Coal rock synchronous loading and CT scanning method based on deep learning adaptive control

Abstract

The invention belongs to the field of coal rock synchronous loading and CT scanning tests and CT image processing, and particularly relates to a coal rock synchronous loading and CT scanning method based on deep learning adaptive control. Comprising the following steps: constructing a coal rock synchronous loading and CT scanning system and an AI edge calculation unit; the model comprises a plurality of feature extraction modules based on a deep convolutional neural network and time sequence information capture modules which are connected with the feature extraction modules in a one-to-one correspondence manner and are based on a long short-term memory network, and the time sequence information capture modules are connected with a state prediction module; the state prediction module is used for predicting fracture states; training the model, embedding the model into an AI edge calculation unit in a software form, and setting a corresponding loading rate adjustment parameter when each fracture state appears; according to the method, the loading rate adjusting time can be accurately mastered, sufficient coal rock test fracture development process images are obtained on the basis of ensuring the overall test efficiency, and subsequent test analysis is facilitated.

Description

Technical Field

[0001] The present invention belongs to the field of coal and rock synchronous loading and CT scanning tests, as well as CT image processing, and particularly relates to a coal and rock synchronous loading and CT scanning method based on deep learning adaptive control. Background Art

[0002] When conducting coal and rock synchronous loading and CT scanning tests, generally, the loading rate of the loading device and the time interval for one scan of the CT scanning device are preset before the test; however, since the deformation rate of the coal and rock specimen is relatively fast after cracks are generated, if the loading rate is fast, the number of images of the crack development process of the coal and rock test obtained by the CT scanning device will be small, affecting subsequent test analysis; and if the loading rate is slow, the loading time before the coal and rock specimen generates cracks will be long, affecting the overall test efficiency; and it is difficult to grasp the adjustment timing of the loading rate by artificially adopting different loading rates at different times during the loading of the coal and rock specimen according to experience; and if the loading rate is adjusted after manually analyzing the crack development situation after obtaining the CT scan images, it will also be disadvantageous for timely adjusting the loading rate due to the long time for manually analyzing the crack development situation. In addition, the crack development rate of the coal and rock specimen is different at different stages of crack development, and the corresponding required loading rate settings are also different, which poses a greater challenge to the adjustment of the loading rate. Summary of the Invention

[0003] To solve the above problems, the present invention proposes a coal and rock synchronous loading and CT scanning method based on deep learning adaptive control, including the following steps: S1. Construct a coal and rock synchronous loading and CT scanning system The coal and rock synchronous loading and CT scanning system includes a loading system, a CT scanning system, and an AI edge computing unit; the AI edge computing unit can real-time obtain the CT images scanned and reconstructed by the CT scanning system, and can also provide the loading rate parameters to the loading system in real time; S2. Construct a coal and rock synchronous loading and CT scanning adaptive control model The model includes several feature extraction modules based on deep convolutional neural networks and temporal information capture modules based on long short-term memory networks connected to each feature extraction module one by one; the output ends of all temporal information capture modules are connected to the state prediction module, and adjacent temporal information capture modules are connected; the state prediction module is a fully connected layer, and outputs the predicted crack state, including crack initiation state, crack propagation state, and crack penetration state; S3. Train the model and embed it in the AI edge computing unit in software form, and set the loading rate adjustment parameters corresponding to each crack state; S4. Perform coal and rock synchronous loading and CT scanning with adaptive control.

[0004] Preferably, in step S1, the loading system includes a fixed base, a fixed gantry is provided on the fixed base, a loading motor is provided on the fixed gantry, a pressure rod is connected to the lower part of the loading motor, a pressure head is rotatably connected to the lower part of the pressure rod, and a rotating base is rotatably connected to the fixed base; a loading control unit is further included, and the loading control unit is at least used to control the loading rate and loading pressure of the loading system.

[0005] Preferably, in step S1, the CT scanning system includes a radiator and a radiation source respectively located on both sides of the loading system, and a scanning control unit is further included, and the scanning control unit is at least used to control the time interval of each scan.

[0006] Preferably, in step S2, the feature extraction module is implemented by convolution and pooling operations based on the ResNet-18 model, the activation function uses the non-linear activation function ReLU, and the pooling operation type is L p pooling; the input of each feature extraction module is a single CT image, and the last layer of the ResNet-18 network model is removed.

[0007] Preferably, in step S2, the fracture state is divided based on the fracture length change rate and the fracture width change rate.

[0008] Preferably, in step S3, CT images obtained after synchronous loading and CT scanning tests are performed on a number of coal and rock specimens are acquired; for each coal and rock specimen, assuming that it has a total of M CT images, the consecutive N CT images starting from the first one are used as sample characteristics, and the fracture state corresponding to the fracture length change rate and width change rate calculated based on the N +1-th CT image and the N -th CT image is used as the sample label, 2 ≤ N +1 ≤ M ; after establishing a number of samples, they are brought into the model for training.

[0009] Preferably, in step S4, a coal and rock synchronous loading and CT scanning system is used for the test. The CT scanning system acquires CT images in real time and inputs the CT images into the trained model. The model predicts the fracture change and the corresponding fracture state between the next CT image and the latest acquired CT image before scanning the next CT image, and adjusts the loading rate when different fracture states appear.

[0010] Beneficial technical effects: Based on the fusion modeling of two deep learning networks, namely convolutional neural network and long short-term memory network, the present invention can predict the fracture state corresponding to the fracture change between the next CT image and the latest acquired CT image before the next CT image is scanned based on the acquired CT images, and timely adjust the loading rate when different fracture states occur. The present invention can accurately grasp the timing of loading rate adjustment, and on the basis of ensuring the overall test efficiency, obtain sufficient images of the fracture development process of coal and rock specimens, which is conducive to the subsequent analysis of test results. Brief Description of the Drawings

[0011] Figure 1 It is a schematic diagram of the loading device and CT scanning device of the present invention; Figure 2 It is an adaptive control model for synchronous loading of coal and rock specimens and CT scanning of the present invention; In the figure: 1 - loading motor; 2 - pressure rod; 3 - pressure head; 4 - coal and rock specimen; 5 - rotating base; 6 - fixed base; 7 - fixed gantry; 8 - radiator (ray source); 9 - ray source. Detailed Embodiments

[0012] In the detailed embodiments section, the technical solutions of the present invention will be described in detail with reference to the drawings.

[0013] As Figure 1-2 shown, the present invention provides a method for synchronous loading of coal and rock specimens and CT scanning based on deep learning adaptive control, including the following steps: S1. Construct a synchronous loading and CT scanning system for coal and rock specimens As Figure 1 shown, the synchronous loading and CT scanning system for coal and rock specimens includes a loading system, a CT scanning system, and an AI edge computing unit; the loading system includes a fixed base 6, a fixed gantry 7 is arranged on the fixed base 6, a loading motor 1 is arranged at the center of the fixed gantry 7, a pressure rod 2 is connected to the lower part of the loading motor 1, a pressure head 3 is rotatably connected to the lower part of the pressure rod 2, a rotating base 5 is rotatably connected to the fixed base 6, the rotating base 5 is located directly below the pressure head 3, and a coal and rock specimen 4 for testing is arranged between the rotating base 5 and the pressure head 3; the loading system further includes a loading control unit, and the loading control unit is at least used to control the loading rate and loading pressure of the loading system; The CT scanning system includes a radiator (ray source) 8 and a ray source 9, which are arranged at the same height as the coal and rock specimen 4 and are located on both sides of the loading system, such as arranged on the left and right sides or arranged on the front and back sides; the CT scanning system further includes a scanning control unit, and the scanning control unit is at least used to control the time interval for one scan; The AI edge computing unit is connected to the CT scanning system to obtain in real time the CT images scanned and reconstructed by the CT scanning system. The AI edge computing unit is also connected to the loading system to provide in real time the loading rate parameters to the loading control unit. The NVIDIA Jetson AGX Xavier is used as the hardware of the AI edge computing unit. S2. Construct an adaptive control model for synchronous coal-rock loading and CT scanning As Figure 2 shown, the model is modeled by fusing two deep learning networks, namely, the convolutional neural network (CNN) and the long short-term memory network (LSTM). It includes several feature extraction modules based on the deep convolutional neural network and the corresponding number of temporal information capture modules based on the long short-term memory network. The feature extraction modules are implemented by convolution and pooling operations based on the ResNet-18 model, the activation function uses the non-linear activation function ReLU, and the pooling operation type is L p pooling. The input of each feature extraction module is a single CT image. In order to fuse the ResNet-18 model with the LSTM model, the last layer of the ResNet-18 network model is removed to output a three-dimensional tensor, and the dimension values of the three dimensions are 1, 1, and 4,096 respectively. The temporal information capture module is mainly constructed based on the long short-term memory network (LSTM model). The long short-term memory network contains three control gates: input, output, and forget. Its structure is well known in the art and will not be elaborated here. The input of each temporal information capture module is the output of the feature extraction module, and the output ends of all temporal information capture modules are connected to the state prediction module, and adjacent temporal information capture modules are connected to each other. The state prediction module is a fully connected layer, and its output dimension is 3, corresponding to three predicted fracture states, including fracture initiation state, fracture propagation state, and fracture penetration state. Extract the fracture features in the CT image through the convolutional neural network (CNN), including the length and width of the fracture. The long short-term memory network (LSTM) predicts the evolution trend of the fracture according to the historical data and judges the fracture state. Among them, the conditions corresponding to the crack initiation state are: the crack length change rate is less than 0.05 mm / min, and the crack width change rate is less than 0.02 mm / min; the conditions corresponding to the crack propagation state are: the crack length change rate is between 0.05 and 0.2 mm / min, or the crack width change rate is between 0.02 and 0.1 mm / min; the crack penetration state is: the crack length change rate is greater than 0.2 mm / min, or the crack width change rate is greater than 0.1 mm / min; the calculation method of the crack length change rate is: calculated by the ratio of the difference between the maximum crack length on the next CT image and the maximum crack length on the previous CT image to the time difference between the two CT image scans; the calculation method of the crack width change rate is: calculated by the ratio of the difference between the maximum crack width on the next CT image and the maximum crack width on the previous CT image to the time difference between the two CT image scans; S3. Train the model and embed it in the AI edge computing unit in the form of software Obtain the CT images obtained after performing synchronous loading and CT scanning tests on several coal and rock specimens; for each coal and rock specimen, assuming it has a total of M CT images, the consecutive N CT images starting from the first one can be used as sample characteristics, and based on the crack length change rate and width change rate corresponding to the N +1th CT image and the N th CT image, the corresponding crack state is used as the sample label, 2 ≤ N +1 ≤ M ; after establishing several samples, bring them into the model for training; Embed the trained model in the AI edge computing unit in the form of software, and set that when the crack initiation state appears, the AI edge computing unit gives an instruction to the loading control unit: adjust the loading rate to 50% of the initial set loading rate; when the crack initiation state appears, the AI edge computing unit gives an instruction to the loading control unit: adjust the loading rate to 20% of the initial set loading rate; when the crack initiation state appears, the AI edge computing unit gives an instruction to the loading control unit: adjust the loading rate to 10% of the initial set loading rate.

[0014] S4. Perform adaptive control of coal and rock synchronous loading and CT scanning Use the coal and rock synchronous loading and CT scanning system to conduct experiments. The CT scanning system obtains CT images in real time and inputs the CT images into the trained model. The model can predict the crack changes and their corresponding crack states between the next CT image and the latest obtained CT image before scanning the next CT image, and adjust the loading rate when different crack states appear.

[0015] The above-provided is only the preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Within the technical scope disclosed by the present invention, any various changes or alternative solutions that can be easily thought of by those skilled in the art should be included within the protection scope of the present invention.

Claims

1. A coal-rock synchronous loading and CT scanning method based on deep learning adaptive control, characterized in that, Including: S1. Construct a coal-rock synchronous loading and CT scanning system: The coal-rock synchronous loading and CT scanning system includes a loading system, a CT scanning system, and an AI edge computing unit; the AI edge computing unit can obtain the CT images scanned and reconstructed by the CT scanning system in real time, and can also provide the loading rate parameters to the loading system in real time. S2. Construct an adaptive control model for coal-rock synchronous loading and CT scanning: The model includes several feature extraction modules based on deep convolutional neural networks and temporal information capture modules based on long short-term memory networks that are connected to each feature extraction module in a one-to-one correspondence; the output ends of all temporal information capture modules are connected to the state prediction module, and adjacent temporal information capture modules are connected to each other; the state prediction module is a fully connected layer that outputs the predicted fracture state, including the fracture initiation state, the fracture propagation state, and the fracture penetration state. S3. Train the model and embed it in the AI edge computing unit in the form of software, and set the loading rate adjustment parameters corresponding to each fracture state. S4. Perform adaptive control of coal-rock synchronous loading and CT scanning.

2. The coal and rock synchronous loading and CT scanning method according to claim 1, characterized in that In step S1, the loading system includes a fixed base, a fixed gantry is provided on the fixed base, a loading motor is provided on the fixed gantry, a pressure rod is connected to the lower part of the loading motor, a pressure head is rotatably connected to the lower part of the pressure rod, and a rotating base is rotatably connected to the fixed base; it also includes a loading control unit, and the loading control unit is at least used to control the loading rate and loading pressure of the loading system.

3. The coal-rock synchronous loading and CT scanning method according to claim 2, wherein In step S1, the CT scanning system includes a radiator and a radiation source respectively located on both sides of the loading system, and also includes a scanning control unit, and the scanning control unit is at least used to control the time interval of one scan.

4. The coal and rock synchronous loading and CT scanning method according to claim 1, characterized in that In step S2, the feature extraction module is implemented by convolution and pooling operations based on the ResNet-18 model. The activation function uses the non-linear activation function ReLU, and the pooling operation type is L p pooling; the input of each feature extraction module is a single CT image, and the last layer of the ResNet-18 network model is removed.

5. The coal-rock synchronous loading and CT scanning method according to claim 4, characterized in that In step S2, the fracture state is divided based on the fracture length change rate and the fracture width change rate.

6. The coal and rock synchronous loading and CT scanning method according to claim 5, characterized in that In step S3, obtain the CT images obtained after performing synchronous loading and CT scanning tests on several coal-rock specimens; for each coal-rock specimen, assume that it has a total of M CT images, and use the consecutive N CT images starting from the first one as sample characteristics, and based on the N +1-th CT image and the N -th CT image, use the fracture state corresponding to the calculated fracture length change rate and width change rate as the sample label, where 2 ≤ N +1 ≤ M ; establish several samples and then bring them into the model for training.

7. The coal-rock synchronous loading and CT scanning method according to claim 6, characterized in that In step S4, an experiment is carried out using the coal-rock synchronous loading and CT scanning system. The CT scanning system obtains CT images in real time, and inputs the CT images into the trained model. The model predicts the fracture changes and their corresponding fracture states between the next CT image and the latest obtained CT image before scanning the next CT image, and adjusts the loading rate when different fracture states occur.

Images