Explosion-proof intelligent electro-hydraulic actuator for coal mine and fault self-diagnosis method thereof

Abstract

The invention relates to the technical field of explosion-proof intelligent equipment for a coal mine, in particular to an explosion-proof intelligent electro-hydraulic actuator for the coal mine and a fault self-diagnosis method of the explosion-proof intelligent electro-hydraulic actuator. The anti-explosion intelligent electro-hydraulic actuator for the coal mine comprises a mining anti-explosion speed regulating motor, a closed hydraulic system without an external pipeline and an intelligent anti-explosion control module, and the closed hydraulic system is composed of a bidirectional hydraulic pump driven by a motor, a hydraulic cylinder, an integrated valve set and an energy accumulator. The problems that according to a traditional coal mine underground hydraulic power station scheme, pipelines are prone to leakage, the occupied area is large, and the open system efficiency is low are solved. A fault self-diagnosis program operated by the intelligent explosion-proof control module firstly acquires multi-dimensional operation parameters such as pressure difference, flow, displacement, speed and rotating speed of an oil pump, converts one-dimensional time sequence data into a two-dimensional feature image by utilizing a glami angle and field method, and then inputs the two-dimensional feature image into a pre-trained fault diagnosis model to output a diagnosis result; and the fault identification accuracy and the convergence speed are effectively improved.

Description

Technical Field

[0001] This invention relates to the field of explosion-proof intelligent equipment technology for coal mines, and in particular to an explosion-proof intelligent electro-hydraulic actuator for coal mines and its fault self-diagnosis method. Background Technology

[0002] With the deepening of intelligent construction in coal mines, the demand for automated operating machinery in underground coal mines is becoming increasingly urgent. However, due to the strict explosion-proof standards required by the extreme underground environment, many mature automated core components on the surface cannot be directly applied, especially actuators that can meet the requirements of heavy load and high-precision linear drive, forming a bottleneck of missing core components.

[0003] Currently, there are two main technical solutions for driving heavy linear loads in underground coal mines, but both have significant shortcomings: The first type is the traditional hydraulic cylinder drive scheme. This scheme typically consists of an independent hydraulic power station, external high-pressure pipelines, and hydraulic cylinders. However, it has the following drawbacks: the system is bulky and has low integration; the hydraulic power station occupies a large area and requires long pipelines to connect it to the hydraulic cylinders, making the entire system less compact and difficult to adapt to the confined space of underground coal mines; the system mostly uses valve-controlled throttling speed regulation, resulting in significant throttling and overflow losses, leading to low overall transmission efficiency; it relies on the adjustment of hydraulic valves, and its response speed and control accuracy are insufficient to meet the requirements of high-precision operations; the numerous external pipelines and joints are high-risk areas for hydraulic oil leakage, which not only pollutes the environment but also easily causes equipment failures and requires a large amount of maintenance.

[0004] The second option is the heavy-duty explosion-proof pure electric actuator solution. To solve the problems of the hydraulic system, pure electric actuators were first adopted. However, pure electric actuators also have certain drawbacks: in order to directly drive heavy loads, the required motor power is huge. When high-power motors meet the explosion-proof standards of coal mines, they face the dilemma of a sharp increase in size, weight and cost, which makes it difficult to miniaturize and lighten the product, thus limiting its application scenarios.

[0005] In summary, existing technologies cannot adequately address the urgent need for heavy-duty, high-precision, high-reliability, and compact linear drive actuators in underground coal mines. As a device that combines the advantages of high hydraulic output and high electrical control precision, the application of electro-hydraulic actuators in underground coal mines requires a systematic solution to a series of key technical issues, including explosion-proof design, system integration, transmission efficiency, and intelligent operation and maintenance. Summary of the Invention

[0006] The technical problem this invention aims to solve is as follows: Existing mining hydraulic actuators suffer from problems such as numerous pipelines leading to leaks, large footprint, low efficiency, and insufficient control precision. High-power pure electric actuators, on the other hand, face difficulties in explosion-proof design and miniaturization. This invention provides an explosion-proof intelligent electro-hydraulic actuator for coal mines and its fault self-diagnosis method. By highly integrating a mining explosion-proof speed-regulating motor, a bidirectional hydraulic pump, a hydraulic cylinder, and a valve group accumulator, a closed-loop hydraulic system without external pipelines is formed, solving the problems of leakage, low efficiency, and large size associated with traditional solutions. Multi-dimensional operating parameters are collected and converted into two-dimensional images using the Gramian angle and field method, providing optimized data for subsequent intelligent diagnosis. A two-dimensional convolutional neural network fault diagnosis model based on particle swarm optimization is constructed to intelligently analyze the two-dimensional feature images, ultimately achieving the actuator's fault self-diagnosis and early warning functions.

[0007] The technical solution adopted by this invention to solve its technical problem is: an explosion-proof intelligent electro-hydraulic actuator for coal mines, comprising: Explosion-proof speed-regulating motor for mining; A closed-loop hydraulic system without external piping consists of a bidirectional hydraulic pump driven by the explosion-proof speed-regulating motor, a hydraulic cylinder, and an integrated hydraulic valve group and accumulator. The intelligent explosion-proof control module performs motion control on the electro-hydraulic actuator and runs a fault self-diagnosis program. The fault self-diagnosis program includes: Multiple operating parameters of the electro-hydraulic actuator are collected, including at least the differential pressure between the inlet and outlet of the oil pump, the oil pump flow rate, the displacement response, the speed response, and the motor speed. The collected one-dimensional time series running parameters are converted into two-dimensional feature images using the Gramian angle and field method; The two-dimensional feature image is input into a pre-trained fault diagnosis model, and the fault diagnosis result is output. The fault diagnosis model is a two-dimensional convolutional neural network model with parameters optimized based on the particle swarm optimization algorithm.

[0008] By integrating the explosion-proof speed-regulating motor, bidirectional hydraulic pump, hydraulic cylinder, hydraulic valve group, and accumulator into one unit, a closed hydraulic system without external pipelines is formed. This not only solves the problem of leakage caused by the large number of external pipelines in traditional hydraulic actuators, but also improves the system's compactness and reduces the floor space. The closed hydraulic system can improve transmission efficiency and reduce energy loss.

[0009] Furthermore, the explosion-proof speed-regulating motor for mining is an intrinsically safe explosion-proof speed-regulating motor for mining, a potted explosion-proof speed-regulating motor for mining, or a flameproof and potted explosion-proof speed-regulating motor for mining.

[0010] The mining explosion-proof speed regulating motor is limited to intrinsically safe or potted explosion-proof speed regulating motors or explosion-proof and potted explosion-proof speed regulating motors. This ensures that the strict explosion-proof requirements in underground mines are met, while effectively overcoming the disadvantages of traditional explosion-proof motors being large in size and heavy in weight. This achieves the miniaturization and weight reduction of the equipment, making it more suitable for installation and use in the space-constrained environment of underground coal mines.

[0011] Furthermore, the hydraulic valve group includes a replenishing shuttle valve, a hydraulically controlled check valve, and a safety relief valve. The hydraulically controlled check valve is connected between the bidirectional hydraulic pump and the hydraulic cylinder and is used to lock the position of the hydraulic cylinder when the power is off. The safety relief valve is connected between the hydraulic cylinder and the accumulator and is used to release pressure when the load exceeds the limit; the oil replenishment shuttle valve is used to compensate for the flow difference between the large and small chambers of the hydraulic cylinder.

[0012] By configuring a hydraulic valve group that includes a replenishing shuttle valve, a hydraulically controlled check valve, and a safety relief valve, multiple functions are ensured: the hydraulically controlled check valve ensures reliable locking of the hydraulic cylinder position in the event of a power failure, improving equipment safety; the safety relief valve releases pressure in a timely manner when the load exceeds the limit, playing an overload protection role; and the replenishing shuttle valve can effectively compensate for the flow difference between the large and small chambers of the hydraulic cylinder, ensuring the stability and reliability of the system operation.

[0013] Furthermore, damping elements are provided at the large and small chamber oil ports of the hydraulic cylinder to suppress pressure fluctuations generated when the cylinder stops due to inertia under heavy load conditions.

[0014] By installing damping elements at the large and small chamber oil ports of the hydraulic cylinder, the pressure shock and mechanical vibration caused by inertial stopping under heavy load conditions can be effectively suppressed. This not only improves the stability of the equipment during heavy load operation, but also extends the service life of the hydraulic components.

[0015] Furthermore, the collected one-dimensional time series running parameters The feature image is converted to an n×n two-dimensional feature image using the Gramian angle and field method. The specific steps are as follows: First, the time series at time t Perform normalization to obtain the normalized value. The calculation formula is: , Then, the normalized values Mapped to angle The calculation formula is: , , , Finally, based on two angles , Cosine formula: , Construct the Gramian angle and field matrix G, whose elements The calculation formula is: , We obtain an n×n symmetric two-dimensional matrix, which serves as the two-dimensional feature image.

[0016] This paper presents specific calculation methods and steps for converting one-dimensional time series operating parameters into two-dimensional feature images, establishes a complete data preprocessing scheme, transforms time series data into a format suitable for image recognition and processing, and provides a good data foundation for subsequent intelligent diagnosis based on deep learning.

[0017] Furthermore, the establishment of the fault diagnosis model includes the following steps: Construct an initial model for a two-dimensional convolutional neural network; The particle swarm optimization algorithm is used to globally optimize the hyperparameters of the initial model of the two-dimensional convolutional neural network, with the fault classification accuracy as the fitness function of the particles. A fault diagnosis model is obtained by configuring and training a two-dimensional convolutional neural network model with optimized hyperparameters.

[0018] By establishing a fault diagnosis model training process based on particle swarm optimization, combining intelligent optimization algorithms with deep learning, and through global optimization of model hyperparameters, the accuracy and generalization ability of the fault diagnosis model are significantly improved, making the diagnosis results more reliable.

[0019] Furthermore, the two-dimensional convolutional neural network model includes: The input layer is used to receive an n×n two-dimensional feature image; The feature extraction module consists of at least one set of convolutional layers and pooling layers connected in sequence; the convolutional layers are used to extract feature maps from the input data through linear operations of convolutional kernels and biases and nonlinear activation functions; the pooling layers are used to downsample the feature maps to reduce the data dimensionality. The classification module includes a fully connected layer and an output layer; the fully connected layer is used to flatten the downsampled feature map into a feature vector, and calculate and output high-level features through weights, biases and activation functions; the output layer is used to output the fault classification result based on the high-level features.

[0020] By optimizing the structure of the two-dimensional convolutional neural network model and adopting an architecture that includes convolutional layers, pooling layers, and fully connected layers, the model can efficiently process input feature images while ensuring the accuracy of feature extraction and fault classification, thus making the diagnostic system more practical.

[0021] Furthermore, the specific steps for globally optimizing the hyperparameters of the initial model of the two-dimensional convolutional neural network using the particle swarm optimization algorithm are as follows: a. Initialize the particle swarm optimization (PSO) algorithm parameters, including particle velocity, position, inertia weight, learning factor, population size, and maximum number of iterations; b. Initialize the particle swarm, where the position of each particle is randomly generated, representing a potential solution for the weights and biases of the two-dimensional convolutional neural network model; c. For each particle, decode and set the weights and biases of the two-dimensional convolutional neural network model based on its position information, and use the training dataset to calculate the fault classification accuracy of the model as the fitness value of the particle. d. Update the individual optimal position and the global optimal position of each particle; e. Update the velocity and position of all particles in the particle swarm according to the velocity and position update formulas of the particle swarm algorithm; f. Iterate through steps c to e until the termination condition is met, and output the globally optimal particle position, which represents the optimized combination of weights and bias parameters.

[0022] Specifically, the implementation steps for optimizing neural network parameters using the particle swarm optimization algorithm are defined, and a complete parameter optimization process is established. Through iterative optimization, the model training is effectively prevented from getting stuck in local optima, thereby further improving the accuracy of fault diagnosis and the convergence speed of the model.

[0023] Furthermore, the intelligent explosion-proof control module is installed inside the mine explosion-proof control box.

[0024] By installing the intelligent explosion-proof control module inside the mine explosion-proof control box, reliable explosion-proof protection is provided for the electrical control system, ensuring the safe operation of equipment in the flammable and explosive environment of underground coal mines, while also facilitating centralized management and maintenance of the system.

[0025] A fault self-diagnosis method for the explosion-proof intelligent electro-hydraulic actuator for coal mines described in the above scheme is also provided, comprising the following steps: S1. Real-time acquisition of multiple operating parameters of the electro-hydraulic actuator to form one-dimensional time series data; S2. Encode one-dimensional time series data into two-dimensional feature images using the Gramian angle and field method; S3. Input the two-dimensional feature image into the pre-trained fault diagnosis model. The fault diagnosis model is a two-dimensional convolutional neural network model with optimized parameters based on the particle swarm optimization algorithm. S4. Obtain and output the diagnostic results of the fault diagnosis model.

[0026] The entire automated process, from data acquisition and feature conversion to intelligent diagnosis, enables equipment to have intelligent functions such as self-sensing of status and self-diagnosis of faults, improving the maintainability and operational safety of the equipment and reducing the workload of manual inspection.

[0027] The beneficial effects of this invention are: This invention highly integrates the explosion-proof speed-regulating motor with various components of the hydraulic system, forming a closed hydraulic system without external pipelines, thus solving the leakage problem caused by the numerous external pipelines in traditional hydraulic systems. This integrated design not only improves the system's compactness and adapts to the needs of use in confined underground spaces, but also significantly improves transmission efficiency and reduces energy loss through the advantages of a closed hydraulic system. At the same time, the use of intrinsically safe or potted explosion-proof motors ensures safety while achieving miniaturization and weight reduction of the equipment, overcoming the disadvantages of large size and weight of traditional explosion-proof motors. This invention effectively improves the reliability and stability of the equipment under various working conditions by configuring a hydraulic valve group with multiple protection functions and a damping element; the hydraulic control check valve ensures position locking in the event of a power failure, the safety relief valve provides overload protection, the oil replenishment shuttle valve ensures flow balance, and the damping element can suppress pressure shocks under heavy load conditions, thus improving safety. This invention combines time-series data conversion with deep learning technology. It converts one-dimensional operating parameters into two-dimensional feature images using the Gramian angle and field method, and then uses a convolutional neural network optimized by the particle swarm optimization algorithm for fault diagnosis. This enables the equipment to have self-sensing and fault self-diagnosis capabilities, greatly improving the maintainability and operational safety of the equipment. Attached Figure Description

[0028] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0029] Figure 1 This is a structural diagram of an existing technical solution; Figure 1 (a) is a traditional hydraulic cylinder drive scheme; Figure 1 (b) is a heavy-duty explosion-proof pure electric drive actuator solution.

[0030] Figure 2 This is a schematic diagram of the structure of the explosion-proof intelligent electro-hydraulic actuator for coal mines according to the present invention.

[0031] Figure 3 This is a control principle diagram of the explosion-proof intelligent electro-hydraulic actuator for coal mines according to the present invention.

[0032] Figure 4 This is a flowchart of the fault self-diagnosis method for the explosion-proof intelligent electro-hydraulic actuator for coal mines according to the present invention.

[0033] In the diagram: 1. Explosion-proof speed-regulating motor; 2. Two-way hydraulic pump; 3. Hydraulic cylinder; 4. Oil replenishment shuttle valve; 5. Hydraulic control check valve; 6. Safety relief valve; 7. Accumulator; 8. Damping element; 9. Mine explosion-proof control box. Detailed Implementation

[0034] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0035] Example 1 like Figure 2 As shown, an explosion-proof intelligent electro-hydraulic actuator for coal mines includes an explosion-proof speed-regulating motor 1, a bidirectional hydraulic pump 2, a hydraulic cylinder 3, a hydraulic valve group, and an accumulator 7. The explosion-proof speed-regulating motor 1 can be intrinsically safe, potted, or explosion-proof. The bidirectional hydraulic pump 2 can be a piston pump or a gear pump. The accumulator 7 can be an airbag type or a micro-positive pressure oil tank. The hydraulic cylinder 3 includes a cylinder barrel 31 and a piston 32. A damping element 8 is provided at the oil ports of the large and small chambers of the hydraulic cylinder 3 to prevent pressure fluctuations in the large and small chambers of the hydraulic cylinder when the cylinder stops due to heavy load inertia, thus maintaining overall stability and avoiding shaking.

[0036] like Figure 3 As shown, the hydraulic valve group includes a replenishing shuttle valve 4, a hydraulically controlled check valve 5, and a safety relief valve 6. The hydraulically controlled check valve 5 is connected to the left and right oil ports of the bidirectional hydraulic pump 2 and the two chambers of the hydraulic cylinder 3, enabling the hydraulic cylinder 3 to be locked in position when the explosion-proof speed-regulating motor 1 is not working during power failure. The safety relief valve 6 is installed between the two chambers of the hydraulic cylinder 3 and the accumulator 7. When the hydraulic cylinder is overloaded and the internal pressure of the hydraulic cylinder exceeds the limit, oil overflows into the accumulator 7 through the relief valve 6, providing safety protection. The replenishing shuttle valve 4 is used to connect to the accumulator 7, the two outlets of the bidirectional hydraulic pump 2, and the two chambers of the hydraulic cylinder 3. Since the hydraulic cylinder 3 is a single-acting cylinder, the areas of the two chambers are not equal, resulting in inconsistent flow rates from the two chambers of the hydraulic cylinder 3. To compensate for this inconsistency, the replenishing shuttle valve 4 is installed to compensate for the inconsistent oil flow rates in the two chambers, ensuring that the flow rates through the inlet and outlet of the hydraulic pump 2 are equal.

[0037] The electro-hydraulic actuator in this embodiment also includes an intelligent explosion-proof control module. In addition to controlling the motion of the electro-hydraulic actuator, it also runs a fault self-diagnosis program. The intelligent explosion-proof control module controls the input signal of the explosion-proof speed-regulating motor 1 and outputs the linear motion of the piston rod 32 of the hydraulic cylinder 3 to drive the load. The fault self-diagnosis program includes: collecting multiple operating parameters of the electro-hydraulic actuator, converting the collected one-dimensional time series operating parameters into a two-dimensional feature image using the Gramian angle and field method; and then inputting the two-dimensional feature image into a pre-trained fault diagnosis model to output the fault diagnosis result.

[0038] The intelligent explosion-proof control module is installed in the mine explosion-proof control box 9 and is explosion-proof. The mine explosion-proof control box 9 contains electrical components such as control circuit boards and transformers, and is electrically connected to the oil replenishment shuttle valve 4, the hydraulic control check valve 5 and the safety relief valve 6.

[0039] This embodiment uses an intrinsically safe, potted, or explosion-proof and potted explosion-proof speed-regulating DC motor 1 for mining. Compared to traditional motors, it is smaller in size and weight. The explosion-proof speed-regulating DC motor 1 integrates a drive circuit board and a control circuit board, and has built-in speed regulation and feedback closed-loop functions. Used in mining explosion-proof electro-hydraulic actuators, it avoids the disadvantages of traditional explosion-proof AC motors used in coal mines, such as large size and weight and lack of closed-loop feedback. Its use in mining explosion-proof electro-hydraulic actuators offers advantages such as high overall integration, high control precision, small size and weight, and a high force-to-weight ratio. The way to achieve equipment miniaturization and lightweighting is through: ①Use intrinsically safe, potted motors, explosion-proof and potted explosion-proof speed-regulating motors, and explosion-proof motors; ② Due to the system principle, the existing solution adopts an open system scheme with a hydraulic power station, such as... Figure 1 As shown in (a), it has drawbacks such as large footprint and large volume; while the existing solution integrates the explosion-proof motor, valve group, controller, etc. into one unit, such as... Figure 1 As shown in (b), it has a small size and a small footprint, and does not require space-consuming equipment such as a hydraulic power station.

[0040] like Figure 1 As shown, taking the coal bunker dredging device in an underground coal mine as an example, the current solution uses a hydraulic power station connected to a tamper cylinder via an oil pipe to perform the dredging task. Replacing the hydraulic power station + tamper cylinder with a mine explosion-proof electro-hydraulic actuator can reduce the footprint of the hydraulic power station, which is conducive to the small and compact design of the equipment and saves underground space in the coal mine.

[0041] Mining explosion-proof electro-hydraulic actuators are complex electromechanical devices. Based on their operating principles, their fault types can be broadly categorized into electrical faults and hydraulic faults. Electrical faults include motor faults, circuit board faults, and sensor faults, while hydraulic faults include pump faults, internal leakage in hydraulic cylinders, external leakage in hydraulic components, and hydraulic valve group faults. Therefore, five key operating parameters of the mining explosion-proof electro-hydraulic actuator can be selected as monitoring targets: the differential pressure between the inlet and outlet of the oil pump, the oil pump flow rate, the displacement response and speed response of the actuator, and the motor speed. By learning from these parameters, faults in the mining explosion-proof electro-hydraulic actuator can be diagnosed.

[0042] Convolutional Neural Networks (CNNs) are a type of neural network in deep learning specifically designed to process data with a grid structure. They are widely known for their tremendous success in the field of computer vision and are one of the core components of modern artificial intelligence systems.

[0043] Therefore, the data of the five monitoring parameters of the mine explosion-proof electro-hydraulic actuator changing over time can be used as a time series and can be input to a one-dimensional convolutional neural network. However, the fault diagnosis rate of a one-dimensional convolutional neural network is significantly lower than that of a two-dimensional convolutional neural network, and convolutional neural networks have significant advantages in two-dimensional image processing. Therefore, a two-dimensional convolutional neural network can be chosen to establish a fault diagnosis model for the mine explosion-proof electro-hydraulic actuator. A whole-machine fault simulation model is established using simulation software to simulate operating parameters. Five typical data points are selected: oil pump inlet and outlet pressure difference, oil pump flow rate, displacement response, speed response, and motor speed of the mine explosion-proof electro-hydraulic actuator. A convolutional neural network model is then established based on the convolutional neural network (CNN). Using the fault dataset established through simulation, a convolutional neural network model is built to perform fault diagnosis of the mine explosion-proof electro-hydraulic actuator, and the final diagnosis result is output. This achieves greater intelligence for the mine explosion-proof electro-hydraulic actuator.

[0044] Example 2 like Figure 4 As shown, the fault self-diagnosis method for the explosion-proof intelligent electro-hydraulic actuator used in Embodiment 1 includes the following steps: S1. Real-time acquisition of multiple operating parameters of the electro-hydraulic actuator to form one-dimensional time series data; S2. Encode one-dimensional time series data into two-dimensional feature images using the Gramian angle and field method; S3. Input the two-dimensional feature image into the pre-trained fault diagnosis model. The fault diagnosis model is a two-dimensional convolutional neural network model with optimized parameters based on the particle swarm optimization algorithm. S4. Obtain and output the diagnostic results of the fault diagnosis model.

[0045] The establishment of a two-dimensional convolutional neural network model based on particle swarm optimization includes the following steps: First, a fault sample dataset of mine explosion-proof electro-hydraulic actuators is obtained, and the original one-dimensional data signal is encoded to generate a two-dimensional feature image using the Gramian Angular Summation Field (GRP) method. Then, the two-dimensional feature image is used as the input layer of the convolutional neural network. After the dataset has undergone feature extraction through multiple convolutional and pooling layers, it is stretched into a feature vector output through a fully connected layer. Ordinary convolutional neural network (CNN) fault diagnosis models are prone to getting trapped in local optima and have slow convergence speeds, which seriously affect the accuracy of fault diagnosis. However, by using the Particle Swarm Optimization (PSO) algorithm to optimize the weights, biases, and other parameters of the aforementioned CNN, it is possible to avoid getting trapped in local optima. In the PSO algorithm, the position information of each particle represents a combination of weights and biases in the CNN. The parameters of the CNN are set based on the decoding results of the particle position information. The fault classification accuracy output by the CNN is used as the fitness value of the corresponding particle and input into the PSO algorithm. After iteration, the PSO algorithm obtains a new particle swarm, and then resets the parameters of the CNN. This process is repeated until the optimal parameter combination is obtained. Finally, the test data is selected as input, and the trained model is used to diagnose the faults and output the fault diagnosis results.

[0046] Specifically, encoding one-dimensional data signals to generate two-dimensional feature images includes: The collected one-dimensional time series running parameters The feature image is converted to an n×n two-dimensional feature image using the Gramian angle and field method. The specific steps are as follows: First, the time series at time t Perform normalization to obtain the normalized value. The calculation formula is: , Then, the normalized values Mapped to angle The calculation formula is: , , , Finally, based on two angles , Cosine formula: , Construct the Gramian angle and field matrix G, whose elements The calculation formula is: , We obtain an n×n symmetric two-dimensional matrix, which serves as the two-dimensional feature image.

[0047] The GASF method, used for transformation, can completely preserve the temporal dependencies and numerical values ​​of a one-dimensional time series in the generated two-dimensional image. The texture and structural features of the image can intuitively reflect changes in the device's operating status, making it ideal for feature learning and fault mode recognition using convolutional neural networks, which excel in image recognition.

[0048] In a preferred embodiment, the sampling frequency can be set to 1kHz to 10kHz, and the time series length n can be selected as needed, such as 256, 512 or 1024 data points.

[0049] Furthermore, the two-dimensional convolutional neural network model includes an input layer, a feature extraction module, and a classification module. The input layer receives an n×n two-dimensional feature image. The feature extraction module consists of at least one set of convolutional layers and pooling layers connected sequentially. The convolutional layers extract feature maps from the input data through linear operations of convolutional kernels and biases, and nonlinear activation functions. The pooling layers downsample the feature maps to reduce the data dimensionality. The classification module includes a fully connected layer and an output layer. The fully connected layer flattens the downsampled feature maps into feature vectors and outputs high-level features through weights, biases, and activation functions. The output layer outputs fault classification results based on the high-level features.

[0050] The formula for operating a convolutional layer is as follows: j = 1, 2, ..., N; In the above formula, For feature maps, The convolution kernel weight matrix, Here, f is the bias, l is the non-linear activation function, M is the number of feature maps, and N is the number of convolutional kernels.

[0051] Following the convolutional layers are the pooling layers. The purpose of the pooling layers is to sample the feature maps after convolution. Pooling can further extract effective features and reduce the dimensionality of parameters, avoiding overfitting in convolutional neural networks. The formula for pooling is as follows: , in: For the t-th feature in the i-th feature surface of the pooling input layer l, w represents the sampling width of the pooling operation; This is the output value of the pooling output layer l.

[0052] The last pooling layer is typically followed by a fully connected layer. Each neuron in the fully connected layer is connected to all feature maps in the last pooling layer. High-level features from the fully connected layer are extracted as input to the classifier, and the calculation formula is as follows: , In the above formula, For output, As weight, For bias.

[0053] In a preferred embodiment, the two-dimensional convolutional neural network model comprises two convolutional-pooling layer groups. The first convolutional layer uses 32 convolutional kernels of size 3×3 with a stride of 1 and employs the ReLU activation function. It is then connected to a 2×2 max pooling layer. The second convolutional layer uses 64 convolutional kernels of size 3×3, and is also connected to a 2×2 max pooling layer. After flattening, it is connected to a fully connected layer containing 128 neurons. Finally, the fault classification result is obtained through a softmax output layer.

[0054] The specific steps for globally optimizing the hyperparameters of the initial model of a two-dimensional convolutional neural network using the particle swarm optimization algorithm are as follows: a. Initialize the particle swarm optimization (PSO) algorithm parameters, including particle velocity, position, inertia weight, learning factor, population size, and maximum number of iterations; b. Initialize the particle swarm, where the position of each particle is randomly generated, representing a potential solution for the weights and biases of the two-dimensional convolutional neural network model; c. For each particle, decode and set the weights and biases of the two-dimensional convolutional neural network model based on its position information, and use the training dataset to calculate the fault classification accuracy of the model as the fitness value of the particle. d. Update the individual optimal position and the global optimal position of each particle; e. Update the velocity and position of all particles in the particle swarm according to the velocity and position update formulas of the particle swarm algorithm; f. Iterate through steps c to e until the termination condition is met, and output the globally optimal particle position, which represents the optimized combination of weights and bias parameters.

[0055] By introducing the particle swarm optimization (PSO) algorithm to globally optimize the weights and biases of a convolutional neural network (CNN), the problem of getting trapped in local optima during training using traditional gradient descent can be effectively overcome. The combination of the PSO algorithm's global search capability and the powerful feature extraction capability of CNNs significantly improves the convergence speed and final classification accuracy of the fault diagnosis model.

[0056] The parameters of the particle swarm optimization algorithm can be set as follows: population size of 50, maximum number of iterations of 200, inertia weight decreasing linearly from 0.9 to 0.4, and individual learning factor and social learning factor both of 2.0.

[0057] To verify the effectiveness of this embodiment, 1000 sets of sample data for each of five states, including normal state, hydraulic cylinder leakage, and pump failure, were generated using a simulation model. After GASF conversion and training with the PSO-CNN model, the average fault diagnosis accuracy on the independent test set reached 98.5%, which is higher than the 92.1% of the CNN model without empirical optimization and the 85.7% of traditional diagnostic methods such as support vector machines.

[0058] Based on the above-described preferred embodiments of the present invention, and through the foregoing description, those skilled in the art can make various changes and modifications without departing from the inventive concept. The technical scope of this invention is not limited to the contents of the specification, but must be determined according to the scope of the claims.

Claims

1. An explosion-proof intelligent electro-hydraulic actuator for coal mines, characterized in that, include: Mining explosion-proof speed-regulating motor (1); The closed hydraulic system without external pipelines consists of a bidirectional hydraulic pump (2) driven by the explosion-proof speed-regulating motor (1), a hydraulic cylinder (3), and an integrated hydraulic valve group and accumulator (7); The intelligent explosion-proof control module performs motion control on the electro-hydraulic actuator and runs a fault self-diagnosis program. The fault self-diagnosis program includes: Multiple operating parameters of the electro-hydraulic actuator are collected, including at least the differential pressure between the inlet and outlet of the oil pump, the oil pump flow rate, the displacement response, the speed response, and the motor speed. The collected one-dimensional time series running parameters are converted into two-dimensional feature images using the Gramian angle and field method; The two-dimensional feature image is input into a pre-trained fault diagnosis model, and the fault diagnosis result is output. The fault diagnosis model is a two-dimensional convolutional neural network model with parameters optimized based on the particle swarm optimization algorithm.

2. The explosion-proof intelligent electro-hydraulic actuator for coal mines according to claim 1, characterized in that: The explosion-proof speed regulating motor (1) used in mining is an intrinsically safe explosion-proof speed regulating motor, a mine-sealed explosion-proof speed regulating motor, or a mine-explosion-proof and sealed explosion-proof speed regulating motor.

3. The explosion-proof intelligent electro-hydraulic actuator for coal mines according to claim 1, characterized in that: The hydraulic valve group includes a replenishing shuttle valve (4), a hydraulic control check valve (5) and a safety relief valve (6). The hydraulic control check valve (5) is connected between the bidirectional hydraulic pump (2) and the hydraulic cylinder (3) and is used to lock the position of the hydraulic cylinder (3) when the power is off. The safety relief valve (6) is connected between the hydraulic cylinder (3) and the accumulator (7) and is used to release pressure when the load exceeds the limit; the oil replenishment shuttle valve (4) is used to compensate for the flow difference between the large and small chambers of the hydraulic cylinder (3).

4. The explosion-proof intelligent electro-hydraulic actuator for coal mines according to claim 1, characterized in that: Damping elements (8) are also provided at the large and small chamber oil ports of the hydraulic cylinder (3) to suppress pressure fluctuations generated when the cylinder stops due to inertia under heavy load conditions.

5. The explosion-proof intelligent electro-hydraulic actuator for coal mines according to claim 1, characterized in that, The collected one-dimensional time series running parameters The feature image is converted to an n×n two-dimensional feature image using the Gramian angle and field method. The specific steps are as follows: First, the time series at time t Perform normalization to obtain the normalized value. The calculation formula is: ， Then, the normalized values Mapped to angle The calculation formula is: ， ， ， Finally, based on two angles , Cosine formula: ， Construct the Gramian angle and field matrix G, whose elements The calculation formula is: ， We obtain an n×n symmetric two-dimensional matrix, which serves as the two-dimensional feature image.

6. The explosion-proof intelligent electro-hydraulic actuator for coal mines according to claim 5, characterized in that, The establishment of the fault diagnosis model includes the following steps: Construct an initial model for a two-dimensional convolutional neural network; The particle swarm optimization algorithm is used to globally optimize the hyperparameters of the initial model of the two-dimensional convolutional neural network, with the fault classification accuracy as the fitness function of the particles. A fault diagnosis model is obtained by configuring and training a two-dimensional convolutional neural network model with optimized hyperparameters.

7. The explosion-proof intelligent electro-hydraulic actuator for coal mines according to claim 6, characterized in that, The two-dimensional convolutional neural network model includes: The input layer is used to receive an n×n two-dimensional feature image; The feature extraction module consists of at least one set of convolutional layers and pooling layers connected in sequence; the convolutional layers are used to extract feature maps from the input data through linear operations of convolutional kernels and biases and nonlinear activation functions; the pooling layers are used to downsample the feature maps to reduce the data dimensionality. The classification module includes a fully connected layer and an output layer; the fully connected layer is used to flatten the downsampled feature map into a feature vector, and calculate and output high-level features through weights, biases and activation functions; the output layer is used to output the fault classification result based on the high-level features.

8. The explosion-proof intelligent electro-hydraulic actuator for coal mines according to claim 6, characterized in that, The specific steps for globally optimizing the hyperparameters of the initial model of a two-dimensional convolutional neural network using the particle swarm optimization algorithm are as follows: a. Initialize the particle swarm optimization (PSO) algorithm parameters, including particle velocity, position, inertia weight, learning factor, population size, and maximum number of iterations; b. Initialize the particle swarm, where the position of each particle is randomly generated, representing a potential solution for the weights and biases of the two-dimensional convolutional neural network model; c. For each particle, decode and set the weights and biases of the two-dimensional convolutional neural network model based on its position information, and use the training dataset to calculate the fault classification accuracy of the model as the fitness value of the particle. d. Update the individual optimal position and the global optimal position of each particle; e. Update the velocity and position of all particles in the particle swarm according to the velocity and position update formulas of the particle swarm algorithm; f. Iterate through steps c to e until the termination condition is met, and output the globally optimal particle position, which represents the optimized combination of weights and bias parameters.

9. The explosion-proof intelligent electro-hydraulic actuator for coal mines according to claim 1, characterized in that, The intelligent explosion-proof control module is installed inside the mine explosion-proof control box (9).

10. A fault self-diagnosis method for the explosion-proof intelligent electro-hydraulic actuator for coal mines according to any one of claims 1 to 9, characterized in that, Includes the following steps: S1. Real-time acquisition of multiple operating parameters of the electro-hydraulic actuator to form one-dimensional time series data; S2. Encode one-dimensional time series data into two-dimensional feature images using the Gramian angle and field method; S3. Input the two-dimensional feature image into the pre-trained fault diagnosis model. The fault diagnosis model is a two-dimensional convolutional neural network model with optimized parameters based on the particle swarm optimization algorithm. S4. Obtain and output the diagnostic results of the fault diagnosis model.

Images