Fault diagnosis method and device thereof

Abstract

A method for diagnosing a fault of a rotating motor comprises: obtaining a time-domain acceleration signal of the rotating motor along a vibration direction and a rotation period (210) of the rotating motor; converting the time-domain acceleration signal into a time-domain velocity signal, cutting out a plurality of velocity signal segments along the time-domain velocity signal with the rotation period as a step length, and sequentially arranging the plurality of velocity signal segments to obtain a velocity signal matrix (220); converting the velocity signal matrix into an image (230); and inputting the image into a trained neural network model to obtain a fault diagnosis result (240). The method converts the fault diagnosis problem into an image recognition problem by converting the velocity signal matrix into an image and inputting the image into the trained neural network model to obtain a fault diagnosis result. The method is applicable to image-oriented neural network models and improves the efficiency and accuracy of fault diagnosis.

Description

Technical Field

[0001] The present invention mainly relates to the field of computer technology, and in particular to a fault diagnosis method and device thereof. Background Art

[0002] In industrial systems, the performance of mechanical devices, such as rotating motors, plays a crucial role. Promptly monitoring the status of these devices and conducting fault diagnosis can help predict abnormal conditions before actual failures occur, ensuring optimal operation.

[0003] When diagnosing mechanical faults, various types of signals are collected from the device. Vibration signals from various locations and directions are among the most commonly used. When using vibration signals for fault diagnosis, the vibration signals are first collected and then subjected to a series of analyses. For example, analysis can be performed according to international standards such as ISO 13373-1, ISO 13373-2, and ISO 13373-3 to determine whether a fault has occurred and what type of fault it is.

[0004] When performing spectral analysis on vibration signals, technicians need to have relevant domain knowledge, which limits the application of fault diagnosis based on vibration signals. With the rapid development of artificial intelligence technology, the industry is trying to use this technology for fault diagnosis. However, in current neural network models, their acceleration hardware (such as network processors NPU) and platforms (such as TensorFlow, Caffe) are mainly oriented towards images, text and voice data, and are not suitable for vibration signals required for fault diagnosis. In addition, different types of vibration sensors have different sampling rates and resolutions, and a single neural network model cannot be applied. In order to obtain more accurate diagnostic results, the neural network model needs to be trained separately using vibration data with different sampling rates and resolutions, which increases the complexity and time of training. Summary of the Invention

[0005] In order to solve the above technical problems, the present invention provides a fault diagnosis method and device thereof to convert the fault diagnosis problem into an image recognition problem, which is applicable to an image-oriented neural network model to improve the efficiency and accuracy of fault diagnosis.

[0006] To achieve the above-mentioned objectives, the present invention proposes a method for fault diagnosis of a rotating motor, the method comprising: obtaining a time-domain acceleration signal of the rotating motor along a vibration direction and a rotation period of the rotating motor; converting the time-domain acceleration signal into a time-domain velocity signal, extracting a plurality of velocity signal segments along the time-domain velocity signal with the rotation period as a step size, and arranging the plurality of velocity signal segments in sequence to obtain a velocity signal matrix; converting the velocity signal matrix into an image; and inputting the image into a trained neural network model to obtain a fault diagnosis result.

[0007] To this end, the speed signal matrix is ​​converted into an image and the image is input into a trained neural network model to obtain the fault diagnosis result, thereby converting the fault diagnosis problem into an image recognition problem, which is suitable for image-oriented neural network models and improves the efficiency and accuracy of fault diagnosis.

[0008] In one embodiment of the present invention, converting the time domain acceleration signal into a time domain velocity signal includes: converting the time domain acceleration signal into a frequency domain acceleration signal using a fast Fourier transform; performing frequency domain integration on the frequency domain acceleration signal to obtain a frequency domain velocity signal; and converting the frequency domain velocity signal into a time domain velocity signal using an inverse fast Fourier transform.

[0009] To this end, frequency domain integration is used to remove high-frequency noise in the vibration signal.

[0010] In one embodiment of the present invention, extracting a plurality of speed signal segments along the time domain speed signal with the rotation period as a step size includes: extracting a plurality of speed signal segments along the time domain speed signal with the rotation period as a step size without overlap.

[0011] To this end, multiple speed signal segments are intercepted along the time domain speed signal without overlap using the rotation period as a step size, and adjacent speed signal segments in the multiple speed signal segments do not overlap, thereby improving data stability and fault diagnosis accuracy.

[0012] In one embodiment of the present invention, converting the velocity signal matrix into an image includes: multiplying the velocity signal matrix by a normalization coefficient to obtain a normalized velocity signal matrix; and mapping the normalized velocity signal matrix to the image.

[0013] To this end, the velocity signal matrix is ​​normalized so that vibration data with different sampling rates and resolutions can be applied, which reduces the complexity of the neural network model and improves the efficiency of fault diagnosis.

[0014] In one embodiment of the present invention, converting the velocity signal matrix into an image includes: converting the velocity signal matrix into a color image.

[0015] To this end, converting the speed signal matrix into a color image can improve the image recognition, making it easier for technicians to mark it when training the neural network model, thereby improving the accuracy of the neural network model in fault diagnosis.

[0016] In one embodiment of the present invention, after converting the speed signal matrix into an image, the method further includes: adjusting the size of the image to a predetermined size of the neural network model.

[0017] To this end, by adjusting the size of the image to the predetermined size of the neural network model, the complexity of the neural network model can be reduced and the efficiency of fault diagnosis can be improved.

[0018] The present invention also proposes a fault diagnosis device for a rotating motor, which includes: an acquisition unit, which acquires the time domain acceleration signal of the rotating motor along the vibration direction and the rotation period of the rotating motor; an alignment and arrangement unit, which converts the time domain acceleration signal into a time domain speed signal, cuts out multiple speed signal segments along the time domain speed signal with the rotation period as a step size, and arranges the multiple speed signal segments in sequence to obtain a speed signal matrix; a conversion unit, which converts the speed signal matrix into an image; and a judgment unit, which inputs the image into a trained neural network model to obtain a fault diagnosis result.

[0019] In one embodiment of the present invention, the acquisition unit converts the time domain acceleration signal into a time domain velocity signal, including: converting the time domain acceleration signal into a frequency domain acceleration signal using a fast Fourier transform; performing frequency domain integration on the frequency domain acceleration signal to obtain a frequency domain velocity signal; and converting the frequency domain velocity signal into a time domain velocity signal using an inverse fast Fourier transform.

[0020] In one embodiment of the present invention, the aligning and arranging unit extracting a plurality of speed signal segments along the time domain speed signal with the rotation period as a step size includes: extracting a plurality of speed signal segments along the time domain speed signal with the rotation period as a step size without overlap.

[0021] In an embodiment of the present invention, the converting unit converts the speed signal matrix into an image, comprising: multiplying the speed signal matrix by a normalization coefficient to obtain a normalized speed signal matrix; and mapping the normalized speed signal matrix to the image.

[0022] In an embodiment of the present invention, the converting unit converting the speed signal matrix into an image includes: converting the speed signal matrix into a color image.

[0023] In one embodiment of the present invention, after the conversion unit converts the speed signal matrix into an image, the method further includes: adjusting the size of the image to a predetermined size of the neural network model.

[0024] The present invention also provides an electronic device, comprising a processor, a memory, and instructions stored in the memory, wherein the instructions implement the above-mentioned method when executed by the processor.

[0025] The present invention also provides a computer-readable storage medium having computer instructions stored thereon, which, when executed, execute the method as described above. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] The following drawings are only intended to illustrate and explain the present invention, and do not limit the scope of the present invention.

[0027] Figure 1 It is a schematic diagram of a fault diagnosis model in the prior art;

[0028] Figure 2 is a flow chart of a fault diagnosis method according to an embodiment of the present invention;

[0029] Figure 3 is a process diagram of a fault diagnosis method according to an embodiment of the present invention;

[0030] Figure 4 is a block diagram of a fault diagnosis device according to an embodiment of the present invention.

[0031] Description of Reference Numerals

[0032] 100 fault diagnosis model

[0033] 110 data acquisition unit

[0034] 120 pre-processing units

[0035] 130 feature extraction units

[0036] 140 diagnostic units

[0037] 141 Diagnostic Algorithm

[0038] 150 training set input units

[0039] 200 Fault Diagnosis Method

[0040] Steps 210-240

[0041] 301 Time Domain Acceleration Signal

[0042] 302 time domain velocity signal

[0043] 302a-302d speed signal segments

[0044] 303 speed signal matrix

[0045] 303a Local enlargement of the speed signal matrix

[0046] 304 Normalized velocity signal matrix

[0047] 304a A local magnified portion of the normalized velocity signal matrix

[0048] 305 images

[0049] 306 Resized Image

[0050] 400 fault diagnosis device

[0051] 410 Get Unit

[0052] 420 alignment units

[0053] 430 conversion unit

[0054] 440 judgment unit DETAILED DESCRIPTION

[0055] In order to have a clearer understanding of the technical features, purposes and effects of the present invention, specific embodiments of the present invention are now described with reference to the accompanying drawings.

[0056] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0057] As used in this application and the claims, unless the context clearly indicates otherwise, the words "a," "an," "an," and / or "the" are not intended to refer to the singular but may include the plural. Generally speaking, the terms "comprises" and "include" only indicate the inclusion of the steps and elements specifically identified, and these steps and elements do not constitute an exclusive list. A method or apparatus may also include other steps or elements.

[0058] Figure 1 Schematic diagram of a fault diagnosis model 100 in the prior art. Figure 1 As shown, the fault diagnosis model 100 includes a data acquisition unit 110 , a pre-processing unit 120 , a feature extraction unit 130 , a diagnosis unit 140 and a training set input unit 150 .

[0059] When the fault diagnosis model 100 performs fault diagnosis, the data acquisition unit 110 (such as an acceleration sensor) collects the vibration signal of the rotating motor along the vibration direction, and the preprocessing unit 120 preprocesses the vibration signal for analysis and processing by subsequent modules. For example, the vibration signal is subjected to fast Fourier transform (FFT) to convert the time domain vibration signal into a frequency domain vibration signal. The feature extraction unit 130 performs spectrum analysis on the frequency domain vibration signal to extract the characteristic frequency. The diagnosis unit 140 uses the diagnosis algorithm 141 to diagnose the input characteristic frequency and output the diagnosis result. The training set input unit 150 inputs the training set to the diagnosis unit 140 to train the diagnosis unit 140.

[0060] For fault diagnosis model 100, different types of data acquisition units 110 have different sampling rates and resolutions, making a single neural network model inadequate. To obtain more accurate diagnostic results, the neural network model must be trained separately using vibration data at different sampling rates and resolutions, which increases training complexity and time.

[0061] Figure 2 is a flowchart of a fault diagnosis method according to an embodiment of the present invention. Figure 3 This is a process diagram of a fault diagnosis method according to an embodiment of the present invention. Figure 2 and Figure 3 The fault diagnosis method in this embodiment is described.

[0062] Step 210 : Acquire a time domain acceleration signal of the rotating motor along the vibration direction and a rotation period of the rotating motor.

[0063] Rotation and vibration are two forms of motion in a rotating motor. A rotating motor rotates around its axis to output torque. The vibration signal along the vibration direction indicates a fault in the rotating motor. By analyzing the vibration signal along the vibration direction of the rotating motor, it can be determined whether the rotating motor has a fault and the type of fault.

[0064] The vibration direction can be axial or radial. Multiple vibration sensors can be arranged in the axial and / or radial directions of the rotating motor to obtain vibration signals of the rotating motor along the vibration direction. In an embodiment of the present invention, the vibration signal can be a time-domain acceleration signal, which can be collected by an acceleration sensor, such as a MEMS (micro-electromechanical system) accelerometer.

[0065] The acceleration sensor is used to collect a time domain acceleration signal of the rotating motor along the vibration direction to obtain the time domain acceleration signal. Figure 3The time domain acceleration signal 301 collected by an accelerometer is shown, where the horizontal axis is time in milliseconds (ms) and the vertical axis is acceleration value in m / s 2 The time domain acceleration signal 301 shows the acceleration values ​​at different time points. It is worth noting that different accelerometers may have different sampling rates, that is, sampling at different time intervals, and different accelerometers may also have different resolutions, that is, the collected acceleration values ​​have different decimal places.

[0066] The rotation period of a rotating motor can be calculated based on the motor's speed. Specifically, an encoder captures the motor's speed, from which the motor's rotation period (the time it takes for the motor to complete one rotation) can be calculated. For example, if the encoder captures a motor speed of 1800 rpm (30 revolutions per second), each revolution takes 1 / 30 second, and the rotation period is 1 / 30 second.

[0067] Step 220: Convert the time-domain acceleration signal into a time-domain velocity signal, extract multiple velocity signal segments along the time-domain velocity signal with a rotation period as a step length, and sequentially arrange the multiple velocity signal segments to obtain a velocity signal matrix;

[0068] The time domain acceleration signal is converted into the time domain velocity signal by performing time domain integration on the time domain acceleration signal or by converting the time domain acceleration signal into the frequency domain and then performing frequency domain integration.

[0069] When performing time-domain integration, the time-domain acceleration signal is directly integrated to obtain the time-domain velocity signal. When performing frequency-domain integration, the fast Fourier transform (FFT) is used to convert the time-domain acceleration signal into a frequency-domain acceleration signal; the frequency-domain acceleration signal is then integrated in the frequency domain to obtain the frequency-domain velocity signal; and the inverse fast Fourier transform (IFFT) is used to convert the frequency-domain velocity signal into the time-domain velocity signal. Therefore, frequency-domain integration can remove high-frequency noise from the vibration signal.

[0070] Figure 3 The time domain acceleration signal 301 is converted into a time domain velocity signal 302 , wherein the horizontal axis represents time in seconds (s) and the vertical axis represents velocity value in m / s. The time domain velocity signal 302 displays velocity values ​​at different time points.

[0071] After obtaining the time-domain velocity signal, multiple velocity signal segments are extracted along the time-domain velocity signal in steps of the rotation period. Multiple velocity signal segments can be extracted along the time-domain velocity signal in steps of the rotation period starting from the initial moment. If the initial time-domain velocity signal has high noise, these noisy signal segments can be skipped to ensure data reliability and improve the accuracy of fault diagnosis.

[0072] In one optional scenario, multiple speed signal segments are extracted along the time domain speed signal without overlap, using the rotation period as a step size. After determining the starting point of the initial signal segment, an initial signal segment with a length of the rotation period is obtained along the time domain speed signal. Then, starting from the next data point after the end of the initial signal segment, a second signal segment with a length of the rotation period is obtained along the time domain speed signal. Similarly, multiple speed signal segments can be extracted, and each extracted speed signal segment has the same amount of speed data. Adjacent speed signal segments among the multiple speed signal segments do not overlap, thereby improving data stability and fault diagnosis accuracy.

[0073] After obtaining multiple velocity signal segments, since each velocity signal segment has the same amount of velocity data, the velocity signal segments are sequentially arranged to form a velocity signal matrix. The velocity signal matrix can be formed by sequentially arranging the multiple velocity signal segments row by row from top to bottom or from bottom to top. Alternatively, the velocity signal matrix can be formed by sequentially arranging the multiple velocity signal segments column by column from left to right or from right to left.

[0074] like Figure 3 As shown, starting from the initial moment, at least four velocity signal segments 302a, 302b, 302c, and 302d are extracted from the time-domain velocity signal 302 without overlap based on a rotation period (e.g., 1 / 30 second in the example above). Adjacent velocity signal segments do not overlap, and velocity signal segments 302a, 302b, 302c, and 302d have the same amount of velocity data. Multiple velocity signal segments 302a, 302b, 302c, and 302d are arranged in rows from bottom to top to obtain a velocity signal matrix 303. From a partially enlarged portion 303a of the velocity signal matrix 303, it can be seen that velocity signal segment 302c contains velocity data of 3.5, 12.5, 12.1, and 9.0, and velocity signal segment 302d contains velocity data of 2.4, 11.3, 11.1, and 8.7.

[0075] Step 230: Convert the velocity signal matrix into an image.

[0076] In this step, the speed signal matrix is ​​converted into an image, and the fault diagnosis problem is converted into an image recognition problem, which can be applied to the image-oriented neural network model, thereby improving the efficiency and accuracy of fault diagnosis.

[0077] In one optional embodiment, converting the velocity signal matrix into an image may include: multiplying the velocity signal matrix by a normalization coefficient to obtain a normalized velocity signal matrix; and mapping the normalized velocity signal matrix to the image. By normalizing the velocity signal matrix, vibration data with different sampling rates and resolutions can be used, reducing the complexity of the neural network model and improving the efficiency of fault diagnosis.

[0078] Optionally, the velocity signal matrix can be converted into a color image. The color image can be a color image of the three primary colors of light (red, green, and blue, RGB) or a color image of the four printing colors (cyan, magenta, and yellow key, CMYK). Converting the velocity signal matrix into a color image can improve image recognition, making it easier for technicians to mark it when training the neural network model, thereby improving the accuracy of the neural network model in fault diagnosis.

[0079] Taking the RGB color image as an example, the speed signal matrix is ​​multiplied by a normalization coefficient so that the maximum value of the speed data in the speed signal matrix is ​​near 255. Then, the speed signal matrix is ​​mapped into an RGB color image using pseudocolor rules such as those in MATLAB. Figure 3 A normalized velocity signal matrix 304 using a normalization coefficient of 10 is shown. A partially enlarged portion 304a of the normalized velocity signal matrix 304 shows that the velocity data of velocity signal segment 302c is normalized to 35, 125, 121, and 90, and the velocity data of velocity signal segment 302d is normalized to 24, 113, 111, and 87. By converting the normalized velocity signal matrix 304 using a pseudo-color function, an RGB image 305 can be obtained.

[0080] In some embodiments, after converting the speed signal matrix into an image, the image size is adjusted to the predetermined size of the neural network model, such as 320mm*320mm. By adjusting the image size to the predetermined size of the neural network model, the complexity of the neural network model can be reduced and the efficiency of fault diagnosis can be improved. Figure 3 In FIG. 3 , the RGB image 305 is resized to become the image 306 , and the size of the image 306 is 320 mm*320 mm.

[0081] Step 240: Input the image into the trained neural network model to obtain a fault diagnosis result.

[0082] The neural network model may be an AlexNet or GoogleNet neural network model. In an embodiment of the present invention, an image dataset may be used to train the neural network model. Training the neural network model may include: (1) marking the fault type to be identified and converting it into a file format required for training; (2) dividing the converted image files into a training set and a test set, training the neural network model using the training set, and testing the trained neural network model using the test set.

[0083] After testing, a trained neural network model can be obtained, and the image output in step 230 can be analyzed and processed using the trained neural network model to obtain a fault diagnosis result. For example, faults such as unbalance, looseness, and eccentricity can be diagnosed based on the input image. If no marked fault type is identified, it is determined that no fault has occurred in the rotating motor. For example, if Figure 3 The image 306 shown is a typical unbalanced image. When the image 306 is input into the neural network model, a fault diagnosis result is obtained that an unbalanced fault occurs.

[0084] This embodiment of the present invention provides a method for fault diagnosis of a rotating motor. By converting a speed signal matrix into an image and inputting the image into a trained neural network model, a fault diagnosis result is obtained, thereby converting the fault diagnosis problem into an image recognition problem. It is suitable for image-oriented neural network models, thereby improving the efficiency and accuracy of fault diagnosis.

[0085] Flowcharts are used herein to illustrate the operations performed by the methods according to the embodiments of the present application. It should be understood that the preceding operations are not necessarily performed in exact order. Instead, the various steps may be performed in reverse order or simultaneously. Furthermore, other operations may be added to these processes, or one or more operations may be removed from these processes.

[0086] Figure 4 FIG. 4 is a block diagram of a fault diagnosis device 400 according to an embodiment of the present invention. Figure 4 As shown, the fault diagnosis device 400 in this embodiment includes:

[0087] An acquisition unit 410 acquires a time-domain acceleration signal of the rotating motor along the vibration direction and a rotation period of the rotating motor; an alignment and arrangement unit 420 converts the time-domain acceleration signal into a time-domain velocity signal, extracts a plurality of velocity signal segments along the time-domain velocity signal with a rotation period as a step length, and sequentially arranges the plurality of velocity signal segments to obtain a velocity signal matrix; a conversion unit 430 converts the velocity signal matrix into an image; and a determination unit 440 inputs the image into a trained neural network model to obtain a fault diagnosis result.

[0088] In an optional case, the acquisition unit 410 converts the time domain acceleration signal into the time domain velocity signal, including: using fast Fourier transform to convert the time domain acceleration signal into a frequency domain acceleration signal; performing frequency domain integration on the frequency domain acceleration signal to obtain a frequency domain velocity signal; and using inverse fast Fourier transform to convert the frequency domain velocity signal into a time domain velocity signal.

[0089] In an optional case, the alignment unit 420 extracting a plurality of speed signal segments along the time domain speed signal with the rotation period as a step size includes: extracting a plurality of speed signal segments along the time domain speed signal with the rotation period as a step size without overlap.

[0090] In an optional case, the converting unit 430 converts the speed signal matrix into an image, including: multiplying the speed signal matrix by a normalization coefficient to obtain a normalized speed signal matrix; and mapping the normalized speed signal matrix to the image.

[0091] In an optional case, the converting unit 430 converting the speed signal matrix into an image includes: converting the speed signal matrix into a color image.

[0092] In an optional case, after the conversion unit 430 converts the speed signal matrix into an image, the method further includes: adjusting the size of the image to a predetermined size of the neural network model.

[0093] The implementation and specific process of the fault diagnosis device 400 can refer to the fault diagnosis method 300 and will not be repeated here.

[0094] The present invention also provides an electronic device, comprising a processor, a memory, and instructions stored in the memory, wherein the instructions implement the above method when executed by the processor.

[0095] The present invention also provides a computer-readable storage medium having computer instructions stored thereon. When the computer instructions are executed, the method according to the above is executed.

[0096] Some aspects of the methods and apparatus of the present invention may be performed entirely by hardware, entirely by software (including firmware, resident software, microcode, etc.), or by a combination of hardware and software. The above hardware or software may be referred to as "data blocks," "modules," "engines," "units," "components," or "systems." The processor may be one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DAPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), processors, controllers, microcontrollers, microprocessors, or combinations thereof. In addition, various aspects of the present invention may be embodied as computer products in one or more computer-readable media, the product including computer-readable program code. For example, computer-readable media may include, but are not limited to, magnetic storage devices (e.g., hard disks, floppy disks, tapes, etc.), optical disks (e.g., compact disks (CDs), digital versatile disks (DVDs), etc.), smart cards, and flash memory devices (e.g., cards, sticks, key drives, etc.).

[0097] A computer-readable medium may include a propagated data signal embodying computer program code, for example, in baseband or as part of a carrier wave. The propagated signal may be in a variety of forms, including electromagnetic, optical, etc., or a suitable combination thereof. A computer-readable medium may be any computer-readable medium other than a computer-readable storage medium that can be connected to an instruction execution system, apparatus, or device to communicate, propagate, or transmit the program for use. The program code on the computer-readable medium may be transmitted via any suitable medium, including radio, cable, fiber optic cable, radio frequency signal, or similar medium, or any combination of the above.

[0098] It should be understood that although this specification is described according to various embodiments, not every embodiment contains only one independent technical solution. This narrative method of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment can also be appropriately combined to form other implementation methods that can be understood by those skilled in the art.

[0099] The above description is only an illustrative embodiment of the present invention and is not intended to limit the scope of the present invention. Any equivalent changes, modifications and combinations made by those skilled in the art without departing from the concept and principle of the present invention shall fall within the scope of protection of the present invention.

Claims

1. A fault diagnosis method (200) for a rotating electrical machine, the fault diagnosis method (200) comprising: Acquiring a time domain acceleration signal of the rotating motor along a vibration direction and a rotation period of the rotating motor (210); Converting the time-domain acceleration signal into a time-domain velocity signal, extracting a plurality of velocity signal segments along the time-domain velocity signal with the rotation period as a step length, and sequentially arranging the plurality of velocity signal segments to obtain a velocity signal matrix (220); Converting the velocity signal matrix into a single image (230); The single image is input into a trained neural network model to obtain a fault diagnosis result (240).

2. The fault diagnosis method according to claim 1, characterized in that: Converting the time-domain acceleration signal into a time-domain velocity signal comprises: Converting the time domain acceleration signal into a frequency domain acceleration signal using fast Fourier transform; Performing frequency domain integration on the frequency domain acceleration signal to obtain a frequency domain velocity signal; And the frequency domain speed signal is converted into a time domain speed signal by using inverse fast Fourier transform.

3. The fault diagnosis method according to claim 1, characterized in that: Taking the rotation period as a step length and extracting a plurality of speed signal segments along the time-domain speed signal includes: taking the rotation period as a step length and extracting a plurality of speed signal segments along the time-domain speed signal without overlapping.

4. The fault diagnosis method according to any one of claims 1 to 3, characterized in that: Converting the velocity signal matrix into a single image includes: multiplying the velocity signal matrix by a normalization coefficient to obtain a normalized velocity signal matrix; and Map the normalized velocity signal matrix to a single image.

5. The fault diagnosis method according to claim 4, characterized in that: Converting the velocity signal matrix into a single image (230) includes converting the velocity signal matrix into a single color image.

6. The fault diagnosis method according to claim 1, characterized in that: After converting the speed signal matrix into a single image (230), the method further includes: adjusting the size of the single image to a predetermined size of the neural network model.

7. A fault diagnosis device (400) for a rotating electrical machine, the fault diagnosis device (400) comprising: An acquisition unit (410) acquires a time domain acceleration signal of the rotating motor along a vibration direction and a rotation period of the rotating motor; an alignment and arrangement unit (420), which converts the time-domain acceleration signal into a time-domain velocity signal, extracts a plurality of velocity signal segments along the time-domain velocity signal with the rotation period as a step length, and sequentially arranges the plurality of velocity signal segments to obtain a velocity signal matrix; A conversion unit (430) converts the velocity signal matrix into a single image; The determination unit (440) inputs the single image into a trained neural network model to obtain a fault diagnosis result.

8. The fault diagnosis device according to claim 7, characterized in that: The acquisition unit (410) converts the time domain acceleration signal into a time domain velocity signal, including: Converting the time domain acceleration signal into a frequency domain acceleration signal using fast Fourier transform; Performing frequency domain integration on the frequency domain acceleration signal to obtain a frequency domain velocity signal; and The frequency domain velocity signal is converted into a time domain velocity signal by using inverse fast Fourier transform.

9. The fault diagnosis device according to claim 7, characterized in that: The alignment arrangement unit (420) extracts a plurality of speed signal segments along the time domain speed signal with the rotation period as a step length, comprising: extracting a plurality of speed signal segments along the time domain speed signal with the rotation period as a step length without overlap.

10. The fault diagnosis device according to any one of claims 7 to 9, characterized in that: The conversion unit (430) converts the speed signal matrix into a single image, comprising: multiplying the velocity signal matrix by a normalization coefficient to obtain a normalized velocity signal matrix; and Map the normalized velocity signal matrix to a single image.

11. The fault diagnosis device according to claim 10, characterized in that: The converting unit (430) converts the speed signal matrix into a single image, including: converting the speed signal matrix into a single color image.

12. The fault diagnosis device according to claim 7, characterized in that: After the conversion unit (430) converts the speed signal matrix into a single image, it also includes: adjusting the size of the single image to a predetermined size of the neural network model.

13. An electronic device comprising a processor, a memory, and instructions stored in the memory, wherein the instructions implement the method according to any one of claims 1 to 6 when executed by the processor.

14. A computer-readable storage medium having computer instructions stored thereon, wherein the computer instructions, when executed, perform the method according to any one of claims 1 to 6.

Images