Stand column hydraulic system fault diagnosis method and system in coal environment

Abstract

The invention discloses a fault diagnosis method and system for a stand column hydraulic system in a coal environment, and belongs to the technical field of intelligent detection of coal mine supporting equipment. The method comprises the following steps: constructing a dust friction correction model and a hydraulic oil temperature-viscosity characteristic model of the stand column hydraulic system in a coal mine operation environment; according to the combination of the dust concentration compensation amount and the temperature compensation viscosity, modeling a porous medium leakage model in the coal mine environment, and solving the leakage rate; constructing time domain fault features; fourier transform is carried out on the time domain fault features, a convolutional neural network into which channel attention and space attention are introduced is input, and an attention convolutional neural network fault diagnosis model combined with Fourier transform is constructed; and inputting the actual hydraulic data into the constructed fault diagnosis model to obtain a fault diagnosis result, thereby realizing fault diagnosis of the column hydraulic system in the coal environment. According to the invention, the accuracy of multi-fault coupling diagnosis is improved by combining the attention convolutional neural network fault diagnosis model of Fourier transform.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent detection of coal mine support equipment, and more specifically, relates to a fault diagnosis method and system for a column hydraulic system in a coal environment. Background Art

[0002] The development of the coal industry has been basically stagnant in recent years, and the technology of its testing center is relatively backward. It can only meet the testing capabilities of routine projects and can no longer meet the current large-scale testing needs of column hydraulic systems.

[0003] The existing hydraulic support one-way valve durability test bench has poor reliability. A single device fails more than 50 times per 100 hours, requires a large amount of maintenance, and has no fault diagnosis function. The troubleshooting time is long, which is difficult to balance with the current work pace. The use of traditional convolutional neural network diagnostic models leads to insufficient generalization performance, failure to integrate frequency domain information, and insufficient feature extraction capabilities under the combined working conditions of dust concentration and temperature changes, resulting in weak fault feature characterization capabilities and a high fault diagnosis misjudgment rate. Therefore, there is an urgent need to develop fault diagnosis for column hydraulic systems in coal environments. Summary of the Invention

[0004] In response to the shortcomings of the existing technology, the present invention provides a fault diagnosis method and system for a pillar hydraulic system in a coal environment, establishes a dust friction correction model and a hydraulic oil temperature-viscosity characteristic model for the pillar hydraulic system, constructs time domain fault features, performs Fourier transform on the time domain fault features, and inputs them into a convolutional neural network that introduces channel attention and spatial attention, constructs a Fourier transform-convolutional block attention module-convolutional neural network fault diagnosis model FFT-CBAM-CNN (Fourier Transform-Convolutional Block Attention Module-Convolutional Neural Network) combined with Fourier transform, and improves the fault diagnosis accuracy and robustness of the coal mine hydraulic system.

[0005] The present invention adopts the following technical solutions.

[0006] A first aspect of the present invention provides a method for diagnosing faults in a column hydraulic system under a coal environment, comprising:

[0007] A dust friction correction model and a hydraulic oil temperature-viscosity characteristic model for the column hydraulic system in a coal mine operating environment were constructed to obtain the dust concentration compensation and temperature-compensated viscosity.

[0008] Based on the combination of dust concentration compensation and temperature compensation viscosity, the porous media leakage model in the coal mine environment is modeled and the leakage rate is solved;

[0009] Construct time domain fault signatures based on temperature compensated viscosity, dust concentration compensation, and leakage rate;

[0010] The time-domain fault features are Fourier transformed and input into a convolutional neural network that introduces channel attention and spatial attention to construct a fault diagnosis model combining Fourier transform and attention convolutional neural network.

[0011] The actual hydraulic data is input into the constructed fault diagnosis model to obtain the fault diagnosis results, thus realizing the fault diagnosis of the column hydraulic system in the coal environment.

[0012] Preferably, the construction of the dust friction correction model for the column hydraulic system in the coal mine operation environment specifically includes:

[0013] Test the standard friction coefficient μ_0 without dust interference;

[0014] The dust concentration data C_d of the working area is continuously collected through the dust sensor array in the column hydraulic system;

[0015] Set the gain coefficient and multiply it by the dust concentration data C_d to compensate it and obtain the dust concentration compensation amount.

[0016] Preferably, the construction of the hydraulic oil temperature-viscosity characteristic model of the column hydraulic system under the coal mine operation environment specifically includes:

[0017] Test anti-wear hydraulic oil samples, collect viscosity-temperature curve data within a set temperature range, and use the least squares method to fit and determine the equation coefficients A and B;

[0018] An explosion-proof PT100 temperature sensor is placed on the outer wall of the hydraulic cylinder in the column hydraulic system. The actual oil temperature is acquired at a sampling frequency of 0.5 Hz. The actual oil temperature is pre-processed using a Kalman filter to eliminate abnormal values ​​caused by transient thermal shock of the tunneling machinery and obtain the measured temperature value T.

[0019] Calculate the logarithm of the measured temperature value T, lgT;

[0020] Finally, the temperature-compensated viscosity v(T)=1010(A-B1gT) is generated by a power function of 10.

[0021] Preferably, the method of modeling a porous medium leakage model in a coal mine environment based on the dust concentration compensation amount and the temperature compensation viscosity to solve the leakage rate specifically includes:

[0022] The pores are corrected according to the dust concentration compensation, and the seepage area is solved by combining the geometric area of ​​the sealing surface;

[0023] According to the contamination level and the decrease in permeability coefficient, the permeability coefficient of the contaminated oil is solved;

[0024] The temperature correction factor is obtained by multiplying the ratio of the standard viscosity to the temperature-compensated viscosity by the collected oil temperature.

[0025] Calculate the driving leakage pressure difference per meter of seepage path length under unit viscous resistance;

[0026] The leakage rate is obtained by multiplying the seepage area, the permeability coefficient of the contaminated oil, the temperature correction factor, and the driving leakage pressure difference per meter of seepage path length under unit viscous resistance.

[0027] Preferably, the porosity is corrected according to the dust concentration compensation amount, and the seepage area is solved in combination with the geometric area of ​​the sealing surface, which specifically includes:

[0028] Obtain the pore structure image of the sealing surface, extract the pore area according to the binarization algorithm, and calculate the initial porosity according to the ratio of pore volume to total volume;

[0029] Subtract the dust concentration compensation from 1 and multiply it by the initial porosity to generate the porosity correction value;

[0030] The seepage area is obtained by multiplying the porosity correction value by the geometric area of ​​the sealing surface in the absence of dust.

[0031] Preferably, solving the driving leakage pressure difference per meter of seepage path length under unit viscous resistance specifically includes:

[0032] The driving leakage pressure difference is obtained by solving the energy loss of the hysteresis loop formed by the basic pressure difference and dust accumulation;

[0033] The dynamic viscosity is obtained by multiplying the increase in viscosity for each level increase in the oil pollution level by the number of pollution level increases, adding 1, and then multiplying it by the temperature compensated viscosity.

[0034] The driving leakage pressure difference per meter of seepage path length under unit viscous resistance is obtained by the ratio of the driving leakage pressure difference to the product of dynamic viscosity and seepage path length.

[0035] Preferably, the time domain fault features are subjected to Fourier transform and input into a convolutional neural network that introduces channel attention and spatial attention, so as to construct an attention convolutional neural network fault diagnosis model combined with Fourier transform, which specifically includes:

[0036] Decompose the time domain fault features into spectrum fault features through Fourier transform;

[0037] The spectral fault features are input into the first convolutional layer. The spectral fault features pass through four one-dimensional convolutional layers in sequence in the first convolutional layer. Each convolutional layer is followed by a ReLU activation function and an average pooling layer to obtain the first convolutional layer features.

[0038] Input the first convolutional layer features into the channel attention layer of the CBAM (Convolutional Block Attention Module) attention layer, multiply the features of the channel attention layer with the first convolutional layer features for weighting to obtain the channel convolution features;

[0039] Input the channel convolution features into the spatial attention layer, multiply the features passing through the spatial attention layer with the channel convolution features for weighting, and obtain the features passing through the attention layer;

[0040] The features that pass through the attention layer are input into the second convolutional layer, and the spectral fault features are extracted through a convolutional layer in the second convolutional layer;

[0041] The features after the second convolutional layer are flattened and passed to the fully connected layer for classification. The batch size, the number of iterative updates on the training dataset, the learning rate, the optimizer, and the loss function are set for training to obtain a trained attention convolutional neural network fault diagnosis model combined with Fourier transform.

[0042] Preferably, the step of inputting the first convolutional layer features into the channel attention layer of the CBAM attention layer, and multiplying the features of the channel attention layer with the first convolutional layer features for weighting to obtain the channel convolutional features specifically includes:

[0043] Input the first convolutional layer features into the channel attention layer, and perform global average pooling and global maximum pooling on the first convolutional layer features respectively;

[0044] The pooled results are sent to the fully connected layer of the channel attention layer for calculation to obtain different weights;

[0045] Add the weights of the fully connected layer outputs and then activate them through the Sigmoid function to obtain the channel attention weight;

[0046] Multiply the obtained channel attention weight with the first convolutional layer feature of the original input, weight each channel, and obtain the channel convolution feature Mc(F).

[0047] Preferably, the channel convolution features are input into the spatial attention layer, and the features passing through the spatial attention layer are multiplied with the channel convolution features for weighting to obtain the features passing through the attention layer, specifically including:

[0048] Input the channel convolution features into the spatial attention layer, perform convolution and pooling operations, and obtain the pooling result;

[0049] The pooled result is input into the spatial attention mechanism, which calculates the weights of features at different spatial positions and outputs the feature MS(F) passing through the spatial attention layer.

[0050] The second aspect of the present invention provides a fault diagnosis system for a column hydraulic system in a coal environment, which runs the fault diagnosis method for a column hydraulic system in a coal environment described in the first aspect, comprising:

[0051] The dust temperature acquisition module is used to build a dust friction correction model and a hydraulic oil temperature-viscosity characteristic model for the column hydraulic system in a coal mine operating environment, and obtain the dust concentration compensation amount and temperature-compensated viscosity;

[0052] The leakage rate solving module is used to model the porous media leakage model in the coal mine environment and solve the leakage rate based on the combination of dust concentration compensation and temperature compensation viscosity;

[0053] A fault signature building module is used to build time-domain fault signatures based on temperature-compensated viscosity, dust concentration compensation, and leakage rate;

[0054] A fault diagnosis model construction module is used to perform Fourier transform on time-domain fault features and input them into a convolutional neural network that introduces channel attention and spatial attention to construct a fault diagnosis model combined with Fourier transform and attention convolutional neural network;

[0055] The output module is used to input the actual hydraulic data into the constructed fault diagnosis model to obtain the fault diagnosis results and realize the fault diagnosis of the column hydraulic system in the coal environment.

[0056] The beneficial effects of the present invention are as follows:

[0057] By integrating frequency domain feature enhancement with channel and spatial attention mechanisms in the FFT-CBAM-CNN model, the fault diagnosis misjudgment rate was reduced under the combined working conditions of dust concentration and temperature changes, significantly improving the fault diagnosis robustness and accuracy of coal mine hydraulic systems.

[0058] The present invention combines the time domain leakage rate, temperature compensated viscosity and dust concentration compensation to construct a composite fault feature, and extracts the harmonic distortion characteristics of the frequency band through Fourier transform to improve the leakage fault recognition rate;

[0059] The present invention constructs a porous medium leakage model by coupling the dust concentration compensation amount and the permeability coefficient of the contaminated oil, quantifies the dust clogging effect in the coal mine environment, and improves the construction accuracy of the leakage rate. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 A schematic flow chart of a method for diagnosing a fault in a column hydraulic system under a coal environment provided by the present invention;

[0061] Figure 2 Schematic diagram of the FFT-CBAM-CNN model provided by the present invention;

[0062] Figure 3 A schematic diagram of the process flow of the channel attention layer provided by the present invention;

[0063] Figure 4 Schematic diagram of the process flow of the spatial attention layer provided by the present invention. DETAILED DESCRIPTION

[0064] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. The described embodiments are only part of the embodiments of the present invention, not all of them. Based on the spirit of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts are within the scope of protection of the present invention.

[0065] like Figure 1 As shown, embodiment 1 of the present invention provides a method for diagnosing faults of a column hydraulic system in a coal environment, comprising the following steps:

[0066] Step 1: Construct a dust friction correction model and a hydraulic oil temperature-viscosity characteristic model for the column hydraulic system in a coal mine operating environment to obtain the dust concentration compensation amount and temperature-compensated viscosity.

[0067] In a preferred but non-limiting embodiment of the present invention, step 1 specifically comprises:

[0068] Step 1.1: Construct a dust friction correction model for the column hydraulic system in a coal mine operating environment and solve it to obtain the dust concentration compensation.

[0069] Further preferably, step 1.1 specifically includes:

[0070] Pass the laboratory standard test of standard friction coefficient μ0 in the state without dust interference;

[0071] The dust concentration data of the working area is continuously collected through the dust sensor array in the column hydraulic system C d ;

[0072] When the dust coverage reaches the critical threshold, the shear resistance between particles and the change in surface roughness form additional friction components. Therefore, the gain coefficient is set to 0.002, which is consistent with the dust concentration data C. d Multiply and compensate them to get the dust concentration compensation amount.

[0073] Step 1.2: Construct a temperature-viscosity characteristic model of the hydraulic oil in the column hydraulic system under coal mine operation environment and solve it to obtain the temperature-compensated viscosity.

[0074] Further preferably, step 1.2 specifically includes:

[0075] Anti-wear hydraulic oil samples were tested using a laboratory rotational viscometer. Viscosity-temperature curve data was collected within the temperature range of 20-80°C. The least squares method was used to fit the equation coefficients A and B. The A value for traditional mining-grade L-HM46 hydraulic oil ranged from 3.8 to 4.2, and the B value ranged from 0.12 to 0.18.

[0076] An explosion-proof PT100 temperature sensor is placed on the outer wall of the hydraulic cylinder in the column hydraulic system. The actual oil temperature is acquired at a sampling frequency of 0.5 Hz. The actual oil temperature is pre-processed using a Kalman filter to eliminate abnormal values ​​caused by transient thermal shock of the tunneling machinery and obtain the measured temperature value T.

[0077] Calculate the logarithm of the measured temperature value T, lgT, to quantify the nonlinear effect of temperature change on viscosity;

[0078] By calculating 10(A-BlgT), and then using the power function of 10 to generate the temperature compensated viscosity v(T) = 10 10(A-BlgT) .

[0079] Step 2: Based on the dust concentration compensation amount and temperature compensation viscosity obtained in step 1, a porous medium leakage model in a coal mine environment is built to solve the leakage rate.

[0080] In a preferred but non-limiting embodiment of the present invention, step 2 specifically comprises:

[0081] Step 2.1: Correct the pores based on the dust concentration compensation obtained in step 1, and calculate the seepage area based on the geometric area of ​​the sealing surface.

[0082] Further preferably, step 2.1 specifically includes:

[0083] The pore structure image of the sealing surface was obtained by scanning electron microscopy, the pore area was extracted according to the binarization algorithm, and the initial porosity was calculated according to the ratio of pore volume to total volume.

[0084] Subtract the dust concentration compensation from 1 and multiply it by the initial porosity to generate the porosity correction value;

[0085] The seepage area is obtained by multiplying the porosity correction value by the geometric area of ​​the sealing surface in the absence of dust.

[0086] Step 2.2: Calculate the permeability coefficient of the contaminated oil based on the contamination level and the decrease in permeability coefficient.

[0087] Further preferably, step 2.2 specifically includes:

[0088] The online particle counter monitors the number of particles in the oil in real time and classifies the oil contamination level according to ISO 4406 standard;

[0089] For each level increase in oil contamination, the permeability coefficient decrease is used as the base, the level of oil contamination increase is used as the index, and the permeability coefficient of clean oil × the permeability coefficient decrease is used as the index. 污染度等级提升的级别数 , solve the permeability coefficient of the contaminated oil.

[0090] In step 2.3, the temperature correction factor is obtained by multiplying the ratio of the standard viscosity to the temperature-compensated viscosity obtained in step 2.2 by the collected oil temperature.

[0091] In step 2.4, calculate the driving leakage pressure difference per meter of seepage path length under unit viscous resistance.

[0092] Further preferably, step 2.4 includes:

[0093] In step 2.4.1, the driving leakage pressure difference is obtained based on the basic pressure difference and the hysteresis loop energy loss caused by dust accumulation.

[0094] More preferably, step 2.4.1 specifically includes:

[0095] Solve the difference between the hydraulic system working pressure and the external pressure to obtain the basic pressure difference;

[0096] Obtain the additional resistance P of dust accumulation, integrate the seepage volume according to the additional resistance P, and calculate the hysteresis loop energy loss formed by dust accumulation;

[0097] The driving leakage pressure difference is obtained by summing the basic pressure difference and the hysteresis loop energy loss formed by dust accumulation.

[0098] Step 2.4.2: Multiply the viscosity increase for each level increase in the oil's contamination level by the number of levels of contamination level increase, add 1, and then multiply by the temperature-compensated viscosity to obtain the dynamic viscosity.

[0099] In step 2.4.3, the driving leakage pressure difference per meter of seepage path length under unit viscous resistance is obtained by taking the ratio of the driving leakage pressure difference to the product of the dynamic viscosity and the seepage path length.

[0100] In step 2.5, multiply the seepage area from step 2.1, the contaminated oil permeability coefficient from step 2.2, the temperature correction factor from step 2.3, and the driving leakage pressure difference per meter of seepage path length under unit viscous resistance to obtain the leakage rate, which represents the leakage intensity caused by dust contamination on the sealing surface of the hydraulic system.

[0101] Step 3: Construct a time domain fault feature based on the temperature compensated viscosity and dust concentration compensation in step 1 and the leakage rate solved in step 2.

[0102] Step 4, such as Figure 2As shown in the figure, the time domain fault features are Fourier transformed and input into the convolutional neural network that introduces channel attention and spatial attention to construct a fault diagnosis model combined with Fourier transform attention convolutional neural network (FFT-CBAM-CNN).

[0103] In a preferred but non-limiting embodiment of the present invention, step 4 specifically comprises:

[0104] Step 4.1, decompose the time domain fault features into spectrum fault features through Fourier transform;

[0105] In step 4.2, the spectral fault features are input into the first convolutional layer. The spectral fault features are sequentially passed through four one-dimensional convolutional layers in the first convolutional layer. Each convolutional layer is followed by a ReLU activation function and an average pooling layer to obtain the first convolutional layer features, which are used to extract features at different levels.

[0106] Step 4.3: Pass the features that pass through the convolutional layer through the channel attention and spatial attention layers in the CBAM attention layer in sequence to obtain the features that pass through the attention layer, which are used to enhance the feature representation through the channel and spatial attention mechanisms.

[0107] Further preferably, step 4.3 includes:

[0108] Step 4.3.1, as Figure 3 As shown, the first convolutional layer features of step 4.2 are input into the channel attention layer, and the features passing through the channel attention layer are multiplied with the first convolutional layer features for weighting to obtain the channel convolution features;

[0109] More preferably, step 4.3.1 comprises:

[0110] Input the first convolutional layer features of step 4.2 into the channel attention layer, and perform global average pooling and global maximum pooling on the first convolutional layer features respectively;

[0111] The pooled results are sent to the fully connected layer of the channel attention layer for calculation to obtain different weights;

[0112] Add the weights of the fully connected layer outputs and then activate them through the Sigmoid function to obtain the channel attention weight;

[0113] Multiply the obtained channel attention weight with the first convolution layer feature of the original input, weight each channel, highlight the important channels, and suppress the unimportant channels, thereby completing the processing of the feature map by the channel attention mechanism and obtaining the channel convolution feature M c (F), expressed as follows:

[0114]

[0115] Where:

[0116] F is the input feature map;

[0117] σ is the Sigmoid function;

[0118] MLP is the fully connected layer calculation;

[0119] AvgPool(*) is the global average pooling calculation;

[0120] MaxPool(*) is the global maximum pooling calculation;

[0121] W1 and W0 are the output weights of the fully connected layer;

[0122] is the result of pooling calculation.

[0123] Step 4.3.2, as Figure 4 As shown in Figure 1, the channel convolution features are input into the spatial attention layer, and the features passing through the spatial attention layer are multiplied with the channel convolution features for weighting to obtain the features passing through the attention layer.

[0124] More preferably, step 4.3.2 comprises:

[0125] Input the channel convolution features into the spatial attention layer, perform convolution and pooling operations, and obtain the pooling result;

[0126] The pooled result is input into the spatial attention mechanism, which calculates the weights of the features at different spatial positions, highlights or suppresses the feature information of certain spatial positions, and finally outputs the feature M that passes the spatial attention layer. S (F), expressed as follows:

[0127]

[0128] Where:

[0129] f β () is the convolution operation with a convolution kernel size of β,

[0130] is the result of pooling calculation.

[0131] In step 4.4, the features that pass through the attention layer are input into the second convolutional layer, and the spectral fault features are extracted through a convolutional layer in the second convolutional layer;

[0132] In step 4.5, the features after the second convolutional layer are flattened and passed to the fully connected layer for classification. The batch size is set to 16 during training, and the training dataset is iterated 50 times with a learning rate of 1×10 -5, the optimizer is Adam, the loss function is cross entropy loss, and training is performed to obtain a trained attention convolutional neural network fault diagnosis model combined with Fourier transform, which finally outputs 7 categories of fault types.

[0133] Step 5: Input the actual hydraulic data into the fault diagnosis model constructed in step 4 to implement fault diagnosis of the hydraulic system.

[0134] A second embodiment of the present invention provides a fault diagnosis system for a hydraulic system of a column in a coal environment, which implements a fault diagnosis method for a hydraulic system of a column in a coal environment, and a one-way valve test device with a fault diagnosis function composed of various functional modules, specifically including:

[0135] The dust temperature acquisition module is used to build a dust friction correction model and a hydraulic oil temperature-viscosity characteristic model for the column hydraulic system in a coal mine operating environment, and obtain the dust concentration compensation amount and temperature-compensated viscosity;

[0136] The leakage rate solving module is used to model the porous media leakage model in the coal mine environment and solve the leakage rate based on the combination of dust concentration compensation and temperature compensation viscosity;

[0137] A fault signature building module is used to build time-domain fault signatures based on temperature-compensated viscosity, dust concentration compensation, and leakage rate;

[0138] A fault diagnosis model construction module is used to perform Fourier transform on time-domain fault features and input them into a convolutional neural network that introduces channel attention and spatial attention to construct a fault diagnosis model combined with Fourier transform and attention convolutional neural network;

[0139] The output module is used to input the actual hydraulic data into the constructed fault diagnosis model to obtain the fault diagnosis results and realize the fault diagnosis of the column hydraulic system in the coal environment.

[0140] The beneficial effects of the present invention are as follows:

[0141] By integrating frequency domain feature enhancement with channel and spatial attention mechanisms in the FFT-CBAM-CNN model, the fault diagnosis misjudgment rate was reduced under the combined working conditions of dust concentration and temperature changes, significantly improving the fault diagnosis robustness and accuracy of coal mine hydraulic systems.

[0142] The present invention combines the time domain leakage rate, temperature compensated viscosity and dust concentration compensation to construct a composite fault feature, and extracts the harmonic distortion characteristics of the frequency band through Fourier transform to improve the leakage fault recognition rate;

[0143] The present invention constructs a porous medium leakage model by coupling the dust concentration compensation amount and the permeability coefficient of the contaminated oil, quantifies the dust clogging effect in the coal mine environment, and improves the construction accuracy of the leakage rate.

[0144] The present disclosure may be a system, method and / or computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for causing a processor to implement various aspects of the present disclosure.

[0145] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered by the scope of protection of the claims of the present invention.

Claims

1. A method for diagnosing faults in a column hydraulic system under a coal environment, characterized by: A dust friction correction model and a hydraulic oil temperature-viscosity characteristic model for the column hydraulic system in a coal mine operating environment were constructed to obtain the dust concentration compensation and temperature-compensated viscosity. Based on the combination of dust concentration compensation and temperature compensation viscosity, the porous media leakage model in the coal mine environment is modeled and the leakage rate is solved; Construct time domain fault signatures based on temperature compensated viscosity, dust concentration compensation, and leakage rate; The time-domain fault features are Fourier transformed and input into a convolutional neural network that introduces channel attention and spatial attention to construct a fault diagnosis model combining Fourier transform and attention convolutional neural network. The actual hydraulic data is input into the constructed fault diagnosis model to obtain the fault diagnosis results, thus realizing the fault diagnosis of the column hydraulic system in the coal environment.

2. The method for diagnosing faults of a column hydraulic system in a coal environment according to claim 1, characterized in that: The construction of the dust friction correction model for the column hydraulic system in the coal mine working environment specifically includes: Test the standard friction coefficient μ0 without dust interference; The dust concentration data of the working area is continuously collected through the dust sensor array in the column hydraulic system C d ; Set the gain coefficient and compare it with the dust concentration data C d Multiply and compensate them to get the dust concentration compensation amount.

3. The method for diagnosing faults of a column hydraulic system in a coal environment according to claim 1, characterized in that: The construction of the hydraulic oil temperature-viscosity characteristic model of the column hydraulic system in the coal mine operation environment specifically includes: Test anti-wear hydraulic oil samples, collect viscosity-temperature curve data within a set temperature range, and use the least squares method to fit and determine the equation coefficients A and B; An explosion-proof PT100 temperature sensor is placed on the outer wall of the hydraulic cylinder in the column hydraulic system. The actual oil temperature is acquired at a sampling frequency of 0.5 Hz. The actual oil temperature is pre-processed using a Kalman filter to eliminate abnormal values ​​caused by transient thermal shock of the tunneling machinery and obtain the measured temperature value T. Calculate the logarithm of the measured temperature value T, lgT; Finally, the temperature-compensated viscosity v(T)=10 is generated by the power function of 10 10(A-BlgT) .

4. The method for diagnosing faults of a column hydraulic system in a coal environment according to claim 1, characterized in that: The method of modeling the porous medium leakage model in the coal mine environment based on the combination of dust concentration compensation and temperature compensation viscosity to solve the leakage rate specifically includes: The pores are corrected according to the dust concentration compensation, and the seepage area is solved by combining the geometric area of ​​the sealing surface; According to the contamination level and the decrease in permeability coefficient, the permeability coefficient of the contaminated oil is solved; The temperature correction factor is obtained by multiplying the ratio of the standard viscosity to the temperature-compensated viscosity by the collected oil temperature. Calculate the driving leakage pressure difference per meter of seepage path length under unit viscous resistance; The leakage rate is obtained by multiplying the seepage area, the permeability coefficient of the contaminated oil, the temperature correction factor, and the driving leakage pressure difference per meter of seepage path length under unit viscous resistance.

5. The method for diagnosing faults of a column hydraulic system in a coal environment according to claim 4, characterized in that: The porosity is corrected according to the dust concentration compensation amount, and the seepage area is solved in combination with the geometric area of ​​the sealing surface, specifically including: Obtain the pore structure image of the sealing surface, extract the pore area according to the binarization algorithm, and calculate the initial porosity according to the ratio of pore volume to total volume; Subtract the dust concentration compensation from 1 and multiply it by the initial porosity to generate the porosity correction value; The seepage area is obtained by multiplying the porosity correction value by the geometric area of ​​the sealing surface in the absence of dust.

6. The method for diagnosing faults of a column hydraulic system in a coal environment according to claim 4, characterized in that: The method of solving the driving leakage pressure difference per meter of seepage path length under unit viscous resistance specifically includes: The driving leakage pressure difference is obtained by solving the energy loss of the hysteresis loop formed by the basic pressure difference and dust accumulation; The dynamic viscosity is obtained by multiplying the increase in viscosity for each level increase in the oil pollution level by the number of pollution level increases, adding 1, and then multiplying it by the temperature compensated viscosity. The driving leakage pressure difference per meter of seepage path length under unit viscous resistance is obtained by the ratio of the driving leakage pressure difference to the product of dynamic viscosity and seepage path length.

7. The method for diagnosing faults of a column hydraulic system in a coal environment according to claim 1, characterized in that: The time domain fault features are Fourier transformed and input into a convolutional neural network that introduces channel attention and spatial attention to construct a fault diagnosis model of an attention convolutional neural network combined with Fourier transform, which specifically includes: Decompose the time domain fault features into spectrum fault features through Fourier transform; The spectral fault features are input into the first convolutional layer. The spectral fault features pass through four one-dimensional convolutional layers in sequence in the first convolutional layer. Each convolutional layer is followed by a ReLU activation function and an average pooling layer to obtain the first convolutional layer features. Input the first convolutional layer features into the channel attention layer of the CBAM attention layer, multiply the features of the channel attention layer with the first convolutional layer features for weighting to obtain the channel convolution features; Input the channel convolution features into the spatial attention layer, multiply the features passing through the spatial attention layer with the channel convolution features for weighting, and obtain the features passing through the attention layer; The features that pass through the attention layer are input into the second convolutional layer, and the spectral fault features are extracted through a convolutional layer in the second convolutional layer; The features after the second convolutional layer are flattened and passed to the fully connected layer for classification. The batch size, the number of iterative updates on the training dataset, the learning rate, the optimizer, and the loss function are set for training to obtain a trained attention convolutional neural network fault diagnosis model combined with Fourier transform.

8. The method for diagnosing faults of a column hydraulic system in a coal environment according to claim 7, characterized in that: The first convolutional layer features are input into the channel attention layer of the CBAM attention layer, and the features of the channel attention layer are multiplied by the first convolutional layer features for weighting to obtain the channel convolution features, specifically including: Input the first convolutional layer features into the channel attention layer, and perform global average pooling and global maximum pooling on the first convolutional layer features respectively; The pooled results are sent to the fully connected layer of the channel attention layer for calculation to obtain different weights; Add the weights of the fully connected layer outputs and then activate them through the Sigmoid function to obtain the channel attention weight; Multiply the obtained channel attention weight with the first convolutional layer feature of the original input, weight each channel, and obtain the channel convolution feature M c (F).

9. The method for diagnosing faults of a column hydraulic system in a coal environment according to claim 7, characterized in that: The channel convolution features are input into the spatial attention layer, and the features passing through the spatial attention layer are multiplied with the channel convolution features for weighting to obtain the features passing through the attention layer, specifically including: Input the channel convolution features into the spatial attention layer, perform convolution and pooling operations, and obtain the pooling result; The pooled result is input into the spatial attention mechanism, which calculates the weights of the features at different spatial positions and outputs the feature M that passes through the spatial attention layer. S (F).

10. A fault diagnosis system for a hydraulic system of a column in a coal environment, which runs a fault diagnosis method for a hydraulic system of a column in a coal environment according to any one of claims 1 to 9, characterized in that: The dust temperature acquisition module is used to build a dust friction correction model and a hydraulic oil temperature-viscosity characteristic model for the column hydraulic system in a coal mine operating environment, and obtain the dust concentration compensation amount and temperature-compensated viscosity; The leakage rate solving module is used to model the porous media leakage model in the coal mine environment and solve the leakage rate based on the combination of dust concentration compensation and temperature compensation viscosity; A fault signature building module is used to build time-domain fault signatures based on temperature-compensated viscosity, dust concentration compensation, and leakage rate; A fault diagnosis model construction module is used to perform Fourier transform on time-domain fault features and input them into a convolutional neural network that introduces channel attention and spatial attention to construct a fault diagnosis model combined with Fourier transform and attention convolutional neural network; The output module is used to input the actual hydraulic data into the constructed fault diagnosis model to obtain the fault diagnosis results and realize the fault diagnosis of the column hydraulic system in the coal environment.

Images