Industrial equipment fault diagnosis prediction method and system based on deep learning

Abstract

The invention discloses an industrial equipment fault diagnosis prediction method and system based on deep learning, and the method comprises the steps: S1, installing high-definition shooting equipment beside industrial equipment, carrying out the video recording of the industrial equipment in a running state, obtaining electronic image data, and dividing an electronic image into a training image set and a test image set; s2, carrying out noise reduction processing on the electronic image, and obtaining the electronic image after noise reduction; s3, performing vibration signal feature extraction on the electronic image after noise reduction to obtain vibration data; s4, carrying out vibration parameter gradient division and intelligent compensation on the collected vibration data; and S5, uploading and summarizing the collected vibration data to a cloud for data proofreading, screening out a proofreading deviation data set, and analyzing production efficiency data characteristics in the operation of the industrial equipment in the proofreading deviation data set. According to the method, the relationship between vibration and production efficiency is introduced, division is carried out according to the vibration parameter gradient, the production efficiency is compensated, industrial equipment with too large deviation of the production efficiency is screened out, the equipment is judged to be abnormal equipment, fault marking is carried out on the equipment, and workers are reminded to check the equipment.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a method and system for industrial equipment fault diagnosis and prediction based on deep learning. Background Art

[0002] Traditional factory operations mainly rely on human monitoring of equipment status and parameters. Operators monitor the constantly changing data 24 hours a day. It is difficult to detect the slow changes in parameters in the early stages of equipment failure, and they can only handle it after the equipment fails. Operators are under great work pressure and high labor intensity.

[0003] Industrial equipment fault inspection requires extremely high professional skills. Only excellent operators can detect equipment anomalies among thousands of data and make correct handling in the early stages of failure. This is rather troublesome and difficult to meet the needs of staff. Summary of the invention

[0004] In order to solve the above technical problems, a method and system for industrial equipment fault diagnosis and prediction based on deep learning are provided. This technical solution solves the problem mentioned in the above background technology that industrial equipment fault inspection requires extremely high professional skills from people. Only excellent operators can perceive equipment anomalies among thousands of data and make correct disposal in the early stage of failure, which is a rather troublesome problem.

[0005] In order to achieve the above purpose, the technical solution adopted by the present invention is:

[0006] A method for industrial equipment fault diagnosis and prediction based on deep learning, comprising the steps of:

[0007] S1, install high-definition camera equipment next to the industrial equipment, record the industrial equipment in operation, obtain electronic image data, and divide the electronic images into a training image set and a test image set;

[0008] S2, performing noise reduction processing on the electronic image to obtain a noise-reduced electronic image;

[0009] S3, extracting vibration signal features from the electronic image after noise reduction to obtain vibration data;

[0010] S4, performing vibration parameter gradient division and intelligent compensation on the collected vibration data;

[0011] S5, uploading and aggregating the collected vibration data to the cloud for data proofreading, screening out a proofreading deviation data group, and analyzing the production efficiency data characteristics of the industrial equipment in the proofreading deviation data group; and

[0012] S6, identifying and marking expected faults of industrial equipment with abnormal data characteristics in the calibration deviation data group.

[0013] Preferably, the step S1 of dividing the electronic images into a training image set and a test image set specifically includes the following steps:

[0014] S11, obtaining red, green and blue color components of each pixel in the electronic image;

[0015] S13, obtaining weighted coefficients of human eye sensitivity of red, green and blue;

[0016] S15, calculating the grayscale value of each pixel of the electronic image according to the grayscale value formula, wherein the grayscale value range is [0, 255];

[0017] S17, replacing each pixel in the electronic image with the grayscale value of each pixel; and

[0018] S19, all electronic images corresponding to the industrial equipment in operation status in the database are equally divided into two parts, which are set as a training image set and a test image set.

[0019] Preferably, the gray value formula is:

[0020] Gray(i,j)=ε1R(i,j)+ε2G(i,j)+ε3B(i,j),

[0021] Where Gray(i, j) is the grayscale of the pixel in the i-th row and j-th column of the electronic image, ε1, ε2, and ε3 are the weighted coefficients of the human eye sensitivity to red, green, and blue, respectively, R(i, j), G(i, j), and B(i, j) are the red, green, and blue components of the pixel in the i-th row and j-th column of the electronic image, respectively, and i and j are both positive integers.

[0022] Preferably, the step 2 of performing noise reduction processing on the electronic image to obtain the electronic image after noise reduction specifically comprises the following steps:

[0023] S21, obtaining the number of pixels M*N of the electronic image in the training image set;

[0024] S22, use element 0 to fill the blank part of the 2M*2N rectangle;

[0025] S23, using the centralization transformation formula to calculate the centralization function corresponding to the 2M*2N rectangle;

[0026] S24, performing discrete Fourier transform on the centralization function corresponding to the 2M*2N rectangle to obtain an intermediate function;

[0027] S25, obtaining an ideal low-pass filter function;

[0028] S26, using a discrete Fourier transform inverse transform formula to calculate an output function; and

[0029] S27, intercepting M*N pixels in the upper left corner of the image corresponding to the output function, and outputting them as a denoised electronic image.

[0030] Preferably, the centralization transformation formula is:

[0031] h(x,y)=(-1) x+y f(x, y),

[0032] Where, h(x, y) is the centering function corresponding to the 2M*2N rectangle, f(x, y) is the input function corresponding to the 2M*2N rectangle, and (x, y) is the coordinate of the pixel point in the 2M*2N rectangle;

[0033] The intermediate function is:

[0034]

[0035] Where F(u, v) is the intermediate function, j is the imaginary unit, and (u, v) is the frequency domain coordinate.

[0036] Preferably, the ideal low-pass filter function is:

[0037]

[0038] Where H(u, v) is the ideal low-pass filter function, d(u, v) is the distance from point (u, v) to the center point of the frequency domain, and r is the preset frequency domain radius;

[0039] The output function is:

[0040]

[0041] Where f′(x, y) is the output function.

[0042] Preferably, the step S3, extracting vibration signal features from the electronic image after noise reduction to obtain vibration data, specifically comprises the following steps:

[0043] S31, decomposing the electronic image by frame and converting it into the frequency domain;

[0044] S33, performing power spectrum analysis on the frequency domain image to obtain a preliminary positioning vibration area;

[0045] S35, determining the vibration direction by performing multi-directional frequency filtering on the time domain signals superimposed on the frequency domain images in different frames and comparing the vibration intensities of the signals;

[0046] S37, filtering the preliminary vibration region of the frequency domain image according to the vibration direction, and comparing the amplitude intensities in the filtered frequency domain and spatial domain to obtain the precise vibration region; and

[0047] S39, performing power spectrum analysis in the time domain on the phase values ​​of points in the precise vibration area on the frequency domain image in the precise vibration area to obtain final vibration data.

[0048] Preferably, the vibration parameter gradient division and intelligent compensation of the collected vibration data specifically include the following steps:

[0049] Obtain the historical production efficiency data of the industrial equipment corresponding to each set of gradient vibration parameter combinations in the set time window, and exhaustively enumerate the compensation strategies for the historical production efficiency data of all industrial equipment, wherein the same compensation strategy is adopted for the historical production efficiency data in the same gradient vibration parameter combination, and the strategies of the historical production efficiency data in different gradient vibration parameter combinations are fully arranged, and a comprehensive compensation calculation is performed for each strategy arrangement and combination;

[0050] Bring the vibration parameters into the compensation strategy algorithm for comprehensive compensation calculation. The specific process is as follows:

[0051] Label the selected k-th strategy as s k , assuming there are c strategies and m sets of gradient vibration parameter combinations, then there are c m There are 3 kinds of strategy combinations, among which the strategy selected by the vth group of combinations is c with an initial value of zero. m * In the m matrix, it is represented by the c-base value corresponding to v. Based on this representation method, the specific compensation expression of the v-th group of strategy combinations is constructed:

[0052]

[0053] In the formula, S v is the total value of current production efficiency data compensation, is the compensation strategy algorithm function corresponding to the i-th group of gradient vibration parameter combination, a i is the value of the i-th group of gradient vibration parameter combination, w i is the number of industrial equipment settings to be compensated corresponding to the i-th group of gradient vibration parameter combinations;

[0054] Calculate the absolute value of the deviation between the historical production efficiency data after compensation and the total production efficiency, select the strategy combination with the smallest absolute value of the deviation, mark it as the optimal strategy combination, and establish the association relationship between the gradient vibration parameter combination and the optimal compensation strategy based on the gradient vibration parameter combination and the optimal strategy combination;

[0055] According to the correlation between the gradient vibration parameter combination and the optimal compensation strategy, the corresponding compensation strategy algorithm is selected to compensate the real-time data and then upload it to the cloud.

[0056] Preferably, the step S6, performing expected fault identification and marking on industrial equipment with abnormal data features in the calibration deviation data group, specifically comprises the following steps:

[0057] S61, taking a single industrial device as a calculation unit, performing Fourier transformation on the data collected by each industrial device in the calibration interval in the calibration deviation data group, calculating the spectrum of the uploaded data of each industrial device, performing average processing on the spectrum values ​​of the uploaded data of all industrial devices at the same frequency, and establishing a comparison spectrum;

[0058] S63, obtaining a control spectrum of the calibration deviation data group, converting the uploaded data spectrum and the control spectrum corresponding to all industrial equipment in the calibration deviation data group into vectors to calculate the Euclidean distance, performing Sigmoid function and inverse conversion on the calculation result, and obtaining the expected failure coefficient of the industrial equipment; and

[0059] S65, setting an expected failure coefficient threshold, marking abnormal failure risks for industrial equipment whose expected failure coefficient is greater than or equal to the expected failure coefficient threshold, and issuing an early warning.

[0060] A deep learning-based industrial equipment fault diagnosis and prediction system, comprising:

[0061] An acquisition module, wherein the acquisition module is used to install a high-definition camera next to the industrial equipment, record the industrial equipment in operation, obtain electronic image data, and divide the electronic images into a training image set and a test image set;

[0062] A data processing module, wherein the data processing module is used to perform noise reduction processing on the electronic image to obtain the electronic image after noise reduction;

[0063] A feature extraction module, wherein the feature extraction module is used to extract vibration signal features from the electronic image after noise reduction to obtain vibration data;

[0064] A screening module, which is used to perform vibration parameter gradient division and intelligent compensation on the collected vibration data, upload and aggregate the collected vibration data to the cloud for data proofreading, screen out proofreading deviation data groups, and analyze production efficiency data characteristics of industrial equipment in operation in the proofreading deviation data groups; and

[0065] A fault identification module is used to identify and mark expected faults of industrial equipment with abnormal data features in the calibration deviation data group.

[0066] Compared with the prior art, the present invention provides an industrial equipment fault diagnosis and prediction method and system based on deep learning, which has the following beneficial effects:

[0067] The present invention performs video monitoring of industrial equipment in operation, adopts deep learning to divide electronic images into a training image set and a test image set, performs noise reduction processing on the electronic images, and extracts vibration signal features in the images, thereby obtaining vibration data. Since the vibration of industrial equipment is closely related to production efficiency, the present invention introduces the relationship between vibration and production efficiency, divides them according to the vibration parameter gradient and compensates for the production efficiency, screens out industrial equipment with excessive deviation in production efficiency, determines the equipment as abnormal equipment, marks its fault, and reminds the staff to check it, thereby meeting the needs of the staff. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 It is a schematic diagram of the fault diagnosis and prediction method of the present invention;

[0069] Figure 2 A schematic diagram of a method for dividing electronic images into a training image set and a test image set in the present invention;

[0070] Figure 3 A schematic diagram of a method for obtaining a noise-reduced electronic image in the present invention;

[0071] Figure 4 A schematic diagram of a method for obtaining vibration data in the present invention;

[0072] Figure 5 It is a schematic diagram of a method for identifying and marking expected faults of industrial equipment with abnormal data features in a calibration deviation data group in the present invention.

[0073] Figure 6 It is a block diagram of the industrial equipment fault diagnosis and prediction system based on deep learning in the present invention. DETAILED DESCRIPTION

[0074] The following description is used to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments described below are only examples, and those skilled in the art may think of other obvious variations.

[0075] Please refer to Figure 1-Figure 5 As shown, a method for industrial equipment fault diagnosis and prediction based on deep learning includes the following steps:

[0076] S1, install high-definition camera equipment next to the industrial equipment, record the industrial equipment in operation, obtain electronic image data, and divide the electronic images into a training image set and a test image set;

[0077] S2, performing noise reduction processing on the electronic image to obtain a noise-reduced electronic image;

[0078] S3, extracting vibration signal features from the electronic image after noise reduction to obtain vibration data;

[0079] S4, performing vibration parameter gradient division and intelligent compensation on the collected vibration data;

[0080] S5, uploading and aggregating the collected vibration data to the cloud for data proofreading, screening out a proofreading deviation data group, and analyzing the production efficiency data characteristics of the industrial equipment in the proofreading deviation data group; and

[0081] S6, identifying and marking expected faults of industrial equipment with abnormal data characteristics in the calibration deviation data group.

[0082] The present invention performs video monitoring on industrial equipment in operation, uses deep learning to divide electronic images into training image sets and test image sets, performs noise reduction processing on electronic images, and extracts vibration signal features in images to obtain vibration data. Since the vibration of industrial equipment is closely related to production efficiency, excessive vibration will lead to increased equipment wear and increased failure frequency, thereby reducing normal operation time. In addition, vibration may affect product quality, lead to increased defective rate, and increase waste and rework costs. Therefore, keeping the equipment within an appropriate vibration range helps to improve production efficiency, extend equipment life, and reduce maintenance costs. Therefore, the present invention introduces the relationship between vibration and production efficiency, divides according to the vibration parameter gradient and compensates for production efficiency, screens out industrial equipment with excessive deviation in production efficiency, determines the equipment as abnormal equipment, marks its fault, and reminds the staff to check it, meeting the needs of the staff.

[0083] The step S1 of dividing the electronic images into a training image set and a test image set specifically includes the following steps:

[0084] S11, obtaining red, green and blue color components of each pixel in the electronic image;

[0085] S13, obtaining weighted coefficients of human eye sensitivity of red, green and blue;

[0086] S15, calculating the grayscale value of each pixel of the electronic image according to the grayscale value formula, and the grayscale value range is [0, 255];

[0087] S17, replacing each pixel in the electronic image with the grayscale value of each pixel; and

[0088] S19, all electronic images corresponding to the industrial equipment in operation status in the database are equally divided into two parts, which are set as a training image set and a test image set.

[0089] The pixels in an electronic image are composed of red, green, and blue. The grayscale value of each pixel is used to replace each pixel in the electronic image so that it is in the same standard and is divided into a training image set and a test image set to facilitate image analysis and research.

[0090] The gray value formula is:

[0091] Gray(i,j)=ε1R(i,j)+ε2G(i,j)+ε3B(i,j),

[0092] Where Gray(i, j) is the grayscale of the pixel in the i-th row and j-th column of the electronic image, ε1, ε2, and ε3 are the weighted coefficients of the human eye sensitivity to red, green, and blue, respectively, R(i, j), G(i, j), and B(i, j) are the red, green, and blue components of the pixel in the i-th row and j-th column of the electronic image, respectively, and i and j are both positive integers.

[0093] The above formula can determine the grayscale of each pixel in the electronic image based on the red, green and blue components of each pixel in the electronic image and the weighted coefficient of the sensitivity of the human eye to the red, green and blue colors. Step S2, performing noise reduction processing on the electronic image, and obtaining the electronic image after noise reduction specifically includes the following steps:

[0094] S21, obtaining the number of pixels M*N of the electronic image in the training image set;

[0095] S22, use element 0 to fill the blank part of the 2M*2N rectangle;

[0096] S23, using the centralization transformation formula to calculate the centralization function corresponding to the 2M*2N rectangle;

[0097] S24, performing discrete Fourier transform on the centralization function corresponding to the 2M*2N rectangle to obtain an intermediate function;

[0098] S25, obtaining an ideal low-pass filter function;

[0099] S26, using a discrete Fourier transform inverse transform formula to calculate an output function; and

[0100] S27, intercepting M*N pixels in the upper left corner of the image corresponding to the output function, and outputting them as a denoised electronic image.

[0101] The centralization transformation formula is:

[0102] h(x,y)=(-1) x+y f(x, y),

[0103] Where, h(x, y) is the centering function corresponding to the 2M*2N rectangle, f(x, y) is the input function corresponding to the 2M*2N rectangle, and (x, y) is the coordinate of the pixel point in the 2M*2N rectangle;

[0104] The intermediate function is:

[0105]

[0106] Where F(u, v) is the intermediate function, j is the imaginary unit, and (u, v) is the frequency domain coordinate.

[0107] The ideal low-pass filter function is:

[0108]

[0109] Where H(u, v) is the ideal low-pass filter function, d(u, v) is the distance from point (u, v) to the center point of the frequency domain, and r is the preset frequency domain radius;

[0110] The output function is:

[0111]

[0112] Where f′(x, y) is the output function.

[0113] Specifically, NumPy and SciPy libraries are used to complete image processing tasks. The code includes the steps of obtaining the number of pixels from the image, creating and centering a 2M*2N matrix, calculating the discrete Fourier transform, applying an ideal low-pass filter, performing an inverse transform, and intercepting the final image. The following is the python code framework:

[0114] import numpy as np

[0115] from scipy.fftpack import fft2,ifft2

[0116] import matplotlib.pyplot as plt

[0117] from scipy import ndimage

[0118] #Assuming there is already a training image image, you can use matplotlib's imread to read it

[0119] #image=plt.imread('path_to_your_image.jpg')

[0120] #For demonstration, we create a simple random image

[0121] image = np.random.randint(0,256,(100,100),dtype = np.uint8) # Example image 100x100;

[0122] #Get the number of pixels M*N of the electronic images in the training image set

[0123] M*N=image.shape

[0124] #Use element 0 to fill the blank part of the 2M*2N rectangle

[0125] padded_image=np.zeros((2*M,2*N),dtype=np.uint8)

[0126] padded_image[M / / 2:M / / 2+M,N / / 2:N / / 2+N]=image

[0127] #Use the centralization transformation formula to calculate the centralization function corresponding to the 2M*2N rectangle

[0128] centered_image=np.fft.fftshift(padded_image.astype(np.float32),axes=(0,1))

[0129] #Perform discrete Fourier transform on the centralization function corresponding to the 2M*2N rectangle to obtain the intermediate function

[0130] f_image=fft2(centered_image)

[0131] #Get the ideal low-pass filter function, assuming the cutoff frequency is D0

[0132] D0=30#cutoff frequency, can be adjusted

[0133] [u,v]=np.meshgrid(np.arange(-N,N),np.arange(-M,M))

[0134] D = np.sqrt((uN)**2+(vM)**2)

[0135] H=D<=D0

[0136] H = H.astype(np.float32)

[0137] #Use the discrete Fourier transform inverse transform formula to calculate the output function

[0138] F_filtered=f_image*H

[0139] output_image=np.real(ifft2(F_filtered))

[0140] #Centralization Back

[0141] output_image=np.fft.ifftshift(output_image,axes=(0,1))

[0142] #Intercept the M*N pixels in the upper left corner of the image corresponding to the output function and output it as the electronic image after noise reduction

[0143] final_image=np.clip(output_image[M / / 2:M / / 2+M,N / / 2:N / / 2+N].astype(np.uint8),0,255)

[0144] # Display the original image and the denoised image

[0145] plt.figure(figsize=(12,6))

[0146] plt.subplot(1,2,1)

[0147] plt.title('Original Image')

[0148] plt.imshow(image,cmap='gray')

[0149] plt.axis('off')

[0150] plt.subplot(1,2,2)

[0151] plt.title('Denoised Image')

[0152] plt.imshow(final_image,cmap='gray')

[0153] plt.axis('off')

[0154] plt.show().

[0155] Step S3, extracting vibration signal features from the electronic image after noise reduction to obtain vibration data, specifically includes the following steps:

[0156] S31, decomposing the electronic image by frame and converting it into the frequency domain;

[0157] S33, performing power spectrum analysis on the frequency domain image to obtain a preliminary positioning vibration area;

[0158] S35, determining the vibration direction by performing multi-directional frequency filtering on the time domain signals superimposed on the frequency domain images in different frames and comparing the vibration intensities of the signals;

[0159] S37, filtering the preliminary vibration region of the frequency domain image according to the vibration direction, and comparing the amplitude intensities in the filtered frequency domain and spatial domain to obtain the precise vibration region; and

[0160] S39, performing power spectrum analysis in the time domain on the phase values ​​of points in the precise vibration area on the frequency domain image in the precise vibration area to obtain final vibration data.

[0161] Specifically, in order to achieve the above steps, we need to perform a series of processing on the electronic image, including frame decomposition, frequency domain conversion, power spectrum analysis, time domain signal superposition, multi-directional frequency filtering, vibration direction determination, precise vibration area positioning and final vibration data analysis. The following is a simplified Python code framework:

[0162]

[0163]

[0164]

[0165] Step S4, performing vibration parameter gradient division and intelligent compensation on the collected vibration data, specifically includes the following steps:

[0166] Obtain the historical production efficiency data of the industrial equipment corresponding to each set of gradient vibration parameter combinations in the set time window, and exhaustively enumerate the compensation strategies for the historical production efficiency data of all industrial equipment, wherein the same compensation strategy is adopted for the historical production efficiency data in the same gradient vibration parameter combination, and the strategies of the historical production efficiency data in different gradient vibration parameter combinations are fully arranged, and a comprehensive compensation calculation is performed for each strategy arrangement and combination;

[0167] Bring the vibration parameters into the compensation strategy algorithm for comprehensive compensation calculation. The specific process is as follows:

[0168] Label the selected k-th strategy as s k , assuming there are c strategies and m sets of gradient vibration parameter combinations, then there are c m There are 3 kinds of strategy combinations, among which the strategy selected by the vth group of combinations is c with an initial value of zero. m * In the m matrix, it is represented by the c-base value corresponding to v. Based on this representation method, the specific compensation expression of the v-th group of strategy combinations is constructed:

[0169]

[0170] In the formula, S v is the total value of current production efficiency data compensation, is the compensation strategy algorithm function corresponding to the i-th group of gradient vibration parameter combination, a i is the value of the i-th group of gradient vibration parameter combination, w i is the number of industrial equipment settings to be compensated corresponding to the i-th group of gradient vibration parameter combinations;

[0171] Calculate the absolute value of the deviation between the historical production efficiency data after compensation and the total production efficiency, select the strategy combination with the smallest absolute value of the deviation, mark it as the optimal strategy combination, and establish the association relationship between the gradient vibration parameter combination and the optimal compensation strategy based on the gradient vibration parameter combination and the optimal strategy combination;

[0172] According to the correlation between the gradient vibration parameter combination and the optimal compensation strategy, the corresponding compensation strategy algorithm is selected to compensate the real-time data and then upload it to the cloud.

[0173] The production efficiency is divided according to the vibration parameter gradient and compensated, and the industrial equipment with excessive deviation in production efficiency is screened out. The equipment is marked as abnormal equipment, and the fault is reminded to check it, thus meeting the needs of the staff.

[0174] The step S5 of uploading and aggregating the collected vibration data to the cloud for data proofreading and screening out the proofreading deviation data group specifically includes the following steps:

[0175] Calculate the deviation ratio between the production efficiency data set by the industrial equipment and the cumulative value of the real-time uploaded data, set the deviation ratio threshold, filter out the data with a deviation ratio greater than or equal to the set deviation ratio threshold, and mark them as the proofreading deviation data group.

[0176] The step S6, identifying and marking expected faults of industrial equipment with abnormal data features in the calibration deviation data group, specifically includes the following steps:

[0177] S61, taking a single industrial device as a calculation unit, performing Fourier transformation on the data collected by each industrial device in the calibration interval in the calibration deviation data group, calculating the spectrum of the uploaded data of each industrial device, performing average processing on the spectrum values ​​of the uploaded data of all industrial devices at the same frequency, and establishing a comparison spectrum;

[0178] S63, obtaining a control spectrum of the calibration deviation data group, converting the uploaded data spectrum and the control spectrum corresponding to all industrial equipment in the calibration deviation data group into vectors to calculate the Euclidean distance, performing Sigmoid function and inverse conversion on the calculation result, and obtaining the expected failure coefficient of the industrial equipment; and

[0179] S65, setting an expected failure coefficient threshold, marking abnormal failure risks for industrial equipment whose expected failure coefficient is greater than or equal to the expected failure coefficient threshold, and issuing an early warning.

[0180] Specifically, the following is a simplified Python code skeleton:

[0181] import numpy as np

[0182] from scipy.fftpack import fft

[0183] from sklearn.preprocessing import MinMaxScaler

[0184] #Assume the following data:

[0185] #device_efficiency_data: dictionary, the key is the device ID, the value is the set production efficiency data list

[0186] #real_time_data: dictionary, the key is the device ID, the value is the list of real-time uploaded data (corresponding to the time point of the set data)

[0187]

[0188]

[0189]

[0190] The present application also discloses a deep learning-based industrial equipment fault diagnosis and prediction system, comprising:

[0191] The acquisition module 10 is used to install a high-definition camera next to the industrial equipment, record the industrial equipment in operation, obtain electronic image data, and divide the electronic images into a training image set and a test image set;

[0192] A data processing module 30, wherein the data processing module 30 is used to perform noise reduction processing on the electronic image to obtain the electronic image after noise reduction;

[0193] A feature extraction module 50, wherein the feature extraction module 50 is used to extract vibration signal features from the electronic image after noise reduction to obtain vibration data;

[0194] A screening module 70, which is used to perform vibration parameter gradient division and intelligent compensation on the collected vibration data, upload and aggregate the collected vibration data to the cloud for data proofreading, screen out proofreading deviation data groups, and analyze production efficiency data characteristics of industrial equipment in operation in the proofreading deviation data groups; and

[0195] The fault identification module 90 is used to identify and mark expected faults of industrial equipment with abnormal data features in the calibration deviation data group.

[0196] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions only describe the principles of the present invention. The present invention may be subject to various changes and improvements without departing from the spirit and scope of the present invention. These changes and improvements fall within the scope of the present invention. The scope of protection claimed by the present invention is defined by the attached claims and their equivalents.

Claims

1. A method for industrial equipment fault diagnosis and prediction based on deep learning, characterized in that: Includes steps: S1, install high-definition camera equipment next to the industrial equipment, record the industrial equipment in operation, obtain electronic image data, and divide the electronic images into a training image set and a test image set; S2, performing noise reduction processing on the electronic image to obtain a noise-reduced electronic image; S3, extracting vibration signal features from the electronic image after noise reduction to obtain vibration data; S4, performing vibration parameter gradient division and intelligent compensation on the collected vibration data; S5, uploading and aggregating the collected vibration data to the cloud for data proofreading, screening out the proofreading deviation data group, and analyzing the production efficiency data characteristics of the industrial equipment in the proofreading deviation data group; as well as S6, identifying and marking expected faults of industrial equipment with abnormal data characteristics in the calibration deviation data group.

2. The method for industrial equipment fault diagnosis and prediction based on deep learning according to claim 1, characterized in that: The step S1 of dividing the electronic images into a training image set and a test image set specifically includes the following steps: S11, obtaining red, green and blue color components of each pixel in the electronic image; S13, obtaining weighted coefficients of human eye sensitivity of red, green and blue; S15, calculating the grayscale value of each pixel of the electronic image according to the grayscale value formula, wherein the grayscale value range is [0, 255]; S17, replacing each pixel in the electronic image with the grayscale value of each pixel; as well as S19, all electronic images corresponding to the industrial equipment in operation status in the database are equally divided into two parts, which are set as a training image set and a test image set.

3. The method for industrial equipment fault diagnosis and prediction based on deep learning according to claim 2 is characterized in that: The gray value formula is: Gray(i,j)=ε1R(i,j)+ε2G(i,j)+ε3B(i,j), Where Gray(i, j) is the grayscale of the pixel in the i-th row and j-th column of the electronic image, ε1, ε2, and ε3 are the weighted coefficients of the human eye sensitivity to red, green, and blue, respectively, R(i, j), G(i, j), and B(i, j) are the red, green, and blue components of the pixel in the i-th row and j-th column of the electronic image, respectively, and i and j are both positive integers.

4. The method for industrial equipment fault diagnosis and prediction based on deep learning according to claim 3 is characterized in that: The step 2, performing noise reduction processing on the electronic image to obtain the electronic image after noise reduction, specifically includes the following steps: S21, obtaining the number of pixels M*N of the electronic image in the training image set; S22, use element 0 to fill the blank part of the 2M*2N rectangle; S23, using the centralization transformation formula to calculate the centralization function corresponding to the 2M*2N rectangle; S24, performing discrete Fourier transform on the centralization function corresponding to the 2M*2N rectangle to obtain an intermediate function; S25, obtaining an ideal low-pass filter function; S26, using the discrete Fourier transform inverse transform formula, calculate the output function; as well as S27, intercepting M*N pixels in the upper left corner of the image corresponding to the output function, and outputting them as a denoised electronic image.

5. The method for industrial equipment fault diagnosis and prediction based on deep learning according to claim 4 is characterized in that: The centralization transformation formula is: h(x,y)=(-1) x+y f(x,y), Where, h(x, y) is the centering function corresponding to the 2M*2N rectangle, f9x, y) is the input function corresponding to the 2M*2N rectangle, and (x, y) is the coordinate of the pixel point in the 2M*2N rectangle; The intermediate function is: Where F(u, v) is the intermediate function, j is the imaginary unit, and (u, v) is the frequency domain coordinate.

6. The method for industrial equipment fault diagnosis and prediction based on deep learning according to claim 4 is characterized in that: The ideal low-pass filter function is: Where H(u, v) is the ideal low-pass filter function, d(u, v) is the distance from point (u, v) to the center point of the frequency domain, and r is the preset frequency domain radius; The output function is: Where f′(x, y) is the output function.

7. The method for industrial equipment fault diagnosis and prediction based on deep learning according to claim 6 is characterized in that: The step S3, extracting vibration signal features from the electronic image after noise reduction to obtain vibration data, specifically includes the following steps: S31, decomposing the electronic image by frame and converting it into the frequency domain; S33, performing power spectrum analysis on the frequency domain image to obtain a preliminary positioning vibration area; S35, determining the vibration direction by performing multi-directional frequency filtering on the time domain signals superimposed on the frequency domain images in different frames and comparing the vibration intensities of the signals; S37, filtering the preliminary vibration region of the frequency domain image according to the vibration direction, and comparing the amplitude intensities in the filtered frequency domain and spatial domain to obtain the precise vibration region; and S39, performing power spectrum analysis in the time domain on the phase values ​​of points in the precise vibration area on the frequency domain image in the precise vibration area to obtain final vibration data.

8. The method for industrial equipment fault diagnosis and prediction based on deep learning according to claim 7 is characterized in that: The vibration parameter gradient division and intelligent compensation of the collected vibration data specifically include the following steps: Obtain the historical production efficiency data of the industrial equipment corresponding to each set of gradient vibration parameter combinations in the set time window, and exhaustively enumerate the compensation strategies for the historical production efficiency data of all industrial equipment, wherein the same compensation strategy is adopted for the historical production efficiency data in the same gradient vibration parameter combination, and the strategies of the historical production efficiency data in different gradient vibration parameter combinations are fully arranged, and a comprehensive compensation calculation is performed for each strategy arrangement and combination; Bring the vibration parameters into the compensation strategy algorithm for comprehensive compensation calculation. The specific process is as follows: Label the selected k-th strategy as s k , assuming there are c strategies and m sets of gradient vibration parameter combinations, then there are c m There are 3 kinds of strategy combinations, among which the strategy selected by the vth group of combinations is c with an initial value of zero. m * In the m matrix, it is represented by the c-base value corresponding to v. Based on this representation method, the specific compensation expression of the v-th group of strategy combinations is constructed: In the formula, S v is the total value of current production efficiency data compensation, is the compensation strategy algorithm function corresponding to the i-th group of gradient vibration parameter combination, a i is the value of the i-th group of gradient vibration parameter combination, w i is the number of industrial equipment settings to be compensated corresponding to the i-th group of gradient vibration parameter combinations; Calculate the absolute value of the deviation between the historical production efficiency data after compensation and the total production efficiency, select the strategy combination with the smallest absolute value of the deviation, mark it as the optimal strategy combination, and establish the association relationship between the gradient vibration parameter combination and the optimal compensation strategy based on the gradient vibration parameter combination and the optimal strategy combination; According to the correlation between the gradient vibration parameter combination and the optimal compensation strategy, the corresponding compensation strategy algorithm is selected to compensate the real-time data and then upload it to the cloud.

9. The method for industrial equipment fault diagnosis and prediction based on deep learning according to claim 8, characterized in that: The step S6, identifying and marking expected faults of industrial equipment with abnormal data features in the calibration deviation data group, specifically includes the following steps: S61, taking a single industrial device as a calculation unit, performing Fourier transformation on the data collected by each industrial device in the calibration interval in the calibration deviation data group, calculating the spectrum of the uploaded data of each industrial device, performing average processing on the spectrum values ​​of the uploaded data of all industrial devices at the same frequency, and establishing a comparison spectrum; S63, obtaining a control spectrum of the calibration deviation data group, converting the uploaded data spectrum corresponding to all industrial equipment in the calibration deviation data group and the control spectrum into vectors to calculate the Euclidean distance, performing Sigmoid function and inverse conversion on the calculation result, and obtaining the expected failure coefficient of the industrial equipment; as well as S65, setting an expected failure coefficient threshold, marking abnormal failure risks for industrial equipment whose expected failure coefficient is greater than or equal to the expected failure coefficient threshold, and issuing a warning.

10. A deep learning-based industrial equipment fault diagnosis and prediction system, used to implement a deep learning-based industrial equipment fault diagnosis and prediction method as described in any one of claims 1 to 9, characterized in that: include: A collection module (10), wherein the collection module (10) is used to install a high-definition camera next to the industrial equipment, record the industrial equipment in operation, obtain electronic image data, and divide the electronic images into a training image set and a test image set; A data processing module (30), the data processing module (30) being used to perform noise reduction processing on the electronic image to obtain the electronic image after noise reduction; A feature extraction module (50), the feature extraction module (50) being used to extract vibration signal features from the electronic image after noise reduction to obtain vibration data; A screening module (70), the screening module (70) being used to perform vibration parameter gradient division and intelligent compensation on the collected vibration data, upload and aggregate the collected vibration data to the cloud for data proofreading, screen out a proofreading deviation data group, and analyze production efficiency data characteristics of industrial equipment in operation in the proofreading deviation data group; as well as A fault identification module (90) is used to identify and mark expected faults of industrial equipment with abnormal data features in a calibration deviation data group.

Images