A multi-component intelligent fault diagnosis method and system for a low-speed mechanical transmission system

Abstract

The application discloses a kind of low-speed mechanical transmission system multi-component intelligent fault diagnosis method and system, belong to mechanical fault diagnosis field, using improved robust local binary operation converts vibration signal into feature enhancement gray texture image, highlights the component feature information in original vibration data.Construct vibration gray texture image denoising model based on auto-encoder, it has good denoising effect to different degrees of noise. Through the optimization of the width of convolutional neural network model, the lightweight of fault diagnosis model is realized, which improves the fault diagnosis efficiency while ensuring the accuracy of fault diagnosis. The application solves the problem of low-speed and noise environment mechanical transmission system vibration signal weak energy and noise interference, which leads to low accuracy of fault diagnosis model. It provides an effective solution for intelligent fault diagnosis of low-speed, noise environment mechanical transmission system components.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of mechanical fault diagnosis, relates to the field of mechanical transmission system fault diagnosis, and in particular to a low-rotating-speed mechanical transmission system multi-component intelligent fault diagnosis method and system. BACKGROUND

[0002] The mechanical transmission system is a key component of mechanical equipment and is a complex system comprising numerous components. Among them, bearings, rotors and gears play the most crucial role in power transmission and motion conversion and are the most vulnerable components in mechanical equipment. Therefore, fault diagnosis of the mechanical transmission system composed of rotors, bearings and gears is of great importance to the safe and reliable operation of mechanical equipment.

[0003] For a long time, the fault diagnosis of mechanical transmission systems mainly uses vibration signals to extract fault features through deep signal analysis, and then completes fault diagnosis through expert experience. The fault diagnosis ability of the fault diagnosis method based on signal analysis is limited, and the use effect seriously depends on artificial experience, and it is impossible to realize automatic diagnosis without human intervention for various faults. With the rapid development of big data analysis and artificial intelligence technology, deep learning has gradually been used in the field of mechanical fault diagnosis due to its advantages of automatically obtaining data features and automatically classifying data.

[0004] In engineering practice, the vibration signals of mechanical equipment are often affected by strong background noise, and the mechanical fault diagnosis method will lose the accuracy of fault recognition when diagnosing the vibration signals containing noise. In addition, the effect of the mechanical fault diagnosis method based on deep learning is a difficulty in adapting to the data analyzed, which needs to be adjusted independently for different research purposes and data distribution rules, and the excessively complex deep learning model is difficult to train and is also difficult to apply in low-resource conditions. When the vibration signal is converted into a two-dimensional image, the above problems of the fault diagnosis method based on deep learning will be aggravated, and it is urgent to design a deep learning fault diagnosis method suitable for vibration images.

[0005] In addition, the performance law of vibration signals at low speed is not obvious, because the existence of the gap between the components in the mechanical transmission system causes the vibration law to be affected by nonlinear characteristics at low speed, so that the vibration signal distribution law of different fault states tends to be similar, causing the classification and recognition ability of the method based on deep learning to decrease for such data. Moreover, in the mechanical transmission system, different parts are sensitive to different forms of vibration signals, rotors and rolling bearing inner rings are more sensitive to vibration displacement signals and are suitable for monitoring by using an eddy current sensor, while gear and rolling bearing non-inner ring parts are more sensitive to vibration acceleration and are suitable for monitoring by using a piezoelectric sensor, so different types of vibration signals need to be used for state monitoring of different parts of the mechanical transmission system to achieve the best fault diagnosis. SUMMARY

[0006] The present application aims to provide a low-speed mechanical transmission system multi-component intelligent fault diagnosis method and system to overcome the problem that the image representation of the vibration signal is not clear and the intelligent fault diagnosis method is not effective when the mechanical equipment is in a low-speed, strong noise environment. The present application designs a method for converting vibration signals under low-speed conditions to gray texture images, combines a self-encoder vibration gray texture image denoising model and a lightweight convolutional neural network intelligent diagnosis model to achieve efficient, accurate and intelligent diagnosis of mechanical transmission system multi-component faults under low-speed and noise conditions.

[0007] To achieve the above-mentioned purpose, the present application adopts the following technical solutions:

[0008] A low-speed mechanical transmission system multi-component intelligent fault diagnosis method, comprising the following steps:

[0009] 1) Constructing a mechanical transmission system vibration gray texture image dataset

[0010] Collect vibration acceleration and displacement signals of bearings, rotors and gears in the mechanical transmission system under normal and fault conditions at low speed, segment and splice the vibration acceleration and displacement signals into feature matrices according to specified points, then sequentially perform standardization, proportional scaling and robust local binary processing to generate mechanical transmission system vibration gray texture images, construct a mechanical transmission system vibration gray texture image dataset from all generated images, and divide the vibration gray texture image dataset into a training set and a test set according to a proportion;

[0011] 2) Constructing a self-encoder vibration gray texture image denoising model

[0012] Based on the mechanical transmission system vibration gray texture image dataset in step 1), a self-encoder vibration gray texture image denoising model is constructed, which includes multi-scale convolution, Ghost convolution and ECB convolution. The training set is used for model training, and the test set is used for model testing. The denoising effect of the denoising model is evaluated by peak signal-to-noise ratio and structural similarity index;

[0013] 3) Constructing a mechanical transmission system fault diagnosis model

[0014] Based on the lightweight convolutional neural network, the model width is optimized, an improved convolutional neural network fault diagnosis model is constructed, based on the vibration gray texture image data set of the mechanical transmission system in step 1), the fault feature data set is established after being labeled with faults, and is divided into training set and test set, the training set is used for model training, the test set is used for testing the final diagnosis result, the final structure and parameters of the fault diagnosis model are saved, after determining the model structure and parameters, the vibration gray texture image data set in step 1) is added with light, medium and heavy noise of different degrees, the vibration gray texture image data set containing noise is established, and according to the best parameters of the noise reduction model in step 2), the best parameters are used to reduce the vibration gray texture image data set with different degrees of noise, the vibration gray texture image after noise reduction is labeled to establish the fault feature data set after noise reduction, and is divided into training set and test set, the training set is used for model training, and the test set is used for testing the diagnosis effect of the fault diagnosis model on the noise reduction image data set.

[0015] Further, the step 1) constructs the vibration gray texture image data set of the mechanical transmission system, specifically:

[0016] 1.1) Collect the vibration acceleration and displacement signals of bearings, rotors and gears in the mechanical transmission system under normal and fault conditions at low speed, segment and splice the vibration acceleration and displacement signals according to the specified number of points respectively to form a feature matrix;

[0017] 1.2) The elements of the feature matrix are standardized and converted to gray images by 0-255, and the conversion relationship between the element value of the feature matrix and the pixel value of the gray image is:

[0018]

[0019] In the formula, i , j represents the position coordinates in the feature matrix or the gray image, is the value of the element located in the feature matrix, i , j X max and X min respectively represent the maximum value and the minimum value in the feature matrix, is the gray value of the corresponding position after conversion;

[0020] 1.3) The converted gray image is scaled to 256x256 pixels in proportion;

[0021] ​1.4) Robust local binary processing is performed on the image obtained after equal proportion scaling to generate a vibrating gray texture image, and a weighted local gray value is introduced in the robust local binary processing to adjust the weight of the center pixel on the threshold selection;

[0022]

[0023] wherein, is the gray value of the center pixel, is the gray value of the adjacent pixel on the circle with a radius of R is a constant, The greater the value, the greater the proportion of the center threshold;

[0024] After threshold selection, the texture features of the image are described by robust local binary processing, and the size of the pixel area of the circle in the image is calculated in turn. The calculation formula of robust local binary processing is:

[0025]

[0026] wherein, p represents the total number of selected field points on the circumference, , represents the gray value of the field point, RLBP P,R is the center point RLBP value processed by the RLBP operator,

[0027] 1.5) Add different degrees of Gaussian noise to the obtained vibrating gray texture image to construct a noisy gray texture image dataset, and divide it into a training set and a test set according to a predetermined proportion.

[0028] Further, the predetermined proportion is 8:2.

[0029] Further, the generated image sample contains prominent texture detail features, and the self-encoder denoising model is used to extract features and reconstruct images from the noisy vibrating gray texture image dataset in step 1.5) to achieve the purpose of denoising.

[0030] Further, the self-encoder vibrating gray texture image denoising model in step 2) is constructed, specifically:

[0031] 2.1) Construct a self-encoder vibrating gray texture image denoising model, which includes multi-scale convolution, Ghost convolution and ECB convolution;

[0032] ​​2.2) Use step 1) to construct the bearing, rotor and gear in the mechanical transmission system in low speed working condition under normal and fault state vibration gray texture image data set, according to the proportion of 8:2, divide the training set and test set, add three different degree of Gaussian noise with standard deviation of 15, 25, 50 to the data set, update the model hyperparameter setting to achieve the optimal denoising effect, and evaluate the denoising effect of the denoising model by peak signal-to-noise ratio and structural similarity index.

[0033] Further, the step 3) of constructing the mechanical transmission system fault diagnosis model, specifically:

[0034] 3.1) Construct a convolutional neural network fault diagnosis model, which includes a feature extractor and a classifier. Optimize the width of the model structure in the feature extraction part and simplify the model structure.

[0035] 3.2) For low speed working condition, use step 1) to construct the bearing, rotor and gear in the mechanical transmission system in normal and fault state vibration gray texture image data set, label the vibration gray texture image after denoising in step 2) to establish the data set, and divide it into training set and test set, input to the fault diagnosis model for training and testing, and the classification accuracy, precision, recall and F1 score of the fault diagnosis model on the test set are used as evaluation indexes, and the calculation formula is:

[0036]

[0037]

[0038]

[0039]

[0040] In the formula, ACC, precision, recall and F1 represent the accuracy, precision, recall and F1 score of the test set respectively; TP represents the number of samples that are actually positive examples and are correctly predicted as positive examples; TN represents the number of samples that are actually negative examples and are correctly predicted as negative examples; FP represents the number of samples that are actually negative examples but are incorrectly predicted as positive examples; FN represents the number of samples that are actually positive examples but are incorrectly predicted as negative examples.

[0041] A low-speed mechanical transmission system multi-component intelligent fault diagnosis system, comprising:

[0042] The mechanical transmission system vibration gray texture image dataset construction module: collects the vibration acceleration and displacement signals of bearings, rotors and gears in the mechanical transmission system under normal and fault conditions at low speed, segments and splices the vibration acceleration and displacement signals into feature matrices according to specified points, and then sequentially performs standardization, equal proportion scaling and robust local binary processing to generate corresponding vibration gray texture images. All generated images form the mechanical transmission system vibration gray texture image dataset, and the vibration gray texture image dataset is divided into a training set and a test set according to a proportion;

[0043] The autoencoder vibration gray texture image denoising model construction module: is used to construct a vibration gray texture image dataset for low speed conditions by using the low speed condition data and the operations in the mechanical transmission system vibration gray texture image dataset construction module. An autoencoder vibration gray texture image denoising model is constructed using the vibration gray texture image dataset for low speed conditions. The autoencoder denoising model includes multi-scale convolution, Ghost convolution and ECB convolution. The training set is used for model training, and the test set is used for model testing. The peak signal-to-noise ratio and structural similarity index are used to evaluate the denoising effect of the denoising model.

[0044] The mechanical transmission system fault diagnosis model construction module: constructs a convolutional neural network fault diagnosis model, which includes a feature extractor and a classifier. The model structure of the feature extraction part is optimized in width, and the model structure is simplified. For low speed condition data, vibration gray texture image datasets of bearings, rotors and gears in the mechanical transmission system under normal and fault conditions are constructed. The vibration gray texture image samples denoised by the denoising model are labeled to establish a fault feature dataset, which is divided into a training set and a test set and input into the fault diagnosis model for training and testing. The classification accuracy, precision, recall and F1 score of the fault diagnosis model on the test set are used as evaluation indicators to complete the fault diagnosis of the mechanical transmission system under low speed conditions and noise interference.

[0045] Further, the construction process of the mechanical transmission system vibration gray texture image dataset is as follows:

[0046] 1.1) Collect the vibration acceleration and displacement signals of bearings, rotors and gears in the mechanical transmission system under normal and fault conditions at low speed, segment and splice the vibration acceleration and displacement signals according to specified points to form feature matrices;

[0047] 1.2) Standardize the elements of the feature matrix to 0-255 to convert it into a gray image. The conversion relationship between the feature matrix element value and the gray image pixel value is:

[0048]

[0049] In the formula,i , j Represents the position coordinates in the feature matrix or grayscale image. For the characteristic matrix located in ( i , j The value of the element, X max and X min represents the maximum and minimum values ​​in the characteristic matrix, respectively. This represents the grayscale value at the corresponding position after conversion;

[0050] 1.3) Scale the converted grayscale image proportionally to 256×256 pixels;

[0051] 1.4) Robust local binarization is performed on the image obtained after proportional scaling to generate a vibration grayscale texture image. Weighted local grayscale values ​​are introduced in the robust local binarization process to adjust the weight of the center pixel on the threshold selection.

[0052]

[0053] In the formula, The grayscale value of the center pixel. For radius R The grayscale values ​​of adjacent pixels on the circle. It is a constant. The larger the value, the greater the proportion of the center threshold;

[0054] After threshold selection, robust local binarization is performed to describe the texture features of the image. The robust local binarization is calculated sequentially based on the size of circular pixel regions in the image. The formula for robust local binarization is:

[0055]

[0056] In the formula, p This represents the total number of points selected on the circumference. , RLBP represents the grayscale value of a neighborhood point. P, R The RLBP value of the center point after processing by the RLBP operator. ;

[0057] 1.5) Add Gaussian noise of different degrees to all the obtained vibration grayscale texture images to construct a noisy grayscale texture image dataset, and divide it into training set and test set according to a preset ratio.

[0058] Furthermore, the construction of the autoencoder vibration grayscale texture image denoising model specifically involves:

[0059] 2.1) Construct an autoencoder vibration grayscale texture image denoising model, including multi-scale convolution, Ghost convolution and ECB convolution;

[0060] 2.2) Use step 1) to construct the bearing, rotor and gear in the mechanical transmission system in low speed working condition under normal and fault state vibration gray texture image data set, according to the proportion of 8:2, divide the training set and test set, add three different degree Gaussian noise with standard deviation of 15, 25, 50 to the data set, update the model hyperparameter setting to achieve the optimal denoising effect, and evaluate the denoising effect of the denoising model by peak signal-to-noise ratio and structural similarity index.

[0061] Further, the construction mechanical transmission system fault diagnosis model, specifically:

[0062] 3.1) Construct a convolutional neural network fault diagnosis model, including a feature extractor and a classifier, optimize the width of the model structure of the feature extraction part, and simplify the model structure;

[0063] 3.2) For low speed working condition, use step 1) to construct the bearing, rotor and gear in the mechanical transmission system in normal and fault state vibration gray texture image data set, label the vibration gray texture image after denoising in step 2) to establish the data set, and divide it into training set and test set, input to the fault diagnosis model for training and testing, and the classification accuracy, precision, recall and F1 score of the fault diagnosis model on the test set are used as evaluation indexes, and the calculation formula is:

[0064]

[0065]

[0066]

[0067]

[0068] In the formula, ACC, precision, recall and F1 represent the accuracy, precision, recall and F1 score of the test set respectively; TP represents the number of samples that are actually positive examples and are correctly predicted as positive examples; TN represents the number of samples that are actually negative examples and are correctly predicted as negative examples; FP represents the number of samples that are actually negative examples but are incorrectly predicted as positive examples; FN represents the number of samples that are actually positive examples but are incorrectly predicted as negative examples.

[0069] Compared with the prior art, the present application has the following beneficial technical effects:

[0070] (1) In order to solve the problem that the fault vibration characteristics are not obvious under low speed working condition, an improved gray texture pattern sample generation method is proposed. The texture features and detail information of the gray image are enhanced by robust local binary operation, and the state characteristics of different components of the mechanical transmission system are highlighted by using the obtained gray texture image information. Compared with the conventional vibration gray image conversion method, the improved vibration gray texture image forms a large difference between different fault categories samples, and compared with the gray texture image generated by the traditional local binary operation, it has clearer texture details, which is conducive to further feature extraction.

[0071] (2) In order to reduce the negative influence of background noise in actual working environment on fault diagnosis effect, a multi-module vibration gray texture image denoising model based on autoencoder is proposed. The multi-scale convolution, ghost convolution and ECB convolution in inception are used to improve the autoencoder, which enhances the feature extraction ability and image reconstruction ability of the vibration gray texture image denoising model. The improved autoencoder model has the advantages of few parameters, small calculation amount, high speed and small model size, which can remove background noise of different degrees and better restore the texture details of the vibration gray texture image.

[0072] (3) An improved convolutional neural network fault diagnosis model is proposed. By optimizing the bottleneck layer structure of the basic lightweight convolutional neural network, the parameter amount, calculation amount and model size of the model are reduced, and the high fault diagnosis accuracy is ensured. The method can still achieve good results under background noise, and has strong robustness to noise, and maintains high fault diagnosis accuracy for the denoised data set.

[0073] In order to solve the problem that the vibration signal energy of mechanical transmission system under low speed working condition is weak and the noise interference is large, resulting in low accuracy of fault diagnosis model, an intelligent fault diagnosis method based on vibration signal mapping gray texture image and autoencoder pre-training convolutional neural network is proposed. The improved gray texture image converted from vibration signal under low speed working condition is used for autoencoder denoising and convolutional neural network fault diagnosis. The accuracy, precision, recall and F1 score of data classification are used as indicators to evaluate the effect of fault diagnosis model. Compared with other fault diagnosis methods of mechanical transmission system, the invention can accurately extract image features and maintain high fault diagnosis accuracy under low speed and noise environment. BRIEF DESCRIPTION OF DRAWINGS

[0074] The drawings for the specification are used to provide further understanding of the invention, and form a part of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention, and do not constitute an improper limitation of the invention.

[0075] Figure 1 It is a flow chart of a low speed mechanical transmission system multi-component intelligent fault diagnosis method.

[0076] Figure 2 vibration gray texture images converted from vibration signals for different component fault states;

[0077] Figure 3 vibration gray texture image auto-encoder denoising model structure diagram constructed by the present application;

[0078] Figure 4 improved convolutional neural network fault diagnosis model structure diagram constructed by the present application;

[0079] Figure 5 classification performance of the present application on data sets containing different degrees of noise and denoised data sets. DETAILED DESCRIPTION

[0080] In order for those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor should be within the scope of protection of the present application.

[0081] It should be understood that the data used in this way can be interchanged under appropriate circumstances, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device containing a series of steps or units does not have to be limited to those steps or units clearly listed, but can include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0082] Embodiment one

[0083] A low-speed mechanical transmission system multi-component intelligent fault diagnosis method, comprising the following steps:

[0084] 1) Construct a mechanical transmission system vibration gray texture image data set:

[0085] The vibration acceleration and displacement signals of bearings, rotors and gears in the mechanical transmission system under normal and fault conditions at low speed are collected, the vibration acceleration and displacement signals are segmented and spliced into feature matrices according to specified points, and then the feature matrices are sequentially subjected to standardization, equal scaling and robust local binary processing to generate vibration gray texture images of the mechanical transmission system, all the generated images are used to form a vibration gray texture image dataset of the mechanical transmission system, and the vibration gray texture image dataset is divided into a training set and a test set according to a proportion;

[0086] The specific steps of constructing the mechanical transmission system fault feature dataset in step 1) are as follows:

[0087] 1.1) Collect the vibration acceleration and displacement signals of bearings, rotors and gears in the mechanical transmission system under normal and fault conditions at low speed, segment and splice the vibration acceleration and displacement signals according to specified points to form feature matrices;

[0088] 1.2) The elements of the feature matrix are subjected to 0-255 standardization processing to convert into a gray image, and the conversion relationship between the element value of the feature matrix and the pixel value of the gray image is:

[0089]

[0090] In the formula, i , j represents the position coordinates in the feature matrix or the gray image, is the value of the element located at i , j in the feature matrix, X max and X min respectively represent the maximum value and the minimum value in the feature matrix, is the gray value of the corresponding position after conversion;

[0091] 1.3) The converted gray image is scaled to 256x256 pixels in size;

[0092] 1.4) The image obtained after equal scaling is subjected to robust local binary processing to generate a vibration gray texture image, and a weighted local gray value is introduced in the robust local binary processing to adjust the weight of the center pixel on the threshold selection;

[0093]

[0094] In the formula, is the gray value of the center pixel, is the gray value of the adjacent pixel on the circle with a radius of R , and is a constant, The greater the value, the greater the proportion of the center threshold value;

[0095] After the threshold selection is completed, the texture features of the image are described by robust local binary, and the size of the pixel area of the circle in the image is calculated in turn. The calculation formula of the robust local binary is:

[0096]

[0097] In the formula, p The total number of field points selected on the circumference is represented by , The gray value of the field point is represented by RLBP P,R The center point RLBP value after the RLBP operator is processed, ;

[0098] 1.5) Add different degrees of Gaussian noise to the obtained all vibration gray texture images to construct a noisy gray texture image dataset, and divide it into a training set and a test set according to a predetermined proportion.

[0099] 2) Constructing an autoencoder vibration gray texture image denoising model: based on the mechanical transmission system vibration gray texture image dataset in step 1), an autoencoder vibration gray texture image denoising model is constructed, which contains multi-scale convolution, Ghost convolution and ECB convolution. The training set is used for model training, and the test set is used for model testing. The denoising effect of the denoising model is evaluated by peak signal-to-noise ratio and structural similarity index;

[0100] The specific steps of constructing the autoencoder vibration gray texture image denoising model in step 2) are as follows:

[0101] 2.1) Constructing an autoencoder vibration gray texture image denoising model contains multi-scale convolution, Ghost convolution and ECB convolution.

[0102] 2.2) Use step 1) to construct the vibration gray texture image dataset of the bearing, rotor and gear in the mechanical transmission system under normal and fault conditions at low speed. Divide the training set and test set according to the ratio of 8:2. Add three different degrees of Gaussian noise with standard deviation of 15, 25 and 50 to the dataset. Update the model hyperparameter settings to achieve the optimal denoising effect. Evaluate the denoising effect of the denoising model by peak signal-to-noise ratio and structural similarity index.

[0103] 3) Constructing a fault diagnosis model: Select MobileNetV3 as the basic model of the embodiment, optimize the model width, and construct an improved convolutional neural network fault diagnosis model. Based on the vibration gray texture image data set of the mechanical transmission system in step 1), the fault feature data set is established after being labeled, and is divided into training set and test set. The training set is used for model training, and the test set is used for testing the final diagnosis result. Save the final structure and parameters of the fault diagnosis model, and determine the model structure and parameters. Add light, medium and heavy noise of different degrees to the vibration gray texture image data set in step 1), and establish a noisy vibration gray texture image data set. According to the best parameters of the noise reduction model obtained in step 2), use the best parameters to reduce the vibration gray texture image data set with different degrees of noise. The vibration gray texture image after noise reduction is labeled to establish a fault feature data set after noise reduction, and is divided into training set and test set. The training set is used for model training, and the test set is used for testing the final diagnosis effect. Save the final structure and parameters of the fault diagnosis model.

[0104] The specific steps of the mechanical transmission system fault diagnosis model in step 3) are as follows:

[0105] 3.1) Construct a convolutional neural network fault diagnosis model based on MobileNetV3, which includes a feature extractor and a classifier. Optimize the model structure of the feature extraction part in width, and simplify the model structure; 3.2) For low speed working condition, use the vibration gray texture image data set of bearings, rotors and gears in normal and fault states in step 1) to establish the data set after labeling the images in step 2), and divide it into training set and test set. Input to the fault diagnosis model for training and testing. The classification accuracy, precision, recall and F1 score of the fault diagnosis model on the test set are used as evaluation indexes, and the calculation formula is:

[0106]

[0107]

[0108]

[0109]

[0110] In the formula, ACC, precision, recall and F1 represent the accuracy, precision, recall and F1 score of the test set, respectively. TP represents the number of samples that are actually positive examples and are correctly predicted as positive examples; TN represents the number of samples that are actually negative examples and are correctly predicted as negative examples; FP represents the number of samples that are actually negative examples but are incorrectly predicted as positive examples; FN represents the number of samples that are actually positive examples but are incorrectly predicted as negative examples.

[0111] Example Two

[0112] Referring to the accompanying drawings, Figure 1 A low-speed mechanical transmission system multi-component intelligent fault diagnosis method of the present application mainly comprises the following steps:

[0113] 1) Use the mechanical transmission system fault diagnosis test bench to perform mechanical transmission system fault simulation experiments. Three bearing fault components, two gear fault components, and two rotor fault components are selected for the experiments, including seven fault states of bearing inner ring fault, bearing outer ring fault, bearing rolling element fault, gear tooth surface wear, gear tooth root crack, rotor imbalance, and rotor misalignment, as well as one normal state. The experimental conditions are set as low speed (700 r / min) and no load (0 N·m). For the bearing outer ring, rolling element fault, and gear fault, an acceleration sensor is used to collect the test bench shell vibration signal; for the bearing inner ring and rotor fault, an eddy current displacement sensor is used to collect the vibration signal;

[0114] 1.1) Collect vibration acceleration and displacement signals of different components and different states to obtain original vibration data. The obtained original vibration data is divided into multiple groups without overlap with 4096 sampling points as one group, and is spliced into a feature matrix in sequence by row;

[0115] 1.2) Perform 0-255 standardization processing on the elements of the feature matrix to convert it into a gray-scale image. The conversion relationship between the element value of the feature matrix and the pixel value of the gray-scale image is:

[0116]

[0117] In the formula, i , j represents the position coordinates in the feature matrix or the gray-scale image, is the value of the element located at i , j in the feature matrix, X max and X min represent the maximum value and the minimum value in the feature matrix, respectively, is the gray-scale value of the corresponding position after conversion;

[0118] 1.3) Scale the converted gray-scale image to 256x256 pixels in proportion;

[0119] 1.4) Perform robust local binary processing on the image obtained after proportional scaling to generate a vibration gray-scale texture image, and introduce a weighted local gray-scale value in the robust local binary processing to adjust the weight of the central pixel on the threshold selection;

[0120]

[0121] wherein, is the gray value of the center pixel, is the gray value of the adjacent pixel on the circle with a radius of R , is a constant, The greater the value, the greater the proportion of the center threshold value;

[0122] After the threshold selection is completed, the texture feature of the image is described by robust local binary, and the size of the pixel area of the circle in the image is calculated in turn, and the calculation formula of the robust local binary is:

[0123]

[0124] wherein, p denotes the total number of field points selected on the circumference, , denotes the gray value of the field point. RLBP P,R is the center point RLBP value processed by the RLBP operator, ;

[0125] 1.5) Add different degrees of Gaussian noise to the obtained all vibration gray texture images to construct a noisy gray texture image dataset, and divide it into a training set and a test set according to a preset proportion.

[0126] 2) Constructing an autoencoder vibration gray texture image denoising model, specifically comprising the following steps:

[0127] 2.1) Constructing an autoencoder vibration gray texture image denoising model, comprising multi-scale convolution, Ghost convolution and ECB convolution;

[0128] 2.2) Using step 1) to construct the vibration gray texture image dataset of the bearing, rotor and gear in the mechanical transmission system under normal and fault conditions at low speed, dividing the training set and the test set according to the ratio of 8:2, adding three different degrees of Gaussian noise with standard deviations of 15, 25 and 50 to the dataset, updating the model hyperparameter settings to achieve the optimal denoising effect, and evaluating the denoising effect of the denoising model through the peak signal-to-noise ratio and the structural similarity index.

[0129] The vibration gray texture image denoising results under different degrees of Gaussian noise by using five comparison methods and the method of the present application are shown in Tables 1 to 3:

[0130] Table 1 Denoising results of the model under Gaussian noise with a standard deviation of 15

[0131]

[0132] Table 2: Denoising results of each model under Gaussian noise with standard deviation 25

[0133]

[0134] Table 3: Model denoising results under Gaussian noise with standard deviation 50

[0135]

[0136] 3) Construct a mechanical transmission system fault diagnosis model, specifically including the following steps:

[0137] 3.1) Construct a convolutional neural network fault diagnosis model, including a feature extractor and a classifier, optimize the model structure of the feature extraction part, simplify the model structure, and the designed model structure is shown in FIG. 3. Figure 4

[0138] 3.2) For low-speed, no-load working condition data, use step 1) to construct the vibration gray texture image data set of bearings, rotors and gears in the mechanical transmission system under low-speed, no-load working conditions, label the vibration gray texture image after denoising in step 2) to establish a data set, and divide it into a training set and a test set, input to the fault diagnosis model for training and testing, and the specific parameter settings of each layer of the final model are shown in Table 4:

[0139] Table 4: Parameter settings of each layer of the model

[0140]

[0141] The classification accuracy, precision, recall and F1 score of the fault diagnosis model on the test set are used as evaluation indicators. The test accuracy of the final model on each denoising data set is above 96%, and the classification results of different levels of noise data sets and denoising data sets are shown in FIG. 4. Figure 5

[0142] To prove the superiority of the method of the application, the method of the application is compared with the original MobileNetV3 basic model and GoogLeNet, AlexNet two kinds of classical convolutional neural networks, also under the same low-speed, no-load working condition, the results are shown in Table 5.

[0143] Table 5: Comparison of results of different diagnosis methods

[0144]

[0145] Example Three

[0146] A low-speed mechanical transmission system multi-component intelligent fault diagnosis system, comprising:

[0147] ​​The mechanical transmission system vibration grayscale texture image dataset construction module: the vibration acceleration and displacement signals of bearings, rotors, and gears in the mechanical transmission system under normal and fault conditions at low speed are collected, the vibration acceleration and displacement signals are segmented and spliced into feature matrices according to specified points, and then the feature matrices are sequentially subjected to standardization, equal proportion scaling, and robust local binary processing to generate mechanical transmission system vibration grayscale texture images, all the generated images form a mechanical transmission system vibration grayscale texture image dataset, and the vibration grayscale texture image dataset is divided into a training set and a test set according to a proportion;

[0148] The autoencoder vibration grayscale texture image denoising model construction module: used for constructing a vibration grayscale texture image dataset of the low speed condition by using the low speed condition data and the operation in the mechanical transmission system vibration grayscale texture image dataset construction module, and constructing an autoencoder vibration grayscale texture image denoising model by using the vibration grayscale texture image dataset of the low speed condition. The autoencoder denoising model contains multi-scale convolution, Ghost convolution, and ECB convolution. The training set is used for model training, and the test set is used for model testing, and the peak signal-to-noise ratio and structural similarity index are used to evaluate the denoising effect of the denoising model.

[0149] The mechanical transmission system fault diagnosis model construction module: a convolutional neural network fault diagnosis model is constructed, containing a feature extractor and a classifier, the model structure of the feature extraction part is optimized in width, and the model structure is simplified. For low speed condition data, vibration grayscale texture image datasets of bearings, rotors, and gears in the mechanical transmission system under normal and fault conditions are constructed, the vibration grayscale texture image samples denoised by the denoising model are labeled to establish a fault feature dataset, and the fault feature dataset is divided into a training set and a test set and input to the fault diagnosis model for training and testing. The classification accuracy, precision, recall, and F1 score of the fault diagnosis model on the test set are used as evaluation indexes to complete the fault diagnosis of the mechanical transmission system under low speed condition and noise interference.

[0150] Those skilled in the art will appreciate that embodiments of the application can be provided as methods, systems, or computer program products. Accordingly, the application can take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the application can take the form of a computer program product on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROMs, optical storage media, etc.) having computer-usable program code embodied in the medium.

[0151] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 2. The flowchart blocks and / or data flow diagram blocks of the figure(s) illustrate the functionality of a possible implementation of the present application. Figure 1 2. The flowchart blocks and / or data flow diagram blocks of the figure(s) illustrate the functionality of a possible implementation of the present application. 2. The flowchart blocks and / or data flow diagram blocks of the figure(s) illustrate the functionality of a possible implementation of the present application.

[0152] 2. The flowchart blocks and / or data flow diagram blocks of the figure(s) illustrate the functionality of a possible implementation of the present application. Figure 1 2. The flowchart blocks and / or data flow diagram blocks of the figure(s) illustrate the functionality of a possible implementation of the present application. Figure 1 2. The flowchart blocks and / or data flow diagram blocks of the figure(s) illustrate the functionality of a possible implementation of the present application. 2. The flowchart blocks and / or data flow diagram blocks of the figure(s) illustrate the functionality of a possible implementation of the present application.

[0153] 2. The flowchart blocks and / or data flow diagram blocks of the figure(s) illustrate the functionality of a possible implementation of the present application. Figure 1 2. The flowchart blocks and / or data flow diagram blocks of the figure(s) illustrate the functionality of a possible implementation of the present application. Figure 1 2. The flowchart blocks and / or data flow diagram blocks of the figure(s) illustrate the functionality of a possible implementation of the present application. 2. The flowchart blocks and / or data flow diagram blocks of the figure(s) illustrate the functionality of a possible implementation of the present application.

[0154] 2. The flowchart blocks and / or data flow diagram blocks of the figure(s) illustrate the functionality of a possible implementation of the present application.Finally, it should be noted that the above-mentioned embodiments are merely intended for describing and illustrating, not limiting, the technical solutions of the present application, and although the present application has been described in detail with reference to the above-mentioned embodiments, those skilled in the art should understand that after reading the present application, the specific implementation manners of the present application can be variously changed, modified or replaced by equivalents, and these changes, modifications or replacements by equivalents are all within the protection scope of the claims of the present application.​​​​​​​​

Claims

1. A multi-component intelligent fault diagnosis method for low-speed mechanical transmission systems, characterized in that, Comprising the following steps: 1) Constructing a mechanical transmission system vibration grayscale texture image dataset Collect the vibration acceleration and displacement signals of bearings, rotors and gears in the mechanical transmission system under normal and fault conditions at low speed, segment and splice the vibration acceleration and displacement signals into feature matrices according to the specified number of points, then sequentially perform standardization, equal scaling and robust local binary processing to generate mechanical transmission system vibration grayscale texture images, and construct a mechanical transmission system vibration grayscale texture image dataset from all generated images, and divide the vibration grayscale texture image dataset into training and test sets according to the proportion; 2) Constructing a vibration grayscale texture image denoising model of the autoencoder Based on the mechanical transmission system vibration grayscale texture image dataset in step 1), a vibration grayscale texture image denoising model of the autoencoder is constructed, which includes multi-scale convolution, Ghost convolution and ECB convolution, the training set is used for model training, and the test set is used for model testing, and the denoising effect of the denoising model is evaluated by peak signal-to-noise ratio and structural similarity index; 3) Constructing a mechanical transmission system fault diagnosis model Based on the lightweight convolutional neural network, the model width is optimized to construct an improved convolutional neural network fault diagnosis model, based on the mechanical transmission system vibration grayscale texture image dataset in step 1), a fault feature dataset is established after labeling, and is divided into training and test sets, the training set is used for model training, and the test set is used for final diagnosis result, the final structure and parameters of the fault diagnosis model are saved, after determining the model structure and parameters, add light, medium and heavy noise of different degrees to the mechanical transmission system vibration grayscale texture image dataset in step 1), establish a noisy vibration grayscale texture image dataset, and use the best parameters obtained from the evaluation index of the denoising model in step 2) to denoise the vibration grayscale texture image dataset with different degrees of noise, label the denoised vibration grayscale texture image to establish a denoised fault feature dataset, and divide it into training and test sets, the training set is used for model training, and the test set is used for testing the diagnosis effect of the fault diagnosis model on the denoised image dataset.

2. The multi-component intelligent fault diagnosis method for low-speed mechanical transmission system according to claim 1, characterized in that, The step 1) constructs a mechanical transmission system vibration grayscale texture image dataset, specifically: 1.1) Collect the vibration acceleration and displacement signals of bearings, rotors and gears in the mechanical transmission system under normal and fault conditions at low speed, segment and splice the vibration acceleration and displacement signals into feature matrices according to the specified number of points; 1.2) Standardize the elements of the feature matrix to 0-255 to convert it into a grayscale image, the conversion relationship between the feature matrix element value and the grayscale image pixel value is: wherein, i , j represents the position coordinate in the feature matrix or the gray image, is the value of the element located at i , j in the feature matrix, X max and X min represent the maximum value and the minimum value in the feature matrix, respectively, is the gray value of the corresponding position after transformation; 1.3) Scale the converted grayscale image to 256x256 pixels in proportion; 1.4) Perform robust local binary processing on the scaled image to generate a vibration grayscale texture image, and introduce a weighted local grayscale value in the robust local binary processing to adjust the weight of the center pixel on the threshold selection; In the formula, The grayscale value of the center pixel. For radius R The grayscale values ​​of adjacent pixels on the circle. It is a constant. The larger the value, the greater the proportion of the center threshold; After the threshold selection is completed, the texture features of the robust local binary description image are performed, and the size of the pixel area of the circle in the image is calculated in turn, and the calculation formula of the robust local binary is: wherein p denotes the total number of field points selected on the circumference, , denotes the gray value of the field point, RLBP P,R is the center point RLBP value after processing by the RLBP operator, ; 1.5) Add different degrees of Gaussian noise to the obtained all vibration gray texture images to construct a noisy gray texture image dataset, and divide it into a training set and a test set according to a predetermined proportion.

3. The multi-component intelligent fault diagnosis method of a low-rotating-speed mechanical transmission system according to claim 2, characterized in that, The predetermined proportion is 8:

2.

4. The multi-component intelligent fault diagnosis method of a low-rotating-speed mechanical transmission system according to claim 2, characterized in that, The generated image samples contain prominent texture detail features, and the feature extraction and image reconstruction of the noisy vibration gray texture image dataset in step 1.5) are performed by the autoencoder denoising model to achieve the purpose of denoising.

5. The multi-component intelligent fault diagnosis method of a low-rotating-speed mechanical transmission system according to claim 1, characterized in that, The step 2) constructs an autoencoder vibration gray texture image denoising model, specifically: 2.1) Constructing an autoencoder vibration gray texture image denoising model includes multi-scale convolution, Ghost convolution and ECB convolution; 2.2) Use step 1) to construct the vibration gray texture image dataset of the bearing, rotor and gear in the mechanical transmission system under normal and fault conditions at low speed, divide the training set and test set according to the ratio of 8:2, add three different degrees of Gaussian noise with standard deviation of 15, 25 and 50 to the dataset, and update the model hyperparameter settings to achieve the optimal denoising effect. The peak signal-to-noise ratio and structural similarity index are used to evaluate the denoising effect of the denoising model.

6. The multi-component intelligent fault diagnosis method of a low-rotating-speed mechanical transmission system according to claim 1, characterized in that, The step 3) constructs a mechanical transmission system fault diagnosis model, specifically: 3.1) Constructing a convolutional neural network fault diagnosis model, including a feature extractor and a classifier, optimizing the model structure of the feature extraction part in width, and simplifying the model structure; 3.2) For low speed conditions, use step 1) to construct the vibration gray texture image dataset of the bearing, rotor and gear in the mechanical transmission system under normal and fault conditions, establish the dataset after labeling the vibration gray texture image denoised in step 2), and divide it into a training set and a test set. Input to the fault diagnosis model for training and testing, and use the classification accuracy, precision, recall and F1 score of the fault diagnosis model on the test set as evaluation indicators, and the calculation formula is: In the formula, ACC, precision, recall and F1 represent the accuracy, precision, recall and F1 score of the test set respectively; TP represents the number of samples that are actually positive examples and are correctly predicted as positive examples, TN represents the number of samples that are actually negative examples and are correctly predicted as negative examples; FP represents the number of samples that are actually negative examples but are incorrectly predicted as positive examples; FN represents the number of samples that are actually positive examples but are incorrectly predicted as negative examples.

7. A multi-component intelligent fault diagnosis system for low speed mechanical drive systems, characterized by, It includes: A mechanical transmission system vibration gray texture image dataset construction module: collect the vibration acceleration and displacement signals of the bearing, rotor and gear in the mechanical transmission system under normal and fault conditions at low speed, segment and splice the vibration acceleration and displacement signals into feature matrices according to the specified number of points, then perform standardization, equal proportion scaling and robust local binary processing in turn, generate the corresponding vibration gray texture image, and construct the mechanical transmission system vibration gray texture image dataset from all generated images. The vibration gray texture image dataset is divided into a training set and a test set according to a predetermined proportion; The self-encoder vibration grayscale texture image denoising model construction module is configured to use low-speed working condition data to construct a vibration grayscale texture image data set of the low-speed working condition through the operation of the mechanical transmission system vibration grayscale texture image data set construction module, and use the vibration grayscale texture image data set of the low-speed working condition to construct a self-encoder vibration grayscale texture image denoising model, wherein the self-encoder denoising model comprises multi-scale convolution, Ghost convolution and ECB convolution, the training set is used for model training, and the test set is used for model testing, and the peak signal-to-noise ratio and the structural similarity index are used to evaluate the denoising effect of the denoising model. The mechanical transmission system fault diagnosis model construction module is configured to construct a convolutional neural network fault diagnosis model comprising a feature extractor and a classifier, to optimize the width of the model structure of the feature extraction part, to simplify the model structure, to construct vibration grayscale texture image data sets of bearings, rotors and gears in the mechanical transmission system under normal and fault conditions for low-speed working condition data, to establish a fault feature data set by labeling the vibration grayscale texture image samples denoised by the denoising model, and to divide the fault feature data set into a training set and a test set to be input into the fault diagnosis model for training and testing, and to use the classification accuracy, precision, recall rate and F1 score of the fault diagnosis model on the test set as evaluation indexes to complete the fault diagnosis of the mechanical transmission system under low-speed working condition and noise interference.

8. The multi-component intelligent fault diagnosis system of a low-rotating-speed mechanical transmission system according to claim 7, characterized in that, The construction process of the mechanical transmission system vibration grayscale texture image data set is as follows: 1.1) Collect vibration acceleration and displacement signals of bearings, rotors and gears in the mechanical transmission system under normal and fault conditions at low speed, segment and splice the vibration acceleration and displacement signals according to a specified number of points respectively to form a feature matrix; 1.2) Standardize the elements of the feature matrix to 0-255 to convert them into grayscale images, and the conversion relationship between the element values of the feature matrix and the pixel values of the grayscale images is: wherein, i , j represents the position coordinate in the feature matrix or the gray image, is the value of the element located at i , j ) in the feature matrix, X max and X min represent the maximum value and the minimum value in the feature matrix, respectively, is the gray value of the corresponding position after transformation; 1.3) Scale the converted grayscale images proportionally to 256x256 pixels in size; 1.4) Perform robust local binary processing on the images obtained after proportional scaling to generate vibration grayscale texture images, and introduce weighted local grayscale values in the robust local binary processing to adjust the weight of the central pixel on the threshold selection; In the formula, the gray value of the center pixel, the gray value of the adjacent pixel on the circle with a radius of R the gray value of the adjacent pixel on the circle with a radius of a constant, the greater the value, the greater the proportion of the center threshold. After completing the threshold selection, the texture features of the robust local binary description image are calculated in the order of the size of the circular pixel region in the image, and the calculation formula of the robust local binary is: wherein p denotes the total number of field points selected on the circumference, , denotes the gray value of the field point, RLBP P, R is the RLBP value of the center point after processing by the RLBP operator, ; 1.5) Add different degrees of Gaussian noise to all the obtained vibration grayscale texture images to construct a noisy grayscale texture image data set, and divide the data set into a training set and a test set according to a predetermined proportion.

9. The multi-component intelligent fault diagnosis system of a low-rotating-speed mechanical transmission system according to claim 7, characterized in that, The self-encoder vibration grayscale texture image denoising model is constructed as follows: 2.1) Construct a self-encoder vibration grayscale texture image denoising model comprising multi-scale convolution, Ghost convolution and ECB convolution; 2.2) Use step 1) to construct the bearing, rotor and gear in the mechanical transmission system in low speed working condition under normal and fault state vibration gray texture image data set, according to the proportion of 8:2 division training set and test set, add three different degree of Gaussian noise with standard deviation of 15, 25, 50 to the data set, update the model hyperparameter setting to achieve the optimal denoising effect, and evaluate the denoising effect of the denoising model by peak signal-to-noise ratio and structural similarity index.

10. The multi-component intelligent fault diagnosis system of a low-rotating-speed mechanical transmission system according to claim 7, characterized in that, The mechanical transmission system fault diagnosis model is constructed, specifically: 3.1) Construct a convolutional neural network fault diagnosis model, which includes a feature extractor and a classifier. Optimize the width of the model structure in the feature extraction part and simplify the model structure. 3.2) For low speed working condition, use step 1) to construct the bearing, rotor and gear in the mechanical transmission system in normal and fault state vibration gray texture image data set, label the vibration gray texture image after denoising in step 2) to establish the data set, and divide it into training set and test set, input to the fault diagnosis model for training and testing, and take the classification accuracy, precision, recall and F1 score of the fault diagnosis model on the test set as evaluation indexes, whose calculation formula is: In the formula, ACC, precision, recall and F1 represent the accuracy, precision, recall and F1 score of the test set respectively; TP represents the number of samples that are actually positive examples and are correctly predicted as positive examples; TN represents the number of samples that are actually negative examples and are correctly predicted as negative examples; FP represents the number of samples that are actually negative examples but are incorrectly predicted as positive examples; FN represents the number of samples that are actually positive examples but are incorrectly predicted as negative examples.

Images