Motor fault intelligent detection system and method based on multi-signal fusion

Abstract

The present application relates to the field of motor fault detection, and provides a motor fault intelligent detection system and method based on multi-signal fusion, comprising: a signal acquisition module for synchronously acquiring three-phase voltage signals, three-phase current signals and bearing vibration signals during operation of an asynchronous motor; a feature extraction module connected to the signal acquisition module, for pre-processing the collected multi-source signals and extracting stator fault features, rotor fault features and bearing fault features; a fault diagnosis module connected to the feature extraction module, which is internally provided with an AI artificial intelligence-based fault diagnosis model, for diagnosing faults according to the stator, rotor and bearing fault features. By fusing three-phase voltage, current and vibration multi-source signals and combining AI artificial intelligence algorithms, comprehensive health monitoring of the three key components of the asynchronous motor, i.e. the stator, rotor and bearing, is realized. The system can accurately identify incipient faults such as inter-turn short circuit, rotor broken bar and bearing wear.

Description

Technical Field

[0001] This invention relates to the field of electric motor fault detection, specifically to an intelligent electric motor fault detection system and method based on multi-signal fusion. Background Technology

[0002] Electric motors, as core power equipment in industrial production, are widely used in key sectors of the national economy such as metallurgy, chemical industry, power generation, and manufacturing. Their operating status directly affects the safety, efficiency, and cost of production lines. Especially in heavy industrial settings such as aluminum and steel plants, where numerous motors operate continuously for extended periods, malfunctions can lead to unplanned downtime, causing significant economic losses and even safety accidents. Therefore, effective condition monitoring and fault diagnosis of electric motors have become crucial for ensuring the continuity and safety of industrial production.

[0003] Currently, the maintenance and management of electric motors generally adopts the traditional model of "periodic maintenance" or "patrol inspection." Periodic maintenance involves disassembling and maintaining the motor according to a fixed time cycle. However, this method often fails to reflect the actual health condition of the motor. For motors in good condition, excessive disassembly and maintenance wastes manpower, resources, and downtime—this is known as "over-inspection." Conversely, motors with early-stage potential problems may not receive timely attention due to not being due for maintenance, leading to minor faults escalating into major accidents—this is known as "missed inspection." Patrol inspection relies on maintenance personnel using handheld instruments to measure parameters such as temperature and vibration on-site, combined with their experience for judgment. However, in harsh industrial environments with dust, high temperatures, and strong electromagnetic interference, manual patrol inspection is not only difficult to implement, but also struggles to accurately locate fault sources and assess fault severity based on single parameters and limited experience, easily leading to misdiagnosis or missed diagnosis. Furthermore, existing detection methods often employ single-signal analysis, lacking the fusion and complementarity of multi-source information, making diagnostic results susceptible to interference from load fluctuations, power supply harmonics, and other operating conditions, resulting in insufficient reliability. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides an intelligent motor fault detection system and method based on multi-signal fusion, which solves the problems of difficulty in implementing manual inspection and insufficient reliability.

[0005] To achieve the above objectives, the present invention provides the following technical solution: an intelligent motor fault detection system based on multi-signal fusion, comprising: The signal acquisition module is used to synchronously acquire the three-phase voltage signal, three-phase current signal, and bearing vibration signal of the asynchronous motor during operation. The feature extraction module, connected to the signal acquisition module, is used to preprocess the acquired multi-source signals and extract stator fault features, rotor fault features, and bearing fault features. The fault diagnosis module, connected to the feature extraction module, has a built-in AI-based fault diagnosis model for calculating the initial stator fault index based on the fault characteristics of the stator, rotor, and bearings. Rotor initial failure index and bearing initial failure index And compare them with the preset normalized diagnostic thresholds respectively; The results output module is used to generate a motor health status detection report and maintenance suggestions based on the comparison results of the fault diagnosis module.

[0006] Preferably, the feature extraction module includes: The stator fault feature extraction unit is used to calculate the stator negative sequence apparent impedance filter value and the stator apparent impedance angle, and use them as the basis for determining the initial stator fault. The rotor fault feature extraction unit is used to introduce high-frequency resolution force spectrum estimation and adaptive filtering technology to extract components of specific frequencies in the stator current signal as rotor fault features. The bearing fault feature extraction unit is used to integrate vibration signal analysis and stator current signal modulus analysis to extract bearing fault characteristic frequencies.

[0007] Preferably, the stator initial failure index The calculation formula is based on the stator negative sequence apparent impedance filtering value. and impedance angle Construct, and the normalized diagnostic threshold is 1.0; when When the value is greater than 1.0, it is determined that there is a short circuit fault between turns in the stator winding.

[0008] Preferably, the detection sensitivity of the stator fault diagnosis model is configured such that when the short-circuit turns ratio reaches 2% and the inter-turn short-circuit loop current reaches 0.5 times the rated current, the output... A fault signal greater than 1.0; and this fault characteristic is immune to changes in load conditions and faults in the rotor and bearings.

[0009] Preferably, the rotor initial failure index The normalized diagnostic threshold is constructed based on the spectral values ​​of specific frequency components in the current signal, and the normalized diagnostic threshold is 1.0; when When the value is greater than 1.0, it is determined that there is a rotor bar breakage fault, and its detection sensitivity covers the initial fault of a single bar breakage.

[0010] Preferably, the bearing initial failure index The normalized diagnostic threshold is constructed based on the fusion analysis results of vibration and current signals, and the normalized diagnostic threshold is 1.0; when If the value is greater than 1.0, it is determined that there is bearing wear or cage failure.

[0011] The method for intelligent fault detection of electric motors based on multi-signal fusion includes the following steps: Step 1: Simultaneously measure the three-phase voltage, three-phase current, and bearing vibration signals of the asynchronous motor within a 50-second time period at the control cabinet using the signal acquisition module; Step 2: Use the feature extraction module to filter and decouple the acquired signals, and extract the initial fault feature quantities for the stator, rotor and bearing respectively. Step 3: Using the fault diagnosis module, input the extracted fault features into the AI ​​artificial intelligence prediction algorithm model to calculate the stator initial fault index. Rotor initial failure index and bearing initial failure index ; Step 4: Compare each fault index with the preset threshold of 1.0. If any index exceeds 1.0, it is determined that the corresponding component has an initial fault, and a test report containing the fault location, fault severity level and maintenance suggestions is generated through the result output module.

[0012] Preferably, in step one, the signal acquisition process is completed through non-intrusive wiring, including using voltage test clamps to connect the three-phase voltage terminals A, B, and C in the control cabinet, and using current test clamps to connect the three-line currents A, B, and C.

[0013] Preferably, the AI ​​artificial intelligence prediction algorithm model in step three has a built-in immune algorithm to eliminate the interference of load conditions, power fluctuations and non-target fault types on the current diagnostic results.

[0014] Preferably, the method further includes step five: automatically generating a motor health status trend curve using historical test data, which is used to dynamically assess the motor life and provide early warning of the optimal maintenance time.

[0015] This invention provides an intelligent fault detection system and method for electric motors based on multi-signal fusion. It has the following beneficial effects: 1. This invention integrates multi-source signals of three-phase voltage, current, and vibration, combined with AI artificial intelligence algorithms, to achieve comprehensive health monitoring of the three key components of an asynchronous motor: stator, rotor, and bearings. The system can accurately identify initial faults such as inter-turn short circuits, rotor bar breakage, and bearing wear, providing sensitive early warnings at the fault's nascent stage. This effectively avoids the waste of over-inspection and sudden accidents caused by missed inspections resulting from traditional periodic maintenance. By calculating a normalized fault index and comparing it with a threshold, it achieves a leap from manual experience-based judgment to intelligent predictive maintenance, significantly improving the safety and continuity of equipment operation.

[0016] 2. This invention employs a non-invasive detection method, requiring only quick wiring within the control cabinet to collect data. It is convenient to operate and unaffected by harsh industrial environments. The system incorporates an immune algorithm that automatically eliminates interference from load changes, power fluctuations, and non-target faults, ensuring high reliability and stability of diagnostic results. By generating detailed health status reports and fault trend curves, it provides a scientific basis for equipment maintenance decisions, helping companies optimize spare parts management, reduce operating costs, extend motor lifespan, and align with green production guidelines, resulting in significant economic and social benefits. Attached Figure Description

[0017] Figure 1 This is a system diagram of the present invention; Figure 2 This is a flowchart of the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Example: As one aspect of the present invention, please refer to the appendix. Figure 1 This invention provides an intelligent motor fault detection system based on multi-signal fusion, comprising: The signal acquisition module includes voltage sensors, current sensors, and vibration sensors. The voltage sensors utilize three high-precision voltage test clamps (Fluke AC220 model), connected to the A, B, and C phase voltage terminal blocks in the asynchronous motor control cabinet for synchronous acquisition of three-phase voltage signals. The current sensors utilize three flexible current test clamps (PEM CWTMini model), clipped onto the A, B, and C phase lines for acquisition of three-phase current signals. The vibration sensor uses a triaxial piezoelectric accelerometer (PCB 356A15 model), magnetically attached to the vertical and horizontal directions of the motor bearing end cover to acquire bearing vibration signals. The signal acquisition module incorporates a synchronous trigger circuit to ensure strict time synchronization of the three-phase voltage, three-phase current, and vibration signals. The sampling frequency is uniformly set to 10kHz (20kHz for vibration signals to improve resolution), with each acquisition lasting 50 seconds. Time-domain waveform data is generated and transmitted to the feature extraction module via shielded cable or wireless transmission.

[0020] The feature extraction module is based on a digital signal processor (DSP, model TMS320C6678), with an integrated FPGA for data preprocessing. The received raw signal first undergoes anti-aliasing filtering (cutoff frequency 5kHz) and analog-to-digital conversion (24-bit resolution), and then is processed in the following units:

[0021] Stator fault feature extraction unit: Performs synchronous coordinate transformation on three-phase voltage and current, uses a phase-locked loop (PLL) to track the fundamental frequency, and calculates the positive and negative sequence components of the 50Hz fundamental frequency using Fourier transform (FFT, Hanning window, 256 points per cycle). Calculates the negative sequence apparent impedance filtering value. and impedance angle After being filtered by moving average, it is used as a stator fault characteristic quantity. .

[0022] Rotor fault feature extraction unit: Employs high-frequency resolution force spectrum estimation (MUSIC) to perform spectral analysis on a single-phase current signal, with a resolution of 0.01Hz. Extracts characteristic frequency components of the broken rotor bars. ,in This is the power supply base frequency. The slip is estimated from rotor slot harmonics in the vibration signal or current. An adaptive filter is used to suppress fundamental and harmonic interference, and the amplitude of this frequency component is extracted. , which serves as a characteristic quantity of rotor faults.

[0023] Bearing fault feature extraction unit: This unit fuses vibration and current signals. After bandpass filtering, the vibration signal undergoes Hilbert transform to obtain the envelope spectrum, from which bearing fault characteristic frequencies (such as outer race fault frequencies) are extracted. Inner ring failure frequency rolling element failure frequency Calculate the energy within ±10Hz of each frequency. Simultaneously, the same envelope analysis is performed on the current signal to extract the amplitude of the modulated sideband components in the current. As a characteristic quantity of bearing failure The feature extraction module packages the extracted feature vectors and sends them to the fault diagnosis module.

[0024] The fault diagnosis module employs an embedded AI computing platform, incorporating a fault diagnosis model based on a deep neural network. This model is a three-layer fully connected network with 6 nodes in the input layer (corresponding to the aforementioned features), 64 and 32 nodes in the hidden layers respectively, using ReLU activation, and three nodes in the output layer, each corresponding to the initial fault index of the stator. Rotor initial failure index and bearing initial failure index The activation function is Sigmoid, and the output range is [0, 2]. The network is trained using a large amount of historical fault data, with mean squared error as the loss function and Adam as the optimizer. During diagnosis, the feature vector is input into the network, and the calculated value is obtained. , , And compare it with the preset normalized diagnostic threshold of 1.0: If If the value is greater than 1.0, it is determined to be a short circuit fault between turns in the stator winding; if If the value is greater than 1.0, the rotor is determined to have a broken bar fault; if A value greater than 1.0 indicates bearing wear or cage failure. The diagnostic module incorporates an immune algorithm that dynamically adjusts the baseline values ​​of characteristic quantities by monitoring load current and voltage fluctuations in real time, eliminating the impact of power supply fluctuations and load changes. The results output module includes a touch screen and a wireless communication unit, generating a motor health status report based on the diagnostic results. The report includes basic motor information, a health status overview, fault details, maintenance and repair recommendations, and original waveforms and spectrum diagrams. The report can be uploaded to a cloud platform via Wi-Fi or exported via USB.

[0025] As another aspect of the present invention, please refer to the appendix. Figure 2 This invention provides a method for intelligent detection of motor faults based on multi-signal fusion, comprising the following steps: Step 1: Signal Acquisition. The signal acquisition module synchronously measures the three-phase voltage, three-phase current, and bearing vibration signals of the asynchronous motor over a 50-second period at the control cabinet. The operator initiates the acquisition task using a handheld tablet. The voltage test clamp is connected to the A, B, and C phase voltage terminals inside the control cabinet, the current test clamp is attached to the corresponding three-phase lines, and the vibration sensor is attached to the motor housing near the bearing. Waveforms are displayed in real-time during the acquisition process to ensure the signal is free of clipping. After acquisition, the data is saved with a timestamp.

[0026] Step 2, Feature Extraction: The feature extraction module is used to filter and decouple the acquired signals. First, the voltage and current signals are low-pass filtered (cutoff frequency 1kHz), and the vibration signals are band-pass filtered (500Hz~10kHz), followed by analog-to-digital conversion. Features are extracted as follows: Stator characteristics: The fundamental positive and negative sequence components are calculated using FFT with a Hanning window to obtain the negative sequence apparent impedance filtering value. and impedance angle .

[0027] Rotor characteristics: The MUSIC algorithm is used to extract current signals. Amplitude of frequency components And estimate energy The same processing is applied to the current signal to extract the modulation sideband amplitude. .

[0028] Step 3: Fault Diagnosis: The feature vector obtained in Step 2 is input into the AI ​​artificial intelligence prediction algorithm model of the fault diagnosis module. The model is a pre-trained neural network, which calculates the initial stator fault index. Rotor initial failure index and bearing initial failure index The calculation process runs on an embedded AI platform and takes less than 0.1 seconds.

[0029] Step 4: Result Generation: Compare each fault index with the preset threshold of 1.0. Assume the results of this test are: , , If the result is positive, the rotor is determined to have a minor broken bar fault. The result output module generates a test report, which includes the fault location (rotor), fault severity level (minor), confidence level (92%), and maintenance recommendations (such as "Inspect the rotor guide bars and end rings, and plan to shut down the machine for further testing in the near future").

[0030] Step 5: Trend Analysis: The system automatically stores the results in the historical database and plots the fault index trend curve for the past 6 months. By analyzing the changing trend of the fault index, the remaining lifespan of the motor is dynamically assessed, and an early warning is issued when the fault index approaches the threshold, enabling predictive maintenance.

[0031] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A motor fault intelligent detection system based on multi-signal fusion, characterized in that, include: The signal acquisition module is used to synchronously acquire the three-phase voltage signal, three-phase current signal, and bearing vibration signal of the asynchronous motor during operation. The feature extraction module, connected to the signal acquisition module, is used to preprocess the acquired multi-source signals and extract stator fault features, rotor fault features, and bearing fault features. The fault diagnosis module, connected to the feature extraction module, has a built-in AI-based fault diagnosis model for calculating the initial stator fault index based on the fault characteristics of the stator, rotor, and bearings. Rotor initial failure index and bearing initial failure index And compare them with the preset normalized diagnostic thresholds respectively; The results output module is used to generate a motor health status detection report and maintenance suggestions based on the comparison results of the fault diagnosis module.

2. The intelligent motor fault detection system based on multi-signal fusion according to claim 1, characterized in that, The feature extraction module includes: The stator fault feature extraction unit is used to calculate the stator negative sequence apparent impedance filter value and the stator apparent impedance angle, and use them as the basis for determining the initial stator fault. The rotor fault feature extraction unit is used to introduce high-frequency resolution force spectrum estimation and adaptive filtering technology to extract components of specific frequencies in the stator current signal as rotor fault features. The bearing fault feature extraction unit is used to integrate vibration signal analysis and stator current signal modulus analysis to extract bearing fault characteristic frequencies.

3. The intelligent motor fault detection system based on multi-signal fusion according to claim 1, characterized in that, The stator initial failure index The calculation formula is based on the stator negative sequence apparent impedance filtering value. and impedance angle Construct, and the normalized diagnostic threshold is 1.0; when When the value is greater than 1.0, it is determined that there is a short circuit fault between turns in the stator winding.

4. The intelligent motor fault detection system based on multi-signal fusion according to claim 3, characterized in that, The detection sensitivity of the stator fault diagnosis model is configured as follows: when the short-circuit turns ratio reaches 2% and the inter-turn short-circuit loop current reaches 0.5 times the rated current, the output... A fault signal greater than 1.0; and this fault characteristic is immune to changes in load conditions and faults in the rotor and bearings.

5. The intelligent motor fault detection system based on multi-signal fusion according to claim 1, characterized in that, The rotor initial failure index The normalized diagnostic threshold is constructed based on the spectral values ​​of specific frequency components in the current signal, and the normalized diagnostic threshold is 1.0; when When the value is greater than 1.0, it is determined that there is a rotor bar breakage fault, and its detection sensitivity covers the initial fault of a single bar breakage.

6. The intelligent motor fault detection system based on multi-signal fusion according to claim 1, characterized in that, The bearing initial failure index The normalized diagnostic threshold is constructed based on the fusion analysis results of vibration and current signals, and the normalized diagnostic threshold is 1.0; when If the value is greater than 1.0, it is determined that there is bearing wear or cage failure.

7. A method for intelligent detection of motor faults based on multi-signal fusion, using the intelligent detection system for motor faults based on multi-signal fusion as described in any one of claims 1-6, characterized in that, Includes the following steps: Step 1: Simultaneously measure the three-phase voltage, three-phase current, and bearing vibration signals of the asynchronous motor within a 50-second time period at the control cabinet using the signal acquisition module; Step 2: Use the feature extraction module to filter and decouple the acquired signals, and extract the initial fault feature quantities for the stator, rotor and bearing respectively. Step 3: Using the fault diagnosis module, input the extracted fault features into the AI ​​artificial intelligence prediction algorithm model to calculate the stator initial fault index. Rotor initial failure index and bearing initial failure index ; Step 4: Compare each fault index with the preset threshold of 1.

0. If any index exceeds 1.0, it is determined that the corresponding component has an initial fault, and a test report containing the fault location, fault severity level and maintenance suggestions is generated through the result output module.

8. The method for intelligent detection of motor faults based on multi-signal fusion according to claim 7, characterized in that, In step one, the signal acquisition process is completed through non-intrusive wiring, including using voltage test clamps to connect the three-phase voltage terminals A, B, and C in the control cabinet, and using current test clamps to connect the three-line current of A, B, and C.

9. The method for intelligent detection of motor faults based on multi-signal fusion according to claim 7, characterized in that, The AI ​​prediction algorithm model in step three has a built-in immune algorithm to eliminate the interference of load conditions, power fluctuations, and non-target fault types on the current diagnostic results.

10. The intelligent detection method for motor faults based on multi-signal fusion according to claim 7, characterized in that, The method also includes step five: automatically generating a motor health status trend curve using historical test data, which is used to dynamically assess the motor life and provide early warning of the best maintenance time.

Images