Methods, equipment and storage media for detecting wear condition of automotive bearings

Abstract

This invention belongs to the field of equipment maintenance technology, specifically relating to a method, equipment, and storage medium for detecting the wear condition of automotive bearings. The method involves acquiring bearing images as a basic dataset; preprocessing the basic dataset; expanding the basic dataset to construct a target dataset; establishing an image mapping matrix based on the target dataset to obtain abnormal bearing images; and performing defect detection on the abnormal bearing images using a defect location mapping matrix to obtain defect detection results. This invention solves the problem of model training's dependence on samples, as well as the problems of algorithm robustness and generalization ability. The algorithm achieves good recognition accuracy even with a small number of samples, and can significantly improve detection accuracy when applied to bearing wear detection scenarios.

Description

Technical Field

[0001] This invention belongs to the field of equipment maintenance technology, specifically relating to a method, equipment, and storage medium for detecting the wear condition of automotive bearings. Background Technology

[0002] Bearings are one of the core basic components of automobiles, and the bearing industry is a fundamental and strategic industry of the nation, playing a vital supporting role in national economic development and national defense. After decades of continuous and rapid development, my country's bearing industry has formed an independent and complete industrial system, becoming the world's third-largest bearing producer in terms of sales and output, with an annual output of nearly 20 billion sets. Therefore, finding a good and fast testing method is particularly important in the bearing testing industry.

[0003] With the widespread adoption of artificial intelligence technology, machine vision systems using cameras as sensors have emerged in large numbers and have been widely applied in fields such as industrial inspection, packaging and printing, and the food industry. Domestic bearing manufacturers already have machine vision-based inspection methods that can quickly inspect bearings, but current inspection methods suffer from a dependence on samples; when training samples are limited, existing algorithms have issues with robustness and generalization ability, resulting in low accuracy in identifying bearing wear conditions. Summary of the Invention

[0004] To overcome the shortcomings of existing technologies, this invention provides a method, device, and storage medium for detecting the wear condition of automotive bearings, thereby addressing the problem of existing detection methods' dependence on samples; and the issues of robustness and generalization ability of existing algorithms when training samples are limited, resulting in low accuracy in identifying bearing wear conditions.

[0005] One aspect of this invention provides a method for detecting the wear condition of automotive bearings, comprising:

[0006] Bearing images were collected as the basic dataset;

[0007] Preprocess the aforementioned basic dataset;

[0008] The base dataset is expanded to construct the target dataset;

[0009] An image mapping matrix is ​​established based on the target dataset to obtain images of abnormal bearings;

[0010] Defect detection results are obtained by performing defect location mapping matrix on abnormal bearing images.

[0011] In one preferred embodiment of the present invention, the preprocessing of the basic dataset includes:

[0012] Perform two-dimensional wavelet multi-scale decomposition on the aforementioned basic dataset;

[0013] Denoising is performed on the base dataset at various scales;

[0014] Perform inverse wavelet transform on the aforementioned basic dataset.

[0015] In one preferred embodiment of the present invention, the expansion processing of the basic dataset includes:

[0016] The images of the base dataset are copied by rotation within a predetermined angle range to obtain the expanded dataset;

[0017] The basic dataset is divided into training set, validation set and test set according to a predetermined ratio and output as target dataset.

[0018] In one preferred embodiment of the present invention, establishing the image mapping matrix includes:

[0019] Convert the images in the target dataset to grayscale;

[0020] The gray values ​​of each row of pixels in the image are added together to calculate and obtain the first image mapping matrix;

[0021] Calculate the maximum value, minimum value, average value, and standard deviation of the first image mapping matrix;

[0022] The first image mapping matrix is ​​transformed to obtain the second image mapping matrix.

[0023] In one preferred embodiment of the present invention, the step of establishing the image mapping matrix based on the target dataset further includes:

[0024] Check whether the coefficients of change in the second image mapping matrix are constants;

[0025] If the coefficients are constants, then the second image mapping matrix is ​​determined to be a linear transformation of the first image mapping matrix, and the bearing has no defects.

[0026] If the coefficient is an uncertain value, then the second image mapping matrix is ​​determined to be a nonlinear transformation of the second image mapping matrix, indicating that the bearing has a defect.

[0027] In one preferred embodiment of the present invention, when a bearing has a defect, the defect detection of the abnormal bearing image through the defect location mapping matrix includes:

[0028] The detection threshold parameters are determined based on the mean and standard deviation of the image mapping matrix.

[0029] Based on the detection threshold parameter and the second image mapping matrix, a defect location mapping matrix is ​​constructed to detect the abnormal parts of the second image mapping matrix and output a first detection result.

[0030] Obtain the position where the grayscale value is 255 in the first detection result;

[0031] The location of a bearing defect is output when the grayscale value is continuously between 255 and a predetermined threshold.

[0032] In one preferred embodiment of the present invention, after obtaining the bearing defect location, calibration is performed using a convolutional neural network, specifically including:

[0033] Input historical bearing images to train the machine learning model and obtain a trained convolutional neural network;

[0034] The bearing image corresponding to the bearing defect location is imported into the convolutional neural network, and a second detection result is output.

[0035] Based on the second detection result, the location of the bearing defect is calibrated, and a bearing wear status report is output.

[0036] In one preferred embodiment of the present invention, the acquisition of bearing images as the basic dataset specifically includes:

[0037] The bearing sample was placed in a black chamber;

[0038] The bearing sample is rotated at a predetermined angle at predetermined time intervals using an electric turntable;

[0039] Images of the bearing sample are captured at the predetermined time intervals using an imaging device;

[0040] The bearing samples include multiple samples, specifically one or more bearings without defects and one or more bearings with defects.

[0041] In one preferred embodiment of the present invention, an automotive bearing wear condition detection device is also provided, comprising:

[0042] The system includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, can implement the steps of any of the above-described methods for detecting the wear condition of automotive bearings.

[0043] In one preferred embodiment of the present invention, a computer-readable storage medium is also provided, wherein a computer program is stored on the storage medium, and when the computer program is executed by a processor, the steps of the automobile bearing wear condition detection method of any one of the above-mentioned embodiments are implemented.

[0044] The method, equipment, and storage medium for detecting the wear condition of automotive bearings provided by the above solutions of the present invention have the following beneficial effects:

[0045] The proposed method for detecting wear conditions of automotive bearings involves: acquiring bearing images as a basic dataset; preprocessing the basic dataset; expanding the basic dataset to construct a target dataset; establishing an image mapping matrix based on the target dataset to obtain abnormal bearing images; and performing defect detection on the abnormal bearing images using a defect location mapping matrix to obtain defect detection results. This invention addresses the problem of model training's dependence on samples, as well as the issues of algorithm robustness and generalization ability. The algorithm achieves good recognition accuracy even with a small number of samples, significantly improving detection accuracy when applied to bearing wear detection scenarios. Attached Figure Description

[0046] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the structures shown in these drawings without creative effort.

[0047] Figure 1 A flowchart illustrating a method for detecting the wear condition of automotive bearings according to one embodiment of the present invention;

[0048] Figure 2 This is a schematic diagram illustrating the preprocessing flow of the basic dataset according to one embodiment of the present invention.

[0049] Figure 3 A schematic diagram illustrating the process of establishing an image mapping matrix according to one embodiment of the present invention;

[0050] Figure 4 A schematic diagram illustrating the process of establishing an image mapping matrix based on the target dataset according to another embodiment of the present invention;

[0051] Figure 5 This is a schematic diagram illustrating the process of detecting defects in abnormal bearing images using a defect location mapping matrix, according to one embodiment of the present invention.

[0052] Figure 6 This is a schematic diagram illustrating the process of acquiring bearing images as a basic dataset according to one embodiment of the present invention;

[0053] Figure 7 This is a schematic diagram illustrating the architecture of an automotive bearing wear condition detection device according to one embodiment of the present invention. Detailed Implementation

[0054] 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 a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0055] It should be noted that if the embodiments of the present invention involve directional indications (such as up, down, left, right, front, back, etc.), the directional indications are only used to explain the relative positional relationship and movement of the components in a specific posture. If the specific posture changes, the directional indications will also change accordingly.

[0056] Furthermore, if the embodiments of this invention involve descriptions such as "first" or "second," these descriptions are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined with "first" or "second" may explicitly or implicitly include at least one of those features. Additionally, the use of "and / or" or "and / or" throughout the text includes three parallel solutions. For example, "A and / or B" includes solution A, solution B, or a solution where both A and B are satisfied simultaneously. Furthermore, the technical solutions of the various embodiments can be combined with each other, but this must be based on the ability of those skilled in the art to implement them. When the combination of technical solutions is contradictory or impossible to implement, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed by this invention.

[0057] Please refer to Figure 1 One embodiment of the present invention provides a method for detecting the wear condition of automotive bearings, comprising:

[0058] S10. Collect bearing images as the basic dataset;

[0059] S20. Preprocess the basic dataset;

[0060] S30. Expand the basic dataset to construct the target dataset;

[0061] S40. Establish an image mapping matrix based on the target dataset to obtain abnormal bearing images;

[0062] S50. Defect detection is performed on the abnormal bearing image using the defect location mapping matrix to obtain the defect detection results.

[0063] Please refer to Figure 2In one preferred embodiment of the present invention, the preprocessing of the basic dataset includes:

[0064] S201. Perform two-dimensional wavelet multi-scale decomposition on the basic dataset;

[0065] S202. Denoise the basic dataset at various scales;

[0066] S203. Perform inverse wavelet transform on the basic dataset.

[0067] During image acquisition, transmission, reception, and processing, external and internal interferences are inevitably generated. Therefore, after acquiring the basic dataset, noisy images need to be processed. In one application scenario of this embodiment, wavelet transform is performed on noisy signals. Wavelet transform replaces the basis of Fourier transform with a finite-length, decaying wavelet basis. This overcomes the shortcomings of short-time Fourier transform and the fact that the window size does not change with frequency. By selecting a suitable basic wavelet, a(τ) has finite support in the time domain, while a(w) is relatively concentrated in the frequency domain. Wavelet transform is capable of characterizing the local characteristics of signals in both the time and frequency domains, which is beneficial for detecting singular and prominent signals in images, thus more effectively removing noise. The wavelet denoising formula is:

[0068]

[0069] Two-dimensional wavelet multi-scale decomposition is achieved by using directional filtering. Wavelet coefficients at each scale are extracted through scaling and translation, wavelet coefficients generated by noise are cleared to zero, and finally, the signal is reconstructed through inverse wavelet transform to obtain the denoised image.

[0070] In this application scenario, wavelet threshold denoising is used to remove noise. The basic idea is to perform wavelet transform on the original signal. The wavelet coefficients of the resulting signal are much larger than those of the noise. By selecting an appropriate threshold, wavelet coefficients generated by the signal with values ​​higher than this threshold are retained, while wavelet coefficients generated by the noise with values ​​lower than this threshold are eliminated. The signal is then reconstructed through inverse wavelet transform to achieve the denoising function. The threshold chosen is the adaptive threshold of the SURE unbiased estimation method, which converges quickly at the minimum mean square error, achieves the highest signal-to-noise ratio, and has the most significant denoising effect.

[0071] In one preferred embodiment of the present invention, the expansion process of the basic dataset includes:

[0072] The images of the base dataset are copied by rotation within a predetermined angle range to obtain the expanded dataset;

[0073] The basic dataset is divided into training set, validation set and test set according to a predetermined ratio and output as target dataset.

[0074] In one application scenario of this embodiment, 30 high-quality bearings and 30 bearings with different defects are used as samples. In a hexahedral black chamber, an electric turntable rotates the bearing samples 90°, and a camera takes a picture every time. A total of 240 images are obtained as the basic dataset. To avoid overfitting and low model robustness due to an insufficient dataset, this paper expands the dataset by rotating the images in the original basic dataset arbitrarily within the range of [-30°, 30°], resulting in 880 images, the expanded dataset. The training, validation, and test sets are then divided in a 6:2:2 ratio and output as the target dataset.

[0075] Please refer to Figure 3 In one preferred embodiment of the present invention, establishing the image mapping matrix includes:

[0076] S401. Convert the images in the target dataset to grayscale;

[0077] S402. Add the gray values ​​of each row of pixels in the image to calculate and obtain the first image mapping matrix;

[0078] S403. Calculate the maximum value, minimum value, average value, and standard deviation of the first image mapping matrix;

[0079] S404. Transform the first image mapping matrix to obtain the second image mapping matrix.

[0080] In one application scenario of this embodiment, the image mapping matrix is ​​calculated from the vertical or horizontal direction to represent the sum of image pixels. The first image mapping matrix is:

[0081]

[0082] In the first image mapping matrix, β is the scaling factor (100 in the experiment). Then, the maximum value Hmax, minimum value Hmin, mean value Havg, and standard deviation Hσ of the first image mapping matrix are calculated.

[0083] The first image mapping matrix is ​​transformed by subtracting the minimum value from each value of the first image mapping matrix to obtain the second image mapping matrix, wherein the second image mapping matrix is:

[0084]

[0085] In the second image mapping matrix, η represents the coefficients of change. When η is a constant, it is a linear transformation; otherwise, it is a nonlinear transformation.

[0086] Please refer to Figure 4In one preferred embodiment of the present invention, the step of establishing an image mapping matrix based on the target dataset further includes:

[0087] S405. Check whether the coefficients of change in the second image mapping matrix are constants;

[0088] S406. If the coefficient is a constant, then the second image mapping matrix is ​​determined to be a linear transformation of the first image mapping matrix, and the bearing has no defects.

[0089] S407. If the coefficient is an uncertain value, then it is determined that the second image mapping matrix is ​​a nonlinear transformation of the second image mapping matrix, and the bearing has a defect.

[0090] Please refer to Figure 5 In one preferred embodiment of the present invention, when a bearing has a defect, the defect detection of the abnormal bearing image through a defect location mapping matrix includes:

[0091] S501. Determine the detection threshold parameters based on the average value and standard deviation of the image mapping matrix;

[0092] S502. Based on the detection threshold parameter and the second image mapping matrix, a defect location mapping matrix is ​​constructed to detect the abnormal part of the second image mapping matrix and output a first detection result.

[0093] S503. Obtain the position where the gray value is 255 in the first detection result;

[0094] S504: Statistically count the positions where the grayscale value is continuously between 255 and greater than the predetermined threshold, and output the bearing defect location.

[0095] In one application scenario of this embodiment, the location of a bearing defect is determined by detecting abnormal parts through a second image mapping matrix. The defect location mapping matrix is ​​as follows:

[0096]

[0097] In the defect location mapping matrix, α is the detection threshold parameter. Then, H(y) is analyzed to detect the positions of 255 consecutive points. When the position is greater than a certain threshold θ, it is considered the bearing defect location.

[0098] In one preferred embodiment of the present invention, after obtaining the bearing defect location, calibration is performed using a convolutional neural network, specifically including:

[0099] Input historical bearing images to train the machine learning model and obtain a trained convolutional neural network;

[0100] The bearing image corresponding to the bearing defect location is imported into the convolutional neural network, and a second detection result is output.

[0101] Based on the second detection result, the location of the bearing defect is calibrated, and a bearing wear status report is output.

[0102] A convolutional layer in a convolutional neural network consists of several feature planes, each composed of neurons arranged in a rectangular pattern. Neurons within the same feature plane share weights, which are called the convolutional kernel. The initial values ​​of the convolutional kernel are randomly generated, and they are updated to the most suitable weights through gradual training. In this embodiment, the convolutional neural network includes two convolutional layers, using the ReLU activation function.

[0103] Please refer to Figure 6 In one preferred embodiment of the present invention, the acquisition of bearing images as a basic dataset specifically includes:

[0104] S101. Place the bearing sample in the black chamber;

[0105] S102. The bearing sample is rotated at a predetermined angle at predetermined time intervals using an electric turntable;

[0106] S103. Using a camera, images of the bearing sample are captured at the predetermined time intervals;

[0107] The bearing samples include multiple samples, specifically one or more bearings without defects and one or more bearings with defects.

[0108] In this embodiment, the black chamber is a hexahedral black chamber. The bearing sample is supported on the rotating end face of the electric turntable. The electric turntable drives the bearing sample to rotate 90° every 1 second through a worm gear transmission. The imaging device, which is fixedly set in the black chamber, takes pictures of the bearing sample from a fixed angle, thereby obtaining an all-round image of the bearing sample. The rotating platform base of the electric turntable adopts a common precision shaft system to ensure the coaxiality of the central through hole and the rotation center of the turntable, which has high rotational stability and prevents the bearing sample from tilting. The imaging device adopts an industrial camera with 20 megapixels and a maximum frame rate of 5fps or higher.

[0109] Please refer to Figure 7In one preferred embodiment of the present invention, an automotive bearing wear condition detection device 200 is also disclosed, comprising at least one processor 210; and a memory 220 communicatively connected to the at least one processor 210; wherein the memory 220 stores instructions executable by the at least one processor 210, the instructions being executed by the at least one processor 210 to enable the at least one processor 210 to implement the steps of the automotive bearing wear condition detection method of any of the above embodiments. The memory 220 stores a computer program 240, and the processor 210 and the memory 220 are connected via a communication bus 230.

[0110] In one preferred embodiment of the present invention, a computer-readable storage medium is also provided, wherein a computer program is stored on the storage medium, and when the computer program is executed by a processor, it implements the steps of the automobile bearing wear condition detection method of any of the above embodiments.

[0111] The proposed method for detecting wear conditions of automotive bearings involves: acquiring bearing images as a basic dataset; preprocessing the basic dataset; expanding the basic dataset to construct a target dataset; establishing an image mapping matrix based on the target dataset to obtain abnormal bearing images; and performing defect detection on the abnormal bearing images using a defect location mapping matrix to obtain defect detection results. This invention addresses the problem of model training's dependence on samples, as well as the issues of algorithm robustness and generalization ability. The algorithm achieves good recognition accuracy even with a small number of samples, significantly improving detection accuracy when applied to bearing wear detection scenarios.

[0112] The above description is merely a preferred embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural transformations made using the contents of the present invention's specification and drawings under the inventive concept of the present invention, or direct / indirect applications in other related technical fields, are included within the patent protection scope of the present invention.

Claims

1. A method of detecting a wear state of an automobile bearing, characterized by, The method comprises the following steps: Collecting bearing images as a basic data set; Preprocessing the basic data set; Augmenting the preprocessed basic data set, including: copying the preprocessed basic data set images through rotation in a predetermined angle range to obtain an augmented data set; dividing the augmented data set into a training set, a validation set and a test set according to a predetermined proportion and outputting the target data set; Establishing an image mapping matrix according to the target data set to obtain an abnormal bearing image; the establishment of the image mapping matrix comprises: grayscale processing the images in the target data set; adding the grayscale values of each row of pixels in the images to obtain a first image mapping matrix; calculating the maximum value, the minimum value, the average value and the standard deviation of the first image mapping matrix; transforming the first image mapping matrix to obtain a second image mapping matrix; querying whether the coefficient of change in the second image mapping matrix is a constant; if the coefficient is a constant, it is determined that the second image mapping matrix is a linear transformation of the first image mapping matrix, and the bearing does not have defects; if the coefficient is an uncertain value, it is determined that the second image mapping matrix is a nonlinear transformation of the second image mapping matrix, and the bearing has defects; Defect detection on the abnormal bearing image through a defect position mapping matrix to obtain a defect detection result, comprising: determining a detection threshold parameter according to the average value and the standard deviation of the first image mapping matrix; constructing a defect position mapping matrix based on the detection threshold parameter and the second image mapping matrix for detecting the abnormal part of the second image mapping matrix and outputting a first detection result; obtaining the position where the grayscale value is 255 in the first detection result; Counting the positions where the grayscale value is continuously 255 and the number of pixels where 255 continuously appears is greater than a preset threshold, and outputting the bearing defect position; Calibrating the bearing defect position through a convolutional neural network after obtaining the bearing defect position, comprising: Inputting historical bearing images to train a machine learning model to obtain a trained convolutional neural network; Importing the bearing image corresponding to the bearing defect position into the convolutional neural network to output a second detection result; Calibrating the bearing defect position based on the second detection result and outputting a bearing wear state report.

2. The automobile bearing wear state detection method according to claim 1, characterized by, The preprocessing of the basic data set comprises: Sequentially performing two-dimensional wavelet multi-scale decomposition, scale denoising and inverse wavelet transform on the basic data set.

3. The automobile bearing wear state detection method according to claim 1, characterized by, The collection of bearing images as a basic data set specifically comprises: Placing bearing samples in a darkroom that can eliminate environmental light interference; Rotating the bearing samples at a predetermined angle at a predetermined time interval through an electric turntable; Using a shooting device to shoot images of the bearing samples at the predetermined time interval; The bearing samples include one or more bearings without defects and one or more bearings with defects.

4. An automobile bearing wear state detecting apparatus characterized by comprising: The method comprises the following steps: A memory, a processor, and a computer program stored on the memory and executable on the processor, which, when executed by the processor, implements the steps of the automobile bearing wear state detection method according to any one of claims 1-3.

5. A computer readable storage medium, characterized in that, The storage medium stores a computer program, which, when executed by a processor, implements the steps of the automobile bearing wear state detection method according to any one of claims 1-3.

Images