Search

CN-122026779-A - Switch reluctance motor monitoring system and method

CN122026779ACN 122026779 ACN122026779 ACN 122026779ACN-122026779-A

Abstract

The invention relates to the technical field of motor monitoring and control, and discloses a system and a method for monitoring a switched reluctance motor. The system and the method for monitoring the switch reluctance motor realize comprehensive and accurate monitoring by constructing an integrated monitoring system integrating multi-physical-quantity synchronous sensing, high-speed acquisition, edge intelligent diagnosis and active fault tolerance, adopt a wideband and high-precision rogowski coil, an isolated voltage sensor and a dual-redundancy position sensor, combine distributed temperature and vibration sensing, realize multi-dimensional synchronous accurate sensing of electric, mechanical and thermal states, provide a high-quality data basis for fault diagnosis, realize real-time intelligence of diagnosis, ensure signal time sequence consistency based on high-speed synchronous acquisition of an FPGA, integrate a wavelet packet decomposition and lightweight convolutional neural network on the edge side, can extract characteristics and identify faults locally and quickly, and greatly reduce cloud dependence and diagnosis delay.

Inventors

  • LIU HUI

Assignees

  • 深蓝探索动力科技无锡有限公司

Dates

Publication Date
20260512
Application Date
20251224

Claims (9)

  1. 1. A switched reluctance motor monitoring system comprising: The multi-physical quantity sensing module is used for synchronously collecting phase current, phase voltage, rotor position, stator winding temperature and mechanical vibration signals of the switched reluctance motor; The high-speed synchronous data acquisition unit adopts an FPGA chip to realize synchronous sampling and preprocessing of multichannel signals, the sampling frequency is not lower than 50kHz, and the synchronous precision is better than +/-50 ns; The edge calculation diagnosis module integrates a wavelet transformation feature extraction unit and a lightweight convolutional neural network and is used for real-time fault diagnosis and health management; the redundant communication interface comprises a CAN-FD bus, an Ethernet interface and a wireless communication module, and supports multipath data transmission; And the cloud analysis platform is used for storing historical data, tracing deep faults and predicting service life.
  2. 2. The system of claim 1, wherein the multi-physical sensor module comprises: The Roche coil type phase current sensor has the bandwidth of DC-100kHz and the precision of +/-0.5%; the isolated phase voltage sensor adopts a voltage division and photoelectric isolation structure, and the withstand voltage grade is not lower than 1.5 times of the rated voltage of the motor; The double-redundancy rotor position sensor comprises a Hall effect sensor and a rotary transformer, and detects the fault of the position sensor through a cross verification mechanism; A distributed temperature sensor array embedded inside the stator winding and the power switch module; And the triaxial MEMS vibration sensor is arranged at the motor bearing seat.
  3. 3. The system of claim 1, wherein the edge calculation diagnostic module performs the steps of: carrying out wavelet packet decomposition on the acquired phase current signals, and extracting energy characteristic vectors of specific frequency bands; Inputting the feature vector into a lightweight convolutional neural network, and outputting fault type probability distribution; Triggering a dual redundant sensor data fusion mechanism when the diagnostic confidence is below a threshold; And generating a preliminary evaluation result of the health state index SOH and the residual service life RUL.
  4. 4. The system for monitoring the switched reluctance motor according to claim 1, wherein the system is provided with a three-stage early warning mechanism: The first-level early warning is that monitoring parameters exceed a normal range but do not reach a dangerous threshold, and the system records an abnormal log and increases sampling frequency; the secondary early warning is that latent fault characteristics are detected, and a system starts redundant sensors to cross-verify and report to the cloud; And three-stage early warning, namely, confirming serious faults, triggering active fault-tolerant control by the system, limiting the output power of the motor and informing maintenance personnel.
  5. 5. The system of claim 4, wherein the active fault tolerance control comprises: When the open-circuit fault of the single-phase winding is detected, the conduction angle and the current waveform are automatically adjusted, and the phase-loss derating operation is realized; Switching to a sensorless control mode based on current waveform estimation when the resolver fails; when the power switch tube is short-circuited, the corresponding bridge arm is quickly turned off and the standby power module is started.
  6. 6. The method for monitoring the switched reluctance motor is characterized by comprising the following steps of: S1, synchronously acquiring multiple physical quantity parameters of motor operation, and denoising and calibrating an original signal; S2, carrying out wavelet packet 3-layer decomposition on the phase current signals, and extracting energy distribution of 8 frequency bands as initial feature vectors; s3, adopting Principal Component Analysis (PCA) to reduce the dimension, and compressing the dimension of the feature vector to a key feature subspace; S4, inputting the feature subjected to dimension reduction into a trained lightweight convolutional neural network, and performing fault mode identification; S5, determining whether redundant sensor data are required to be fused or early warning is triggered based on a confidence evaluation result; And S6, outputting fault diagnosis results, health state indexes and maintenance suggestions to a human-computer interface and a cloud platform.
  7. 7. The method for monitoring a switched reluctance motor according to claim 6, wherein the lightweight convolutional neural network structure is: An input layer, a feature map of 32×32×1; a first convolution layer, which is composed of 16 3×3 convolution kernels, and batch normalization and ReLU activation; A second convolution layer, which is 32 3×3 convolution kernels, and performs batch normalization and ReLU activation; A maximum pooling layer, namely a2 multiplied by 2 pooling window; 64 neurons with a Dropout rate of 0.3; And the output layer is a Softmax classifier, and corresponds to five types of normal, winding short circuit, winding open circuit, position sensor fault and power converter fault.
  8. 8. The method for monitoring a switched reluctance motor according to claim 6, wherein the confidence level evaluation in step S5 is performed using the following criteria: If the confidence coefficient of the highest probability category is more than 85% and the probability of the next highest category is less than 10%, the diagnosis result is accepted; If the confidence is between 60% and 85%, triggering redundant sensor data fusion and recalculating the characteristics; if the confidence coefficient is less than 60%, marking the unknown anomaly, and uploading the original waveform to a cloud for deep analysis.
  9. 9. The method for monitoring a switched reluctance motor according to claim 6, wherein the state of health index SOH is calculated by: wherein alpha, beta and gamma are weight coefficients, which satisfy alpha+beta+gamma=1, and are dynamically adjusted according to the operation condition of the motor.

Description

Switch reluctance motor monitoring system and method Technical Field The invention relates to the technical field of motor monitoring and control, in particular to a system and a method for monitoring a switched reluctance motor. Background The switched reluctance motor (Switched Reluctance Motor, SRM) has the advantages of simple and firm structure, wide speed regulation range, high efficiency and the like, and is widely applied to the fields of electric automobiles, aerospace and the like. However, SRM is highly dependent on accurate rotor position information and complex power converter control, and problems such as winding insulation degradation, position sensor failure, and power switch tube failure seriously affect its operational reliability. The switched reluctance motor has the advantages of simple and firm structure, low cost, wide speed regulation range and the like, and is increasingly widely applied to the fields of electric automobiles, aviation industry, mining machinery and the like. However, the double salient structure and the pulse power supply mode cause the problems of large torque pulsation, noise, vibration protrusion and the like in operation, and key components such as a power converter, a winding, a bearing and the like are easy to generate faults such as overheating, short circuit, open circuit, abrasion and the like, so that the reliability and the service life of the system are directly influenced. The traditional monitoring method is mostly dependent on threshold judgment of single physical quantity (such as current), lacks synchronous accurate acquisition and depth correlation analysis of multi-dimensional signals, and is difficult to realize early and accurate fault diagnosis, so that the system and the method for monitoring the switch reluctance motor are provided. Disclosure of Invention The invention aims to provide a system and a method for monitoring a switched reluctance motor, which are used for solving the problems in the background technology. In order to solve the technical problems, the invention provides a switch reluctance motor monitoring system, which comprises: The multi-physical quantity sensing module is used for synchronously collecting phase current, phase voltage, rotor position, stator winding temperature and mechanical vibration signals of the switched reluctance motor; The high-speed synchronous data acquisition unit adopts an FPGA chip to realize synchronous sampling and preprocessing of multichannel signals, the sampling frequency is not lower than 50kHz, and the synchronous precision is better than +/-50 ns; The edge calculation diagnosis module integrates a wavelet transformation feature extraction unit and a lightweight convolutional neural network and is used for real-time fault diagnosis and health management; the redundant communication interface comprises a CAN-FD bus, an Ethernet interface and a wireless communication module, and supports multipath data transmission; And the cloud analysis platform is used for storing historical data, tracing deep faults and predicting service life. Preferably, the multi-physical quantity sensing module includes: The Roche coil type phase current sensor has the bandwidth of DC-100kHz and the precision of +/-0.5%; the isolated phase voltage sensor adopts a voltage division and photoelectric isolation structure, and the withstand voltage grade is not lower than 1.5 times of the rated voltage of the motor; The double-redundancy rotor position sensor comprises a Hall effect sensor and a rotary transformer, and detects the fault of the position sensor through a cross verification mechanism; A distributed temperature sensor array embedded inside the stator winding and the power switch module; And the triaxial MEMS vibration sensor is arranged at the motor bearing seat. Preferably, the edge calculation diagnosis module performs the steps of: carrying out wavelet packet decomposition on the acquired phase current signals, and extracting energy characteristic vectors of specific frequency bands; Inputting the feature vector into a lightweight convolutional neural network, and outputting fault type probability distribution; Triggering a dual redundant sensor data fusion mechanism when the diagnostic confidence is below a threshold; And generating a preliminary evaluation result of the health state index SOH and the residual service life RUL. Preferably, the system is provided with a three-level early warning mechanism: The first-level early warning is that monitoring parameters exceed a normal range but do not reach a dangerous threshold, and the system records an abnormal log and increases sampling frequency; the secondary early warning is that latent fault characteristics are detected, and a system starts redundant sensors to cross-verify and report to the cloud; And three-stage early warning, namely, confirming serious faults, triggering active fault-tolerant control by the system, limiting the output power of the m