A multi-source data information interaction fusion processing method based on a track driving system

Abstract

The present application belongs to the field of information collection and processing of track driving system, and solves the problem of error in judging the state of track driving system by only one sensor test data. A multi-source data information interaction fusion processing method based on track driving system is provided, a three-dimensional model of track driving system is established, and is imported into a dynamics simulation software to form a dynamics simulation model of track driving system; a road test is conducted on the track driving system, and experimental data is collected; the preprocessed experimental data is divided into training sample data and test sample data; a DBN model is constructed, and the training sample data is input into the DBN model; the test sample data is input into the trained DBN model; the dynamics simulation model of track driving system is simulated, and the simulation data is input into the trained DBN model; the credibility value output by the DBN model is compared with the simulation result, and the DBN model structure is improved. The present application eliminates the error in judging the state of track driving system by only one sensor test data.

Description

Technical Field

[0001] This invention belongs to the field of information acquisition and processing technology for tracked driving systems, specifically relating to a method for multi-source data information interaction and fusion processing based on tracked driving systems. Background Technology

[0002] Tracked walking mechanisms are widely used in engineering machinery, tractors and other field operation vehicles. In coal mining machinery, the working environment of tracked walking mechanisms is harsh, and they operate for a long time on waterlogged and uneven roads. Therefore, the walking mechanism is required to have sufficient strength and rigidity, as well as good traveling and steering capabilities. The performance testing of tracked equipment is crucial when it is put into use.

[0003] Currently, due to the complex working environment of tracked travel mechanisms, it is difficult to accurately collect and analyze various data related to the operating status of tracked travel mechanisms. If only one sensor test data is used to judge the driving status of the tracked travel system, there will be certain errors, which often leads to misjudgment of tracked travel system failures.

[0004] Therefore, for tracked driving mechanisms, it is of great significance to develop a method for multi-source sensor data fusion based on virtual reality. Summary of the Invention

[0005] To address the shortcomings of relying solely on test data from a single sensor to determine the state of a tracked driving system, which inherently involves some error, this invention aims to provide a multi-source data fusion processing method based on tracked driving systems. This method uses a DBN (Database-Based Network) model to fuse data from multiple sensors in a physical tracked driving system experiment. Furthermore, it verifies the accuracy of the trained DBN model through virtual simulation of the tracked driving system and improves the DBN model accordingly. This effectively eliminates the error associated with relying solely on test data from a single sensor to determine the state of a tracked driving system, and overcomes the limitations of analyzing the motion state of a real tracked driving system based solely on simulation data.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a multi-source data information interaction and fusion processing method based on a tracked driving system, comprising the following steps:

[0007] S1: Establish a three-dimensional model of the tracked driving system, import the established three-dimensional model into the dynamic simulation software Recurdyn, establish various constraints and contacts, and establish a cement road surface to form a dynamic simulation model of the tracked driving system.

[0008] S2: Conduct road tests on the physical tracked driving system, and collect at least the torque and speed data of the DC brushless motor, the stress data of the drive wheel, the current data of the DC brushless motor, and the vertical vibration data of the box as experimental data.

[0009] S3: Preprocess the collected experimental data, extract feature values, and divide the preprocessed experimental data into training samples and test samples;

[0010] S4: Build a DBN model by inputting the training sample data into the DBN model and training the DBN model;

[0011] S5: Input the test sample data into the trained DBN model. After the test sample data is interactively fused, the output of the DBN model is used as the confidence value for comprehensively evaluating the status of the physical tracked driving system.

[0012] S6: Perform dynamic simulation on the tracked driving system dynamic simulation model in the Recurdyn software, collect simulation data, and ensure that the simulation data corresponds one-to-one with the test sample data and the experimental conditions are the same. Input the collected simulation data into the trained DBN model, and after the simulation data interaction and fusion, use the output of the DBN model as a quantity for comprehensive evaluation of the virtual tracked driving system state.

[0013] S7: Compare the confidence value of the DBN model output, which can comprehensively evaluate the state of the physical tracked driving system, with the quantitative value of the DBN model output, which can comprehensively evaluate the state of the virtual tracked driving system, and further correct the parameters of the DBN model until the simulation data and test sample data in the DBN model output meet the set conditions.

[0014] S8: Collect multi-source data from the actual operation of the tracked driving system, preprocess it, and then input it into the corrected DBN model for interactive fusion to accurately determine the operating status of the tracked driving system.

[0015] Preferably, in step S7, the condition is set as follows: the difference between the confidence value that can comprehensively evaluate the status of the physical tracked driving system and the quantity value that can comprehensively evaluate the status of the virtual tracked driving system is less than a set threshold.

[0016] Preferably, in step S7, the outputs of all test sample data and simulation data in the DBN model are compared until all outputs meet the set conditions.

[0017] Preferably, in step S7, the parameters for DBN model correction include the number of hidden layer nodes, the number of training iterations, the weights of the input and output layers, and the learning rate during the DBN model training process.

[0018] Preferably, in step S3, the torque and speed data of the brushless DC motor are analyzed using a multi-layer wavelet packet analysis method to calculate the ratio of the energy of multiple wavelet packet components to the total energy as a feature value; the stress data of the drive wheel is processed by fast Fourier transform; the current data of the brushless DC motor is analyzed using a threshold comparison method to select the maximum value at a fixed time as a feature value; the vertical vibration data of the housing is processed by wavelet transform for noise reduction; and the training sample data and the test sample data are normalized.

[0019] Preferably, in step S4, a DBN model is constructed using MATLAB software; training sample data is input into the DBN model, and each input training sample data is a set of five feature parameters at the same time, including the torque and speed feature values ​​of the brushless DC motor, the stress feature value of the drive wheel, the current feature value of the brushless DC motor, and the vertical vibration feature value of the housing. The model is trained layer by layer through RBM, and then the parameters of the entire network are fine-tuned in reverse through BP neural network to achieve the training of the DBN model.

[0020] Preferably, in step S5, the trained DBN model is validated using test sample data. Each input test sample data consists of a set of five characteristic parameters occurring simultaneously, including the torque and speed characteristic values ​​of the brushless DC motor, the stress characteristic value of the drive wheel, the current characteristic value of the brushless DC motor, and the vertical vibration characteristic value of the housing. In step S6, simulation data is input into the trained DBN model. Each input simulation data consists of a set of five characteristic parameters occurring simultaneously, including the torque and speed characteristic values ​​of the brushless DC motor, the stress characteristic value of the drive wheel, the current characteristic value of the brushless DC motor, and the vertical vibration characteristic value of the housing.

[0021] Preferably, in step S7, the parameters for DBN model improvement are mainly the number of hidden layer nodes, the number of training iterations, the weights of the input and output layers, and the learning rate during the DBN model training process.

[0022] Preferably, the physical tracked travel system consists of a tracked travel mechanism and a data acquisition device. The tracked travel mechanism includes a housing, drive wheels, guide wheels, track rollers, carrier rollers, track pads, a screw tensioning device, a DC brushless motor, a planetary geared motor, a lithium battery, and a driver. The data acquisition device includes a dynamic torque sensor, a stress acquisition and transmission device, an acceleration sensor, and a universal three-phase intelligent power meter.

[0023] A dynamic torque sensor is positioned between the brushless DC motor and the planetary geared motor to collect torque and speed data from the brushless DC motor. The sensor transmits torque data to the host computer via a 485-to-USB adapter cable, and also transmits speed data via a frequency-to-analog isolator and another 485-to-USB adapter cable. Strain gauges of the stress acquisition and transmission device are positioned at the tooth root of the drive wheel. This device collects stress data from the drive wheel and uploads it to the host computer. An acceleration sensor is positioned at the bottom of the enclosure to collect vertical vibration signal data. A universal three-phase intelligent power meter is positioned at the brushless DC motor to collect current change data passing through it.

[0024] Preferably, the dynamic torque sensor, stress acquisition and transmission device, acceleration sensor, and AC / DC universal three-phase intelligent power meter are all arranged on the same side of the tracked travel system.

[0025] Compared with the prior art, the beneficial effects of the present invention are:

[0026] 1. This invention fuses multiple sensor data from a physical tracked driving system experiment using a DBN model to obtain a reliability value that can comprehensively evaluate the state of the physical tracked driving system; simultaneously, it fuses virtual simulation data of the tracked driving system using the DBN model to obtain a quantity that can comprehensively evaluate the state of the virtual tracked driving system; by comparing the two, the accuracy of the trained DBN model during the experiment is verified, and the DBN model is improved; effectively eliminating the error of judging the state of the tracked driving system based on test data from a single sensor, and overcoming the shortcomings of relying solely on simulation data to analyze the motion state of the real tracked driving system.

[0027] 2. This invention provides a more accurate assessment of the operating status of the tracked driving system and can also intelligently diagnose complex faults within the tracked driving system from the fused data in a timely manner. This effectively reduces the erroneous understanding of a complex fault based solely on a single fault mode, and is beneficial for fault diagnosis and detection during the actual operation of the tracked system of a continuous mining machine.

[0028] 3. This invention simulates the motion state of a tunneling machine in a real coal mine by performing dynamic simulation on a dynamic simulation model of a tracked driving system, and makes up for the difficulty of data acquisition. Attached Figure Description

[0029] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0030] Figure 1 This is a flowchart of the multi-source data processing method for the tracked driving system in this embodiment;

[0031] Figure 2 This is a first-view structural schematic diagram of the tracked driving system in this embodiment;

[0032] Figure 3 This is a second-view structural schematic diagram of the tracked driving system in this embodiment;

[0033] Figure 4 This is a flowchart of the training process of the DBN model in this embodiment.

[0034] In the diagram: 1-Drive wheel; 2-Stress acquisition device; 3-Drag chain wheel; 4-Support roller; 5-Track plate; 6-Guide wheel; 7-Rubber block; 8-Driver; 9-Box; 10-Screw tensioning device; 11-Lithium battery; 12-DC brushless motor; 13-Sensor mounting bracket; 14-Dynamic torque sensor; 15-Planetary geared motor. Detailed Implementation

[0035] The technical solutions of the embodiments of the present invention will be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other implementation methods obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0036] It should be noted that the structures, proportions, sizes, etc., illustrated in the accompanying drawings of this specification are only used to complement the content disclosed in the specification for those skilled in the art to understand and read, and are not intended to limit the conditions under which the present invention can be implemented. Therefore, they have no substantial technical significance. Any modifications to the structure, changes in the proportional relationships, or adjustments to the size, without affecting the effects and objectives that the present invention can produce, should still fall within the scope of the technical content disclosed in the present invention. It should be noted that in this specification, relational terms such as "first" and "second" are only used to distinguish one entity from several other entities, and do not necessarily require or imply any such actual relationship or order between these entities.

[0037] like Figures 1 to 4 As shown, the present invention provides an embodiment:

[0038] A method for multi-source data information interaction and fusion processing based on a tracked driving system includes the following steps:

[0039] S1: Establish a three-dimensional model of the tracked driving system, import the established three-dimensional model into the dynamic simulation software Recurdyn, establish various constraints and contacts, and establish a cement road surface to form a dynamic simulation model of the tracked driving system.

[0040] S2: Conduct road tests on the physical tracked driving system, and collect at least the torque and speed data of the DC brushless motor 12, the stress data of the drive wheel 1, the current data of the DC brushless motor 12, and the vertical vibration data of the housing 9 as experimental data.

[0041] S3: Preprocess the collected experimental data, extract feature values, and divide the preprocessed experimental data into training samples and test samples;

[0042] S4: Build a DBN model by inputting the training sample data into the DBN model and training the DBN model;

[0043] S5: Input the test sample data into the trained DBN model. After the test sample data is interactively fused, the output of the DBN model is used as the confidence value for comprehensively evaluating the status of the physical tracked driving system.

[0044] S6: Perform dynamic simulation on the tracked driving system dynamic simulation model in the Recurdyn software, collect simulation data, and ensure that the simulation data corresponds one-to-one with the test sample data and the experimental conditions are the same. Input the collected simulation data into the trained DBN model, and after the simulation data interaction and fusion, use the output of the DBN model as a quantity for comprehensive evaluation of the virtual tracked driving system state.

[0045] S7: Compare the confidence value of the DBN model output, which can comprehensively evaluate the state of the physical tracked driving system, with the quantitative value of the DBN model output, which can comprehensively evaluate the state of the virtual tracked driving system, and further correct the parameters of the DBN model until the simulation data and test sample data in the DBN model output meet the set conditions.

[0046] S8: Collect multi-source data from the actual operation of the tracked driving system, preprocess it, and input it into the corrected DBN model for interactive fusion to accurately determine the operating status of the tracked driving system. The multi-source data here includes the torque and speed data of the DC brushless motor 12, the stress data of the drive wheel 1, the current data of the DC brushless motor 12, and the vertical vibration data of the housing 9 in the actual operation of the tracked driving system.

[0047] In step S1, a high-fidelity 3D model of the tracked driving system is created using SolidWorks software. The model includes: a housing 9, drive wheel 1, guide wheel 6, track roller 4, track chain wheel 3, 230mm track pads 5 and matching rubber blocks 7, a lead screw tensioning device 10, a DC brushless motor 12, a planetary geared motor 15, a dynamic torque sensor 14, a lithium battery 11, a sensor mounting bracket 13, a driver 8, and a stress acquisition device 2. Based on the actual drawings of the purchased tracked driving system, a 1:1 scale model of the tracked driving system is created in SolidWorks and saved as a Parasolid (*.x_t) file. The Parasolid (*.x_t) file is then imported into the Track (LM) module of RecurDyn software, where various constraints, contacts, and concrete pavement are applied.

[0048] In step S3, the torque and speed data of the brushless DC motor 12 are analyzed using a multi-layer wavelet packet analysis method to calculate the ratio of the energy of multiple wavelet packet components to the total energy as the feature value; the stress data of the drive wheel 1 is processed by fast Fourier transform; the current data of the brushless DC motor 12 is analyzed using a threshold comparison method to select the maximum value at a fixed time as the feature value; the vertical vibration data of the housing 9 is processed by wavelet transform noise reduction to filter out irrelevant frequency signals and obtain the actual vibration data.

[0049] In this embodiment, the training samples include 300 data samples, and the test samples include 50 data samples. To speed up program execution and improve data processing, the sample data is normalized using the following formula:

[0050] x*=(xx min ) / (x max -x min )

[0051] In the formula, x min x is the minimum value of the sample data. max Let x be the maximum value of the sample data, and x be the sample data. * Each sample data is normalized to 0-1.

[0052] In step S4, a DBN model is constructed using MATLAB software. The DBN model is a generative model of deep belief networks, which is composed of multiple Restricted Boltzmann Machines (RBMs) stacked together. During the stacking process, the output of the first RBM is used as the input of the next RBM. Pre-training is performed layer by layer using an unsupervised learning greedy method. Finally, a complete unsupervised learning process is performed to determine the network weights. Then, based on the sample data, the backpropagation algorithm is used for optimization to complete a supervised learning process, generating the bottom-level states and optimizing the weights obtained from the pre-training.

[0053] RBM is an energy model where an RBM has n visible cells and m hidden cells, where v = (v1, v2, ..., vm). n ), h=(h1,h2,…,h m If ), then the energy function under this configuration is

[0054]

[0055] In the formula, E θ (v,h) is the energy function; a i and b j The biases of explicit element i and implicit element j are respectively; w ij The connection weights between layers; θ = {w ij ,a i ,b j} represents the model parameters; n v and n h denoted by , where v is the visible layer composed of dominant neurons and h is the hidden layer composed of hidden neurons.

[0056] The joint probability distribution is expressed as

[0057]

[0058] In the formula, P θ (v,h) is the joint probability distribution; E is the normalization factor. θ (v,h) represents the energy function; v is the visible layer, composed of dominant neurons, and h is the hidden layer, composed of hidden neurons.

[0059] The probability of hidden layer and visible layer is

[0060]

[0061]

[0062] In the formula, v is the visible layer, composed of dominant neurons, and h is the hidden layer, composed of hidden neurons. P θ (h│v) represents the probability of the hidden layer; P θ (v│h) represents the probability of the visible layer; P θ (v,h) is the joint probability distribution; P θ (v) and P θ (h) represents the distribution of a set of observation data v and h.

[0063] Further to the activation function

[0064]

[0065]

[0066] In the formula, P θ (v i =1|h) is the probability of the hidden layer in the hidden neuron; P θ (h j =1|v) represents the probability of the visible layer in a dominant neuron; where The activation function of the neural network maps the variable x to the range of 0 to 1; a i and b j The biases of explicit element i and implicit element j are respectively; w ij represents the connection weights between layers; v is the visible layer, composed of explicit neurons, and h is the hidden layer, composed of implicit neurons.

[0067] Given training samples, the samples are input into the first RBM for training. The parameters θ of the RBM in this layer are adjusted, and then the initial parameter values ​​θ of the DBN pre-training are obtained layer by layer, so that the probability distribution represented by the RBM under the control of these parameters is as consistent as possible with the distribution of the training data.

[0068] The training sample data is input into the DBN model. Each input training sample data is a set of five feature parameters at the same time, including the torque and speed feature values ​​of the brushless DC motor 12, the stress feature value of the drive wheel 1, the current feature value of the brushless DC motor 12, and the vertical vibration feature value of the housing 9. The model is trained layer by layer through RBM, and then the parameters of the entire network are fine-tuned in reverse through BP neural network to achieve the training of DBN model. Finally, the trained network model parameters are saved.

[0069] In step S5, the trained DBN model is validated using test sample data. Each input test sample data consists of a set of five characteristic parameters at the same time, including the torque and speed characteristic values ​​of the brushless DC motor 12, the stress characteristic value of the drive wheel 1, the current characteristic value of the brushless DC motor 12, and the vertical vibration characteristic value of the housing 9. In step S6, simulation data is input into the trained DBN model. Each input simulation data consists of a set of five characteristic parameters at the same time, including the torque and speed characteristic values ​​of the brushless DC motor 12, the stress characteristic value of the drive wheel 1, the current characteristic value of the brushless DC motor 12, and the vertical vibration characteristic value of the housing 9.

[0070] In step S7, the parameters for DBN model improvement mainly include the number of hidden layer nodes, training iterations, weights of the input and output layers, and the learning rate during DBN model training. The set condition is that the difference between the confidence value that can comprehensively evaluate the state of the physical tracked driving system and the quantitative value that can comprehensively evaluate the state of the virtual tracked driving system is less than a set threshold (or the ratio is less than a set threshold), meaning that the two are within the same order of magnitude. The outputs of all test sample data and simulation data in the DBN model are compared until all outputs meet the set condition. Here, the outputs being compared are the output values ​​(i.e., confidence value and quantitative value) of test sample data and simulation data in the DBN model under the same experimental conditions and at the same time. The test sample data can be divided into multiple groups until the output of each group meets the set condition.

[0071] The physical tracked driving system upon which this method is based consists of a tracked driving mechanism and a data acquisition device. The tracked driving mechanism includes a housing 9, a drive wheel 1, a guide wheel 6, a support roller 4, a track chain wheel 3, track pads 5, a lead screw tensioning device 10, a DC brushless motor 12, a planetary geared motor 15, a lithium battery 11, a driver 8, and a remote controller. The data acquisition device includes a dynamic torque sensor 14, a stress acquisition and transmission device 2, an acceleration sensor, and a universal three-phase intelligent power meter.

[0072] A dynamic torque sensor 14 is positioned between the brushless DC motor 12 and the planetary geared motor 15 to collect torque and speed data from the brushless DC motor 12. The dynamic torque sensor is a DYN-200 model. It is fixed to a sensor mounting bracket 13, with a support block at the lower end supporting the sensor housing. The sensor mounting bracket 13 is bolted to the housing 9 of the tracked drive system. The dynamic torque sensor transmits torque data to a host computer via a 485-to-USB adapter cable, and also transmits speed data to the host computer via a frequency-to-analog isolator and another 485-to-USB adapter cable. A stress acquisition and transmission device 2 is used to collect drive data. The stress data of wheel 1 is collected and uploaded to the host computer. The stress acquisition and transmission device 2 includes a DH5905N power module, a DH5905N acquisition module, and a strain gauge. The strain gauge is fixed at the tooth root of the drive wheel 1. The DH5905N acquisition module and the DH5905N power module are integrated and fixed to the surface of the drive wheel 1 with strong double-sided adhesive. The DH5905N acquisition module directly transmits the collected stress data to the host computer via a wireless AP. An acceleration sensor is arranged at the bottom of the housing 9 to collect vertical vibration signal data of the housing. A three-phase intelligent AC / DC power meter is arranged at the brushless DC motor 12 to collect current change data through the brushless DC motor 12.

[0073] In order to ensure that the data collected from multiple sensors correspond to each other and to achieve efficient fusion of multi-source data, the dynamic torque sensor 14, the stress acquisition and transmission device 2, the acceleration sensor, and the AC / DC universal three-phase intelligent power meter are all arranged on the same side of the tracked driving system.

[0074] The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for multi-source data information interaction and fusion processing based on a tracked driving system, characterized in that, Includes the following steps: S1: Establish a three-dimensional model of the tracked driving system, import the established three-dimensional model into the dynamic simulation software Recurdyn, establish various constraints and contacts, and establish a cement road surface to form a dynamic simulation model of the tracked driving system. S2: Conduct road tests on the physical tracked driving system, and collect at least the torque and speed data of the DC brushless motor (12), the stress data of the drive wheel (1), the current data of the DC brushless motor (12), and the vertical vibration data of the box (9) as experimental data. S3: Preprocess the collected experimental data, extract feature values, and divide the preprocessed experimental data into training samples and test samples; S4: Build a DBN model by inputting the training sample data into the DBN model and training the DBN model; S5: Input the test sample data into the trained DBN model. After the test sample data is interactively fused, the output of the DBN model is used as the confidence value for comprehensively evaluating the status of the physical tracked driving system. S6: Perform dynamic simulation on the tracked driving system dynamic simulation model in the Recurdyn software, collect simulation data, and ensure that the simulation data corresponds one-to-one with the test sample data and the experimental conditions are the same. Input the collected simulation data into the trained DBN model, and after the simulation data interaction and fusion, use the output of the DBN model as a quantity for comprehensive evaluation of the virtual tracked driving system state. S7: Compare the confidence value of the DBN model output, which can comprehensively evaluate the state of the physical tracked driving system, with the quantitative value of the DBN model output, which can comprehensively evaluate the state of the virtual tracked driving system, and further correct the parameters of the DBN model until the simulation data and test sample data in the DBN model output meet the set conditions. S8: Collect multi-source data from the actual operation of the tracked driving system, preprocess it, and then input it into the corrected DBN model for interactive fusion to accurately determine the operating status of the tracked driving system.

2. The method for multi-source data information interaction and fusion processing based on a tracked driving system as described in claim 1, characterized in that: In step S7, the condition is set as follows: the difference between the confidence value that can comprehensively evaluate the status of the physical tracked driving system and the quantity value that can comprehensively evaluate the status of the virtual tracked driving system is less than a set threshold.

3. The method for multi-source data information interaction and fusion processing based on a tracked driving system as described in claim 1, characterized in that: In step S7, the outputs of all test sample data and simulation data in the DBN model are compared until all outputs meet the set conditions.

4. The method for multi-source data information interaction and fusion processing based on a tracked driving system according to claim 1, characterized in that: In step S7, the parameters for DBN model correction include the number of hidden layer nodes, the number of training iterations, the weights of the input and output layers, and the learning rate during the DBN model training process.

5. The method for multi-source data information interaction and fusion processing based on a tracked driving system as described in claim 1, characterized in that: In step S3, the torque and speed data of the brushless DC motor (12) are analyzed using a multi-layer wavelet packet analysis method to calculate the ratio of the energy of multiple wavelet packet components to the total energy as a feature value; the stress data of the drive wheel (1) is processed by fast Fourier transform; the current data of the brushless DC motor (12) is analyzed using a threshold comparison method to select the maximum value at a fixed time as a feature value. The vertical vibration data of the box (9) is processed by wavelet transform for noise reduction; The training and test sample data were then normalized.

6. The method for multi-source data information interaction and fusion processing based on a tracked driving system according to claim 2, characterized in that: In step S4, a DBN model is constructed using MATLAB software. Training sample data is input into the DBN model. Each input training sample data consists of a set of five feature parameters at the same time, including the torque and speed feature values ​​of the brushless DC motor (12), the stress feature value of the drive wheel (1), the feature parameters of the current of the brushless DC motor (12), and the vertical vibration feature value of the housing (9). The model is trained layer by layer through RBM, and then the parameters of the entire network are fine-tuned in reverse through BP neural network to achieve the training of the DBN model.

7. The method for multi-source data information interaction and fusion processing based on a tracked driving system according to claim 3, characterized in that: In step S5, the trained DBN model is verified by test sample data. Each input test sample data is a set of five feature parameters at the same time, including the torque and speed feature values ​​of the brushless DC motor (12), the stress feature value of the drive wheel (1), the current feature value of the brushless DC motor (12), and the vertical vibration feature value of the housing (9). In step S6, the simulation data is input into the trained DBN model. Each input simulation data consists of a set of five characteristic parameters at the same time, including the torque and speed characteristic values ​​of the brushless DC motor (12), the stress characteristic value of the drive wheel (1), the current characteristic value of the brushless DC motor (12), and the vertical vibration characteristic value of the housing (9).

8. A method for multi-source data information interaction and fusion processing based on a tracked driving system according to any one of claims 1 to 7, characterized in that: The physical tracked driving system consists of a tracked driving mechanism and a data acquisition device. The tracked driving mechanism includes a housing (9), a drive wheel (1), a guide wheel (6), a support roller (4), a track chain wheel (3), track plates (5), a screw tensioning device (10), a DC brushless motor (12), a planetary geared motor (15), a lithium battery (11), and a driver (8). The data acquisition device includes a dynamic torque sensor (14), a stress acquisition and transmission device (2), an acceleration sensor, and a three-phase AC / DC universal intelligent power meter. A dynamic torque sensor (14) is arranged between the brushless DC motor (12) and the planetary geared motor (15) to collect torque and speed data of the brushless DC motor (12). The dynamic torque sensor transmits torque data to the host computer via a 485 to USB adapter cable, and transmits speed data to the host computer via a frequency to analog isolator cable and then via a 485 to USB adapter cable. The strain gauge of the stress acquisition and transmission device (2) is arranged at the tooth root of the drive wheel (1). The stress acquisition and transmission device (2) is used to collect stress data of the drive wheel (1) and upload it to the host computer. An acceleration sensor is arranged at the bottom of the housing (9) to collect vertical vibration signal data of the housing. A three-phase AC / DC universal intelligent power meter is arranged at the brushless DC motor (12) to collect current change data through the brushless DC motor (12).

9. The method for multi-source data information interaction and fusion processing based on a tracked driving system according to claim 8, characterized in that: The dynamic torque sensor (14), stress acquisition and transmission device (2), acceleration sensor and AC / DC universal three-phase intelligent power meter are all arranged on the same side of the tracked travel system.

Images