Fault Diagnosis Method and System for Two-Dimensional Imaging of Belt Conveyor Idler Sound Signals

Abstract

The present invention provides a fault diagnosis method and system for two-dimensional imaging of the sound signal of a belt conveyor idler, which relates to the field of underground coal mine equipment. The method performs noise reduction preprocessing on the acoustic vibration signal based on a wavelet threshold denoising model of a dual-parameter three-segment threshold function, improving the generalization ability of the denoising model; an acoustic signal imaging method based on MTF image coding, and on this basis, by combining the Laplace algorithm and mean pixel enhancement, the time-domain characteristics of the acoustic signal are incorporated into the two-dimensional imaging process of the signal, improving the effectiveness of fault texture feature extraction; combining the RegNet network model and the Biformer attention mechanism to build a lightweight two-dimensional convolutional neural network model to achieve low-power consumption, high efficiency, and high-precision classification of idler fault types.

Description

Technical Field

[0001] The present invention relates to the technical field of underground coal mine equipment. Specifically, it relates to a fault diagnosis method and system for two-dimensional imaging of the acoustic signal of a belt conveyor idler. Background Art

[0002] The intelligent construction of coal mines is the only way for the high-quality development of coal mines. During the coal mine transportation process, intelligent timing and quantitative inspection of belt conveyors is one of the key measures to ensure operation safety, and its stable and reliable operation is directly related to the efficiency of coal production safety. Traditional fault detection usually relies on manual inspection, which is time-consuming and laborious, and is easily affected by human subjective factors, resulting in low detection efficiency. At the same time, traditional fault detection mainly diagnoses by analyzing signals such as the rotation speed, torque, and vibration of idlers. If traditional vibration acceleration sensors are used to collect their vibration data, two problems will be faced: one is that a large number of sensors need to be arranged, which will increase the difficulty of sensor maintenance and the arrangement cost; the other is that the amount of data generated by numerous sensors is huge, posing new challenges to the computer hardware system in the process of processing these complex data.

[0003] Traditional contact vibration sensors can no longer meet the realization of intelligent inspection processes. In addition, due to the serious interference of underground environmental noise, the effect of fault analysis of belt conveyor idlers by collecting acoustic signals is limited. Therefore, when performing intelligent inspection on idlers, how to improve the noise reduction ability of data preprocessing during the inspection process and ensure the accuracy of diagnosis results has also become an important challenge at the present stage. Summary of the Invention

[0004] To solve the above problems, an embodiment of the present invention provides a fault diagnosis method for two-dimensional imaging of the acoustic signal of a belt conveyor idler. The method includes: acquiring the acoustic vibration signal when the belt conveyor idler is working; performing noise reduction preprocessing on the acoustic vibration signal based on a wavelet threshold denoising model of a dual-parameter three-segment threshold function; performing Markov image coding on the noise reduction preprocessed acoustic vibration signal to obtain a coded image, and performing image enhancement processing on the coded image; the image enhancement processing uses the Laplace algorithm and the mean pixel enhancement algorithm; inputting the image after image enhancement processing into a pre-trained double-layer routing attention mechanism Biformer-RegNet idler fault diagnosis model to obtain a fault diagnosis result; the Biformer-RegNet idler fault diagnosis model introduces a Biformer attention mechanism in the Block module of the RegNet network, and the Biformer-RegNet idler fault diagnosis model is trained by the training set and test set divided from the image after image enhancement processing.

[0005] In the embodiment of the present invention, a wavelet threshold denoising model based on a dual-parameter three-segment threshold function is used to perform denoising preprocessing on the acoustic vibration signal, improving the generalization ability of the denoising model; an acoustic signal imaging method based on MTF image coding, and on this basis, by combining the Laplace algorithm and pixel averaging enhancement, the time-domain characteristics of the acoustic signal are integrated into the two-dimensional imaging process of the signal, improving the effectiveness of fault texture feature extraction; the RegNet network model and the Biformer attention mechanism are combined to build a lightweight two-dimensional convolutional neural network model to achieve low-power, high-efficiency, and high-precision classification of idler fault types.

[0006] Optionally, the image enhancement processing of the encoded image includes: performing second-order spatial differentiation on the pixel values of the encoded image according to the Laplace algorithm, or performing a convolution operation on the encoded image.

[0007] In the embodiment of the present invention, the image coding method is improved, and image enhancement processing based on the Laplace algorithm is added, integrating the time-domain characteristic parameters of the acoustic signal into the two-dimensional imaging process of the signal.

[0008] Optionally, the image enhancement processing of the encoded image includes: calculating the time-domain characteristic parameters of the acoustic signal to obtain the time-domain mean of the acoustic signal; using the time-domain mean as a magnification factor to expand the intensity of each channel of each pixel point to obtain the pixel values of each pixel point; normalizing the pixel values of each pixel point and then redistributing each channel value between (0, 255) to obtain new pixel values and reconstructing an image according to the new pixel values.

[0009] In the embodiment of the present invention, image enhancement processing based on pixel averaging enhancement is added, integrating the time-domain characteristic parameters of the acoustic signal into the two-dimensional imaging process of the signal.

[0010] Optionally, the Biformer-RegNet idler fault diagnosis model includes: an input layer CBR, a backbone layer Body, and an output layer RegHead; the backbone layer Body includes a plurality of Restage modules connected in sequence, and the Restage module includes at least one BotteleNeck_x sub-module, and a Biformer sub-module is added in the BotteleNeck_x sub-module.

[0011] In the embodiment of the present invention, the RegNet network model and the Biformer attention mechanism are combined to improve the diagnostic recognition accuracy of the model.

[0012] Optionally, the BotteleNeck_x sub-module includes: a 3×3 depth convolution sub-module, a double-layer routing attention sub-module, a normalization layer, and a multi-layer perceptron.

[0013] In the embodiment of the present invention, a Biformer sub-module is added to the BotteleNeck_x sub-module, which improves the accuracy of model diagnosis and recognition.

[0014] Optionally, the output layer RegHead includes a 1×1 convolution layer, a GAP (Global average pooling) layer, a Dropout layer, and a fully connected layer connected in sequence; the fully connected layer is used to finally output the working state classification label.

[0015] Optionally, the wavelet threshold denoising model based on the double-parameter three-segment threshold function performs denoising preprocessing on the acoustic vibration signal, including: performing multi-resolution decomposition on the acoustic vibration signal by means of wavelet decomposition, and using the Mallat tower algorithm to perform downsampling decomposition on the signal; selecting a threshold according to the noise level and signal characteristics, and performing zeroing or proportional reduction on the wavelet coefficients according to the threshold.

[0016] In the embodiment of the present invention, a new wavelet threshold denoising algorithm is adopted to highlight the useful components of the signal, optimize the selection of wavelet basis functions, thresholds, and decomposition layers on the basis of retaining the advantages of traditional hard and soft threshold function denoising, and perform preprocessing of effective signal denoising in a strong noise background.

[0017] The embodiment of the present invention provides a fault diagnosis system for two-dimensional imaging of the sound signal of a belt conveyor idler, which is used to execute any one of the above-mentioned fault diagnosis methods for two-dimensional imaging of the sound signal of a belt conveyor idler.

[0018] Optionally, it includes: a login system module and a fault diagnosis module; the login system module is used to set usage permissions, and can timely feedback error information to the login interface when the username or password is entered incorrectly. After a user with usage permissions successfully logs in, it enters the main fault diagnosis interface; the fault diagnosis module is configured with a trained Biformer-RegNet network model, which can identify and classify pictures when pictures are imported into the fault diagnosis module, and display the recognition results at a specified position on the interface.

[0019] The fault diagnosis system for two-dimensional imaging of the sound signal of a belt conveyor idler provided by the embodiment of the present invention can achieve the same technical effects as the above-mentioned fault diagnosis method for two-dimensional imaging of the sound signal of a belt conveyor idler. Description of the Drawings

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only the embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained according to the provided drawings.

[0021] Figure 1 It is the fault diagnosis method architecture for two-dimensional imaging of the sound signal of the idler of the belt conveyor provided by the embodiment of the present invention;

[0022] Figure 2 It is the schematic flowchart of a fault diagnosis method for two-dimensional imaging of the sound signal of the idler of the belt conveyor in the embodiment of the present invention;

[0023] Figure 3 It is the schematic diagram of the MTF image coding of the vibration signal in the embodiment of the present invention;

[0024] Figure 4 It is the schematic diagram of the MTF image coding of the acoustic signal in the embodiment of the present invention;

[0025] Figure 5 It is the schematic flowchart of the pixel enhancement algorithm of mean value in the embodiment of the present invention;

[0026] Figure 6 It is the Biformer-RegNet model structure in the embodiment of the present invention;

[0027] Figure 7 It is the spatial structure diagram of the RegNet model in the embodiment of the present invention;

[0028] Figure 8 It is the schematic structure diagram of the Block in the RegNet model in the embodiment of the present invention;

[0029] Figure 9 It is the schematic diagram of the classification process of the RegNet model in the embodiment of the present invention;

[0030] Figure 10 It is the Biformer network model structure diagram in the embodiment of the present invention;

[0031] Figure 11 It is the Biformer attention mechanism structure diagram in the embodiment of the present invention;

[0032] Figure 12 It is the functional structure diagram of the idler fault diagnosis system in the embodiment of the present invention. Detailed implementation manners

[0033] To make the above objects, features, and advantages of the present invention more obvious and understandable, the following will describe in detail the specific embodiments of the present invention with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0034] To solve the above technical problems, the present invention provides a fault diagnosis method for two-dimensional imaging of the acoustic signal of a belt conveyor idler, which has the following characteristics:

[0035] (1) Aiming at the non-linear and non-stationary characteristics of the acoustic signal during the operation of the idler, a new wavelet threshold denoising algorithm is adopted to highlight the useful components of the signal. On the basis of retaining the advantages of traditional hard and soft threshold function denoising, the selection of wavelet basis functions, thresholds, and decomposition levels is optimized, and preprocessing of effective signal denoising is realized under a strong noise background.

[0036] Specifically, according to the existing environment in the coal mine, a wavelet threshold denoising model based on a two-parameter three-segment threshold function is used to realize the preprocessing of acoustic signal denoising under complex environmental conditions.

[0037] (2) The image coding method of one-dimensional signals is introduced into the feature extraction of acoustic signals and vibration signals, and image enhancement processing is carried out by using the Laplace algorithm and time-domain feature fusion method; fault texture features are extracted from the perspective of the image domain in a Convolutional Neural Networks (CNN) model, and the pictures are used as the training set and test set of the CNN model to verify the effectiveness and reliability of this image processing method.

[0038] Optionally, a two-dimensional image coding method based on Markov Transition Field (MTF) is used to realize the feature dimension elevation of the acoustic signal; on this basis, image enhancement processing is carried out by using the Laplace algorithm and time-domain feature fusion method, and the time-domain features of the acoustic and vibration signals are integrated into the two-dimensional image of the signal to improve the effectiveness of fault texture feature extraction. In this embodiment, the standard MTF image coding method is not used, but it is improved.

[0039] (3) Build a lightweight two-dimensional convolutional neural network model to achieve low-power, high-efficiency, and high-precision classification of idler fault types.

[0040] Optionally, the RegNet network model and the Biformer attention mechanism are combined, and the acoustic signal coding images of the idler under four working conditions are used as the data set to build and train a RegNet two-dimensional convolutional neural network model based on the Biformer attention mechanism, and a fault diagnosis system for the idler is built on the basis of the trained model.

[0041] Figure 1 This is the architecture of the fault diagnosis method for the two-dimensional imaging of the belt conveyor idler acoustic signal provided by the embodiments of the present invention. As Figure 1 shown, this method includes three parts: a noise reduction method based on a dual-parameter three-segment wavelet threshold function, a signal feature transformation method based on MTF imaging coding, and a fault diagnosis method for idlers based on the Biformer-RegNet model.

[0042] Among them, the noise reduction method based on the dual-parameter three-segment wavelet threshold function includes: analyzing the influencing factors of wavelet threshold noise reduction, theoretical verification of the dual-parameter three-segment wavelet threshold function and noise reduction simulation experiments, and finally verifying the noise reduction ability of the BT-WTD algorithm.

[0043] The signal feature transformation method based on MTF imaging coding includes: time-frequency domain feature extraction, acoustic signal imaging based on the Markov transfer field, and image enhancement processing integrating time-domain features. Finally, a training set is constructed to train the Biformer-RegNet model.

[0044] The fault diagnosis method for idlers based on the Biformer-RegNet model includes: analysis of the idler fault diagnosis results and model optimization based on the Biformer attention mechanism, and finally realizing the design of the idler fault diagnosis system based on the RegNet model. Based on this idler fault diagnosis system based on the RegNet model, lightweight fault diagnosis of the belt conveyor idlers is realized.

[0045] Figure 2 This is a schematic flowchart of a fault diagnosis method for the two-dimensional imaging of the belt conveyor idler acoustic signal in the embodiments of the present invention. This method includes the following steps:

[0046] S202, acquiring the acoustic and vibration signals when the belt conveyor idler is working.

[0047] For the fault diagnosis of the belt conveyor, the acoustic and vibration signals when its idler is working can be collected. The acoustic and vibration signals can include acoustic signals and vibration signals.

[0048] S204, performing noise reduction preprocessing on the acoustic and vibration signals using a wavelet threshold noise reduction model based on a dual-parameter three-segment threshold function.

[0049] Among them, the wavelet threshold denoising (BT-WTD) model of the biparameter and trisegment can flexibly process the acoustic and vibration signals under different noise levels by setting two thresholds and segmenting the wavelet decomposition signals within the range of each threshold. The double adjustment parameters can flexibly adjust the length of the denoising threshold interval and the asymptote of the threshold function in the non-threshold interval, improving the generalization ability of the denoising model.

[0050] Specifically, the acoustic and vibration signals are decomposed by wavelet decomposition in a multi-resolution manner, and the Mallat pyramid algorithm is used to decompose the signals to a lower order. Then, the threshold is selected according to the noise level and signal characteristics, and the wavelet coefficients are set to zero or reduced proportionally according to the threshold.

[0051] The wavelet decomposition uses the Mallat pyramid algorithm to decompose the signal to a lower order. The wavelet decomposition process is to select a wavelet basis function to perform multi-scale wavelet decomposition on the signal to be processed. The decomposition at each scale can be expressed as:

[0052]

[0053] In the formula, N is the number of sampling points; i is the number of decomposition layers; j is the scale measurement space of the filter; h j is the low-pass filter, and g j is the high-pass filter; c i is the low-frequency wavelet coefficient, and b i is the high-frequency wavelet coefficient.

[0054] Many wavelet coefficients can be obtained in each decomposition process. After the useful signal is decomposed by n layers, the wavelet coefficients are slightly larger, that is, the low-frequency approximation component (CA), which is characterized by a larger amplitude and a smaller fluctuation frequency; the wavelet coefficients obtained after the noise decomposition are smaller, that is, the high-frequency detail component (CD), which is characterized by a smaller amplitude and a larger fluctuation frequency. The noise energy is also mainly concentrated in the high-frequency detail components after wavelet decomposition. As the number of decomposition layers of the approximation coefficient (low-frequency approximation component CA) increases, the weak noise signal in the upper layer will be continuously dispersed into the CA of the lower layer, further reducing the noise level in the signal. However, when the number of layers is too high, the useful signal will be misinterpreted as a noise signal and processed, thus affecting the signal reconstruction. To balance both, generally 3 to 5 layers of decomposition are selected.

[0055] The threshold function denoising process is a key step in the wavelet threshold denoising process. Its principle is to select a certain threshold according to the noise level and signal characteristics, and perform zeroing or proportional reduction on wavelet coefficients of different magnitudes to achieve the purpose of filtering out the noise signal.

[0056] In wavelet threshold denoising, the threshold function reflects different processing methods for wavelet coefficients outside the threshold range. After the acoustic signal is decomposed by wavelet, it is necessary to use the threshold function to process the detail coefficients at different decomposition levels to obtain the wavelet estimation coefficients. Therefore, the design of the threshold function is the key to the wavelet threshold denoising method.

[0057] Aiming at the deficiencies of the classical soft threshold and hard threshold functions, this paper introduces the exponential function and adjustable parameters α and β into the threshold function, and divides the processing interval into three segments: the proportional reduction interval, the transition reduction interval, and the zeroing interval in the form of a double threshold, and proposes a two-parameter three-segment wavelet denoising threshold function (The Bifactorial and Three-Segment function, TBTS) with the characteristics of continuity, constant deviation, symmetry, and flexibility. The expression of the new threshold function is as follows:

[0058]

[0059] where e is the natural constant, λ and λ 0 are both thresholds, and satisfy λ 0 = βλ, α and β are flexibly adjustable parameters, α ∈ (0, +∞), β ∈ (0, 1).

[0060] This threshold function realizes the flexible adjustment of the threshold interval and the constant deviation by introducing two adjustable parameters: when the wavelet decomposition coefficient is greater than the threshold, it can quickly approach the hard threshold function curve, reduce the constant deviation and avoid signal distortion (proportional reduction interval); when the wavelet decomposition coefficient is between the two thresholds, it can avoid the pseudo-Gibbs problem (transition reduction interval); when the wavelet decomposition coefficient is less than the threshold, the wavelet coefficient is set to zero for noise components (zeroing interval), so as to retain more useful signal features, which not only enhances the correlation between wavelet coefficients of the soft threshold function, but also makes up for the discontinuity defect of the hard threshold function at the threshold point.

[0061] In the improved threshold function, the two flexibly adjustable parameters introduced are the slope adjustment parameter α and the threshold range adjustment parameter β. When the threshold range adjustment parameter β takes 0.95, as the value range of α increases from 0 to positive infinity, the asymptote of the improved threshold function approaches y = x, which is the same as the asymptote of the hard threshold function, greatly reducing the deviation of wavelet coefficients outside the threshold range in the wavelet denoising process.

[0062] When the fixed α = 2 and the threshold range adjustment parameter β varies from 0 to 1, the zero-setting interval of the threshold function increases from small to large, and the processing ability of the improved threshold function for wavelet coefficients within a given threshold range decreases from strong to weak. Therefore, by adjusting the range of the adjustment parameter β, it is possible to flexibly process noise signals in various frequency bands, effectively retain the useful features of the original signal during the noise reduction process, and minimize the noise signal to the greatest extent.

[0063] S206, perform Markov image coding on the vibration and acoustic signals after noise reduction preprocessing to obtain the encoded image, and perform image enhancement processing on the encoded image.

[0064] The Markov transfer field MTF is a time series image coding method based on the Markov chain. This method regards the change of the time series as a Markov process, that is, based on the known characteristics of the current signal, the change of the future signal time-domain waveform does not depend on the past characteristics. The principle of the Markov transfer field image representation is based on the first-order Markov chain. By using the transfer matrix to represent the Markov transfer probabilities of the acoustic signal and the vibration signal in the time domain range in turn to retain the original time-domain information of the signal, and at the same time using the transition probability of the amplitude quantile interval to analyze the dynamic characteristics of the time series signal, finally realizing the conversion of the one-dimensional time series signal into a two-dimensional image representation.

[0065] When the one-dimensional acoustic signal or vibration signal time series is processed, a Markov matrix is finally obtained. This matrix reflects the probability transfer relationship between each quantile interval. The expression is shown in the following formula:

[0066]

[0067] Among them, w ij (i, j ∈ {1, 2, …, Q}) represents the probability that the element in q j is followed by the element in q i . Its size is determined by the frequency of the data amplitude neighbors in q j . The calculation formula is shown in the following formula:

[0068] w ij = P(x t ∈ q i |x t-1 ∈ q j )

[0069] After normalizing the elements in each row, the obtained transition matrix W is a Markov matrix. However, this way of constructing the matrix ignores the conditional relationship between the time series distribution and the time step dependence. Removing the time dependence will result in the loss of a large amount of signal - contained information. To overcome this limitation, according to the order of the time series, each calculated transition probability is rearranged, thereby expanding the original Markov matrix. Through this process, a matrix of size [N, N] can be obtained, which is called the Markov transition field matrix M. This matrix records the complete probability information of each possible state transition in the time series in more detail, and its expression is shown in the following formula:

[0070]

[0071] Among them, the quantile intervals q i and q j are the data at time steps i and j respectively; M ij is the transition probability from q i to q j .

[0072] The Markov transition field matrix obtained through this reconstruction method makes up for the shortcomings of the original transition matrix through time arrangement and has time correlation. The two - dimensional image constructed by the Markov transition field matrix coding method can describe the dynamic changes of time and frequency of one - dimensional signals, making the constructed picture dataset have generalization performance.

[0073] Perform Markov image coding on one - dimensional acoustic signals and vibration signals respectively, Figure 3 shows the schematic diagram of MTF image coding of vibration signals, Figure 4 shows the schematic diagram of MTF image coding of acoustic signals. Figure 3 、 Figure 4 Shown respectively are the images of acoustic signals and vibration signals after coding under four working conditions (normal condition, idler jam, bearing abnormality, and support frame abnormality),

[0074] Since the images of vibration signals under four different working conditions are all different, but the images of acoustic signals after coding are relatively similar under normal condition and idler jam condition, and the images of bearing abnormality and support frame abnormality conditions are relatively similar. Therefore, in order to better realize the fault image recognition and classification of idler acoustic signals, it is still necessary to perform enhancement processing on the coded images.

[0075] Specifically, the image enhancement processing can adopt the Laplace algorithm and the mean - pixel enhancement algorithm.

[0076] Among them, Laplace algorithm is adopted for image enhancement processing, including: performing second-order spatial differentiation on the pixel values of the encoded image according to the Laplace algorithm, or performing a convolution operation on the encoded image. The two-dimensional image enhancement method can utilize the Laplace algorithm to enhance the image edges by performing second-order spatial differentiation on the pixel values of the image, or enhance the high-frequency details in the image through image convolution operation to improve the image clarity and edge information.

[0077] Among them, the average pixel enhancement algorithm is adopted for image enhancement processing, including: calculating the time-domain characteristic parameters of the acoustic signal to obtain the time-domain average value of the acoustic signal; using the time-domain average value as a magnification factor to expand the intensity of each channel of each pixel point to obtain the pixel value of each pixel point; performing normalization processing on the pixel values of each pixel point and then redistributing each channel value between (0, 255) to obtain new pixel values and reconstructing the image according to the new pixel values.

[0078] Figure 5 The flowchart of the average pixel enhancement algorithm is shown. As Figure 5 shown, for the MTF image, first calculate the channel values (R channel, G channel, and B channel) of each pixel point, calculate the time-domain characteristic parameters of the acoustic signal to obtain the time-domain average value of the signal, and use this average value as a magnification factor to expand the intensity of each channel at the same time. Then, after normalizing the pixel values, redistribute each channel value between (0, 255) to obtain new pixel values and then reconstruct a new MTF image. If the amplified channel value is greater than 255, normalization processing is performed. If the amplified channel value is less than 255, it is directly output. Reconstruct the MTF image based on the output pixel values.

[0079] After Figure 5 the above processing flow, it is possible to only enhance the magnitude of the image pixel values on the basis of using the time-domain characteristic parameters of the signal without changing the contrast between pixel points, and still be able to retain the time-frequency domain characteristics of the acoustic signal after image encoding to a certain extent.

[0080] S208, input the image after image enhancement processing into the pre-trained Biformer-RegNet idler fault diagnosis model to obtain the fault diagnosis result. The Biformer-RegNet idler fault diagnosis model is trained by the training set and test set divided from the image after image enhancement processing.

[0081] The Biformer-RegNet idler fault diagnosis model is characterized in that a Biformer double-layer routing attention mechanism is introduced into the Block module of the RegNet network.

[0082] In this embodiment, the RegNet and the transformer network based on double-layer routing attention are first analyzed, then the Biformer attention mechanism is introduced into the Block module of the RegNet network, and the model is trained and optimized through the existing data set, and the results are compared with other lightweight networks; finally, a idler fault diagnosis system is built on the optimized model.

[0083] Specifically, the Biformer-RegNet idler fault diagnosis model includes: an input layer CBR, a backbone layer Body, and an output layer RegHead.

[0084] Among them, the input layer CBR refers to the combination of convolution (Conv), batch normalization (Batch Normalization, BN), and ReLU activation function (ReLU). Convolution is used to extract the features of the input data, batch normalization is used to accelerate the training process and improve the generalization ability of the model, and the ReLU activation function is used to increase the nonlinearity of the model and improve the expression ability of the model.

[0085] The backbone layer Body includes a plurality of Restage modules connected in sequence. The Restage module includes at least one BotteleNeck_x sub-module, and a Biformer sub-module is added to the BotteleNeck_x sub-module. The above BotteleNeck_x sub-module includes: a 3×3 depth convolution sub-module, a double-layer routing attention sub-module, a normalization layer, and a multi-layer perceptron. In this embodiment, the Biformer attention mechanism is introduced into the Block module of the RegNet network. Specifically, the Biformer attention mechanism is added to the BotteleNeck_x sub-module in Restage. The Biformer sub-module located after the grouped convolution can focus on the feature distribution of the acoustic signal image and improve the feature extraction ability of the model.

[0086] The output layer RegHead includes a 1×1 convolution, a GAP layer, a Dropout layer, and a fully connected layer connected in sequence; the fully connected layer is used to finally output the working state classification label.

[0087] The fault diagnosis method for two-dimensional imaging of the sound signal of the idler of the belt conveyor provided by the embodiment of the present invention performs noise reduction preprocessing on the vibration signal based on the wavelet threshold denoising model of the double-parameter three-segment threshold function, which improves the generalization ability of the denoising model; based on the MTF image coding method for acoustic signal imaging, and on this basis, by combining the Laplace algorithm and mean pixel enhancement, the time-domain characteristics of the acoustic signal are integrated into the two-dimensional imaging process of the signal, which improves the effectiveness of fault texture feature extraction; combining the RegNet network model and the Biformer attention mechanism to build a lightweight two-dimensional convolutional neural network model to achieve low-power, high-efficiency, and high-precision classification of the idler fault types.

[0088] Using traditional feature extraction methods to analyze the preprocessed signal in the time domain and frequency domain, some time-domain parameter indicators such as peak value and root mean square value can be used to preliminarily judge the time-domain characteristics of the signal under different working conditions. Traditional feature extraction methods can simply judge and classify the above four working conditions. By analyzing the time-domain and frequency-domain results, it is known that the traditional time-domain and frequency-domain feature extraction has limitations and is not suitable for the intelligent inspection process of future belt conveyor inspection robots. Therefore, the embodiment of the present invention provides a one-dimensional signal imaging method based on MTF image coding, integrates the time-domain characteristic parameters of the acoustic signal into the two-dimensional imaging process of the signal, and uses the CNN model for recognition and classification testing.

[0089] For the imaging feature recognition of the acoustic signal of the idler working state, in this embodiment, the Biformer attention mechanism is introduced on the basis of the RegNet network model to build an improved RegNet network recognition model RegNet-Biformer. Figure 6 Shows the Biformer-RegNet model structure in the embodiment of the present invention.

[0090] In the input module CBR_x (x represents the number of CBR modules in series), the input image set is processed through three sub-modules: the Conv2d two-dimensional convolutional layer, the BN batch normalization layer, and the ReLU activation function layer to obtain the feature image set; in the Body layer composed of four Restage modules, the feature image set goes through four Restage stages in sequence. The Biformer attention mechanism is added to the BotteleNeck_x sub-module in Restage. The Biformer sub-module located after the grouped convolution can focus on the feature distribution of the acoustic signal image and improve the feature extraction ability of the model; in the RegHead layer, the feature image set processed by the Body layer goes through a 1×1 convolution, global average pooling, Dropout layer, and fully connected layer in sequence, and finally outputs the model classification label, that is, the idler working state type.

[0091] Figure 7The spatial structure diagram of the RegNet model in the embodiments of the present invention is shown. When the RegNet network processes different classification tasks, different network model structures need to be used. The main approach is to use adjustment parameters to change the model distribution, so as to achieve the purpose of adjusting the model structure. Figure 7 The parameters involved include: the number of times d of the Block in each RegNet Stage i , the number of channels w of the output matrix i and the width g of each Group in the Block.

[0092] In the RegNet model structure, as shown in Table 1, the specific input, output sizes and structure types of each layer of the model are given, where GAP is global average pooling, Dropout is the random inactivation ratio of neuron functions during the training process of the model, and FC is the fully connected layer of the model. The main body of the spatial network is mainly composed of the input layer Steam, the backbone layer Body and the output layer RegHead.

[0093]

[0094] Table 1

[0095] The Steam layer is used to preprocess the input picture set of different sizes. The size of the training set is adjusted to an input set of 3×224×224 dimensions through methods such as central cropping, padding around the edges and triple cropping. The input image undergoes a normal two-dimensional convolution with a stride of 2 and a convolution kernel size of 3×3, and outputs a feature map of 32×112×112 dimensions. The backbone layer Body improves the model capacity through the accumulation of four Regstages, and can extract the two-dimensional data features of signals under four working conditions, enhancing the representation ability of the model.

[0096] Figure 8 The structural schematic diagram of the Block in the RegNet model in the embodiments of the present invention is shown. Each Regstage is stacked by multiple Block layers. The main branches of the Block layer are all grouped convolutions of 1×1 and 3×3; when the stride is equal to 1 in the shortcut branch, no processing is done, and when the stride is equal to 2, a 1×1 convolution is added to downsample in the shortcut branch, as Figure 8 shown (the left figure is the structure with a stride size equal to 1, and the right figure is the structure with a stride equal to 2). The output layer RegHead is composed of three structures: 1×1 convolution, GAP, Dropout and FC, and its function is to classify and diagnose the feature information of the working state of the idler.

[0097] Figure 9The figure shows a schematic diagram of the classification process of the RegNet model in an embodiment of the present invention. The picture input set goes through four Restage stages in sequence and finally outputs classification labels.

[0098] In the RegNet model structure, the widths and depths of the Block modules are given by a fitted linear function. Its expression is:

[0099] u j = w 0 + w a ·j, 0 ≤ j < d

[0100] where the initial width w 0 > 0 of the model, the slope w a > 0 of the fitted straight line of the number of channels change, the depth of the network is d, and the u value calculated by the above formula 5.1 is the number of channels of each layer.

[0101] Considering that the number of channels of the layer structure of each Block module in the training model should be equal integers, so it is necessary to introduce hyperparameters w m and intermediate variable s j , and calculate the intermediate variable of the j-th network layer to satisfy the conditions of the following formula.

[0102]

[0103] To express u j as an integer, round the s j in the above formula and express it as Substituting it in, it can be known that the number of channels calculated therefrom is:

[0104]

[0105] For further quantization calculation of the number of channels u j , for each Block module, their widths are kept consistent, and the depth of the layer (the number of Blocks) is equal to the number of i, as follows:

[0106]

[0107] It can be seen from this that the parameterization of the RegNet model space is to use grid search and automatically optimize d, w 0 , w a and w m , and the value range of the parameters is: d < 64, w 0 < 256, w a < 256 and 0 ≤ w m ≤ 3. The classification process of the input set for the RegNet network model is asFigure 9 as shown

[0108] Figure 10 It shows the structural diagram of the Biformer network model in an embodiment of the present invention. The core of the Biformer network model lies in the attention calculation in two stages: the first stage is to divide the input image into multiple coarse-grained regions of a certain size, perform self-attention within this region, then calculate the correlation between adjacent regions and obtain their relationship matrix, and then sparsify these matrices and retain the K elements with the largest values; the second stage is to further perform self-attention calculation on the fine-grained regions based on the coarse-grained sparse matrix.

[0109] Its network structure is as Figure 10 shown. The input image dataset passes through 4 feature extraction stages from left to right, and after multiple linear mappings and matrix calculations, feature layers with different widths and heights are obtained. It mainly consists of four modules: a 3×3 depth convolution module, a bi-level routing attention (BRA) module, a normalization layer, and a multi-layer perceptron. The core module is the BRA module.

[0110] Figure 11 It shows the structural diagram of the Biformer attention mechanism in an embodiment of the present invention. BRA is essentially a dynamic sparse attention mechanism, which realizes the computational sparsification of the attention module through the bi-level routing method and realizes the resource allocation of the computer and the perception of the picture content. Since it was first proposed and applied in the Biformer network model, it is also called the Biformer attention mechanism. Its core idea is to filter out the irrelevant key-value pairs in the coarse-grained regions of the picture input set, and realize the screening by constructing and pruning the region-level directed graph and the fine-grained token to token attention.

[0111] as Figure 11 shown, the BRA module in the figure is the structural diagram of the Biformer attention mechanism. In the figure, S is the number of non-overlapping regions into which the feature map is divided, Q, K, and V are the linear mappings after the deformation operation of each region, k represents the number of routing regions, A is the affinity adjacency matrix between each region, Q represents the value obtained after the attention operation on K and V, and its calculation method is shown in formula 5.11.

[0112] O = Attention(Q, K g , V g ) + LCE(V)

[0113] where K g , Vg It represents the key-value tensor collected from K and V. LCE(V) is the local context enhancement term. The function LCE(·) is parameterized using depth convolution, and the convolution kernel size is 5.

[0114] In this embodiment, the diagnostic recognition accuracy of the Biformer-RegNet model and other lightweight networks can be compared through experiments. Specifically, datasets with the same sample size are respectively put into the Shufflenet V2 network model, RegNetx_200MF network model, Efficientnet V2 network model, and Mobilenet V3 network model for training. After 50 times of the same training, the model diagnostic recognition accuracy is obtained.

[0115] It can be seen from the experimental results that the network model built in this embodiment has a higher classification accuracy compared with other lightweight convolutional networks. As can be seen from Table 2, although the Biformer-RegNet network model has a higher number of parameters compared with the ShufflenetV2 model, RegNetx_200MF model, and Mobilenet V3 model, the accuracy of fault recognition and classification has increased by 1.99%, 1.48%, and 2.66% respectively; moreover, when compared with the Efficientnet V2 network model with a larger number of parameters and computational complexity, the fault diagnosis accuracy of the Biformer-RegNet model has also increased by 1.15% relatively.

[0116]

[0117] Table 2

[0118] An embodiment of the present invention also provides a fault diagnosis system for two-dimensional imaging of the sound signal of a belt conveyor idler, which is used to execute the above-mentioned fault diagnosis method for two-dimensional imaging of the sound signal of the belt conveyor idler.

[0119] Figure 12 The functional structure diagram of the idler fault diagnosis system provided by the embodiment of the present invention is shown. The idler fault diagnosis system includes two functional modules: a system login module and a fault diagnosis module. The system login module includes the following functions: username verification, password verification, and administrator verification. The fault diagnosis module includes the following functions: picture import, idler fault recognition, and result display.

[0120] Optionally, the system includes: a login system module and a fault diagnosis module;

[0121] The login system module is used to set usage permissions, and can timely feedback error information to the login interface when the username or password is input incorrectly. After a user with usage permissions successfully logs in, they enter the main interface of fault diagnosis;

[0122] The fault diagnosis module is configured with a trained Biformer-RegNet network model, which can identify and classify pictures when pictures are imported into the fault diagnosis module, and display the recognition results at a specified position on the interface.

[0123] The fault diagnosis system for two-dimensional visualization of the sound signal of the idler of the belt conveyor provided in the above embodiment can implement each process in the embodiment of the above method for two-dimensional visualization of the sound signal of the idler of the belt conveyor. To avoid repetition, it will not be elaborated here.

[0124] An embodiment of the present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements each process in the embodiment of the above method for two-dimensional visualization of the sound signal of the idler of the belt conveyor, and can achieve the same technical effects. To avoid repetition, it will not be elaborated here. Among them, the computer-readable storage medium is, for example, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disc, etc.

[0125] Of course, those skilled in the art can understand that all or part of the processes in the above embodiment methods can be completed by a computer program instructing a control device. The program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the above method embodiments. Among them, the storage medium can be a memory, a magnetic disk, an optical disc, etc.

[0126] In this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "including one..." does not exclude the existence of another identical element in the process, method, article or device including the element.

[0127] In this specification, the various embodiments are described in a progressive manner. Each embodiment focuses on the differences from other embodiments. The same or similar parts among the various embodiments can be referred to each other.

[0128] Based on the above description of the disclosed embodiments, those skilled in the art can implement or use the present invention. Various modifications to these embodiments will be obvious to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to these embodiments shown herein, but rather will be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A fault diagnosis method for belt conveyor roller acoustic signal two-dimensional imaging, characterized in that: The method comprises: Obtain the sound and vibration signals of the belt conveyor rollers when they are working; Performing denoising preprocessing on the acoustic vibration signal based on a wavelet threshold denoising model with a two-parameter three-stage threshold function; Performing Markov image coding on the acoustic vibration signal after noise reduction preprocessing to obtain an encoded image, and performing image enhancement processing on the encoded image; the image enhancement processing adopts Laplace algorithm and average pixel enhancement algorithm; Inputting the image after image enhancement processing into a pre-trained Biformer-RegNet roller fault diagnosis model to obtain a fault diagnosis result; the Biformer-RegNet roller fault diagnosis model introduces a Biformer attention mechanism in the Block module of the RegNet network, and the Biformer-RegNet roller fault diagnosis model is trained by a training set and a test set divided by the image after image enhancement processing; The expression of the two-parameter three-stage threshold function is as follows: Among them, ω λ is the wavelet coefficient after denoising preprocessing, ω is the wavelet decomposition coefficient, e is a natural constant, λ and λ0 are both thresholds, and satisfy λ0=βλ, α and β are flexibly adjustable parameters, α∈(0,+∞), β∈(0,1).

2. The method according to claim 1, characterized in that: The performing image enhancement processing on the encoded image comprises: A second-order spatial differentiation is performed on the pixel values ​​of the encoded image according to the Laplace algorithm, or a convolution operation is performed on the encoded image.

3. The method according to claim 1 or 2, characterized in that: The performing image enhancement processing on the encoded image comprises: Calculating the time domain characteristic parameters of the acoustic signal to obtain the time domain mean of the acoustic signal; The intensity of each channel of each pixel point is expanded by using the time domain mean as a multiplication factor to obtain a pixel value of each pixel point; After normalizing the pixel values ​​of each pixel point, each channel value is redistributed between (0, 255) to obtain a new pixel value and reconstruct the image according to the new pixel value.

4. The method according to claim 1, characterized in that: The Biformer-RegNet roller fault diagnosis model includes: an input layer CBR, a backbone layer Body and an output layer RegHead; The backbone layer Body includes a plurality of Restage modules connected in sequence, and the Restage module includes at least one BotteleNeck_x submodule, and a Biformer submodule is added to the BotteleNeck_x submodule.

5. The method according to claim 4, characterized in that The BotteleNeck_x submodule includes: a 3×3 deep convolution submodule, a two-layer routing attention submodule, a normalization layer and a multi-layer perceptron.

6. The method according to claim 4, characterized in that The output layer RegHead includes a 1×1 convolution, a global average pooling layer, a Dropout layer and a fully connected layer connected in sequence; the fully connected layer is used to finally output the working status classification label.

7. The method according to claim 1, characterized in that The wavelet threshold denoising model based on a two-parameter three-segment threshold function performs denoising preprocessing on the acoustic vibration signal, including: The acoustic vibration signal is decomposed by multi-resolution by wavelet decomposition, and the signal is decomposed by reducing order by using Mallat pyramid algorithm; A threshold is selected according to the noise level and signal characteristics, and the wavelet coefficients are zeroed or scaled down according to the threshold.

8. A fault diagnosis system for belt conveyor roller acoustic signals in two-dimensional images, characterized in that: A fault diagnosis method for two-dimensional imaging of belt conveyor roller acoustic signals according to any one of claims 1 to 7; The system comprises: a login system module and a fault diagnosis module; The login system module is used to set the usage authority. When the user name or password is entered incorrectly, the error information can be fed back to the login interface in time. After the user with usage authority successfully logs in, he / she can enter the fault diagnosis main interface; The fault diagnosis module is configured with a trained Biformer-RegNet network model. When an image is imported into the fault diagnosis module, it can recognize and classify the image and display the recognition result at a specified position on the interface.

Images