 1, characterized in that The self-diagnosis linkage unit constructs a fault level-response action mapping table and generates a comprehensive risk index based on the fault confidence output by the current overload protection model and the bearing fault feature deviation: ， Among them, R is the comprehensive risk index, which is used to evaluate the overall risk level of the system; α is the weight coefficient of the current overload protection model, which is used to adjust the influence of current overload on the comprehensive risk index; β is the weight coefficient of the bearing fault characteristic deviation, which is used to adjust the influence of the bearing fault characteristic on the comprehensive risk index; I over is the actual current overload value, indicating the extent to which the current in the system exceeds the normal range, I th is the current threshold, which indicates the maximum normal current value allowed by the system. F is the bearing fault characteristic deviation, which indicates the difference between the current bearing state and the normal state. base is the baseline fault characteristic value, which represents the characteristic value of the bearing under normal conditions.

4. The method according to claim 1, wherein The adaptive processing mechanism is a three-level response mode, including predictive load reduction, active thermal balancing, and hardware protection; The predictive load reduction dynamically adjusts the target torque of the pressure sensor through a PID algorithm based on the estimated temperature of the winding over time. The active thermal balance triggers the variable frequency operation of the cooling system, and optimizes the cooling fan speed distribution in real time through finite element simulation based on the temperature gradient distribution model. The hardware protection is used to directly cut off the IGBT drive signal when the high-frequency energy of the vibration signal and the current harmonic distortion rate exceed the specified threshold.

5. The method according to claim 1, wherein The bearing fault warning model uses improved variational mode decomposition (VMD) for noise reduction, optimizes VMD parameters using an energy difference grid search method to suppress endpoint effects, and uses a support vector regression machine to correct and reconstruct the signal. The amplitude-frequency modulated (AM / FM) signal is then demodulated using Hilbert envelope demodulation to extract the modulated high-frequency resonance components.

6. The method according to claim 1, wherein The deviation degree fuzzy comprehensive evaluation method quantifies the degree to which the characteristic frequency amplitude deviates from a preset baseline.

7. The method according to claim 1, characterized in that The collaborative allocation method of fusion weight information entropy-time decay quantifies the deviation, energy loss coefficient, and overload identification result feature credibility, and sets the feature effectiveness coefficient according to the time decay factor; the final fusion weight is generated by combining the modal contribution, the feature effectiveness coefficient and the quantized information entropy weight.

8. The method according to claim 1, characterized in that The intelligent diagnosis model receives the comprehensive fault feature vector and processes the heterogeneous data of vibration, current, temperature and torque through parallel channels. Among them, the vibration signal uses one-dimensional depthwise separable convolution to extract multi-scale impact features, the current and temperature data are captured with a bidirectional gated recurrent unit to capture the temporal dependency, and the torque signal uses a self-attention mechanism to model the spatial correlation.

9. The method according to claim 8, characterized in that The training process of the intelligent diagnosis model adopts a collaborative optimization strategy of transfer learning and adversarial domain adaptation.

10. A multi-sensor fusion intelligent actuator fault self-diagnosis system, used to perform the method according to any one of claims 1 to 9, characterized in that: It includes multi-sensor data synchronous acquisition module, overload initial judgment and adaptive response module, bearing fault feature demodulation module, multi-modal feature fusion weight module, intelligent diagnosis and strategy triggering module; The multi-sensor data synchronization acquisition module synchronizes the clocks of multiple sensors, uses Hall sensors to provide real-time feedback on motor speed and position information, uses pressure sensors to dynamically monitor output shaft torque, uses temperature sensors to detect motor winding and controller temperatures, uses current sensors to monitor motor current in real time, uses vibration sensors to monitor mechanical vibration signals in real time, centrally collects sensor data, and generates a synchronized time series by interpolating asynchronous data at different sampling rates from multiple sensors. The overload preliminary judgment and adaptive response module extracts motor current data and motor winding and controller temperature from the sensor data based on the synchronous time series and performs preliminary fault identification through the current overload protection model; based on the result of the preliminary fault identification, triggers the corresponding adaptive processing mechanism in the self-diagnosis linkage unit; The bearing fault feature demodulation module demodulates the amplitude-frequency modulation signal of the vibration sensor after noise reduction in the sensor data through the bearing fault early warning model, extracts the high-frequency energy of the demodulated signal and detects the deviation of the rolling bearing fault feature frequency amplitude from the preset baseline; The multimodal feature fusion weight module detects the energy loss coefficient caused by bearing abnormality by performing correlation analysis on the torque signal of the pressure sensor and the speed fluctuation signal of the Hall sensor; constructs a unified spatiotemporal coordinate system based on the synchronized time series, aligns the feature delays of different modal data, and sets the fusion weight based on the deviation, energy loss coefficient, and the overload identification result output by the current overload protection model; The intelligent diagnosis and strategy triggering module performs weighted fusion on the features of different modal data according to the fusion weight to generate a comprehensive fault feature vector; The comprehensive fault feature vector is input into an intelligent diagnosis model, and the intelligent diagnosis model performs fault pattern recognition and outputs fault type and fault location information; and a fault handling strategy is triggered according to the fault type and fault location information.

Images