Loader load process prediction method based on MBD-DEM coupling simulation data driving

Abstract

The invention discloses a loader load process prediction method based on MBD-DEM coupling simulation data driving, and is applied to the technical field of loader load prediction. Comprising the following steps: establishing a multi-rigid-body dynamic model of a loading machine working device; constructing a discrete element model of the material particles by adopting discrete element simulation software; mBD-DEM coupling is realized based on the multi-rigid-body dynamic model and the discrete element model; different working condition combinations are set, and load history data are obtained; preprocessing the load history data and the corresponding working condition data; constructing and training an LSTM neural network; and predicting the loading process of the loader based on the trained LSTM neural network. According to the method, the mechanical kinematics and dynamics characteristics of the loading machine working device and the discrete characteristics of materials can be considered at the same time, and the load prediction accuracy is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of loader load prediction, and more particularly to a loader load history prediction method based on MBD-DEM coupled simulation data drive. Background Art

[0002] Loaders are widely used in mining, water conservancy projects, defense industry and other fields. Their operating conditions are complex and harsh, and the performance and reliability of their working devices directly affect the operating efficiency and equipment life. In the design and analysis process of loaders, it is crucial to accurately obtain the load history of the working device under different working conditions. Traditional load calculation methods based on empirical formulas or simplified mechanical models are difficult to accurately simulate the complex material-working device interaction, resulting in limitations in the design optimization and fatigue analysis of the loader working device.

[0003] The existing load prediction methods for the working device of the loader rely on empirical formulas and mechanical balance calculations to provide basic load estimates. The data collected at the work site is collected on-site through sensors installed on the working device of the loader, such as pin sensors, displacement sensors, pressure sensors, and speed sensors. However, in theoretical calculations, although basic load estimates can be provided, the load is too idealized and often lacks the ability to accurately simulate complex working conditions and dynamic changes. When faced with complex working conditions and variable material properties, the prediction accuracy is not high; during the process of collecting data at the work site, the engineering working environment is complex, there are many interference factors, the noise is large, the load fluctuates greatly, and the collection process requires a lot of manpower, material resources, and financial resources, and the cost is high.

[0004] With the development of computer simulation technology and artificial intelligence technology, MBD-DEM coupled simulation provides a new way to deeply study the mechanical behavior of loader working devices, but how to effectively use the large amount of data generated by coupled simulation to predict the load history under unknown working conditions still needs further exploration. Therefore, how to provide a loader load history prediction method driven by MBD-DEM coupled simulation data is an urgent problem that technicians in this field need to solve. Summary of the invention

[0005] In view of this, the present invention provides a loader load history prediction method driven by MBD-DEM coupled simulation data to solve the problems in the prior art.

[0006] In order to achieve the above object, the present invention provides the following technical solutions:

[0007] A method for predicting load history of a loader based on MBD-DEM coupled simulation data drive includes the following steps:

[0008] S1. Establish a multi-rigid body dynamics model of the loader working device in the multi-body dynamics modeling software;

[0009] S2. Use discrete element simulation software to construct a discrete element model of material particles;

[0010] S3, realize MBD-DEM coupling based on multi-rigid body dynamics model and discrete element model;

[0011] S4. Set different working condition combinations and obtain load history data based on MBD-DEM coupling simulation;

[0012] S5. preprocessing the load history data and corresponding working condition data obtained from the MBD-DEM coupling simulation;

[0013] S6. Build and train LSTM neural network;

[0014] S7. Predict the load history of the loader based on the trained LSTM neural network.

[0015] Optionally, S1 is:

[0016] According to the CAD drawings of the 5t loader, a 3D model was drawn on Soildworks in a simplified manner without affecting the motion process and calculation results. The boom, rocker arm, connecting rod and bucket were precisely drawn and then imported into Adams to define the material properties and mass distribution. Kinematic constraints such as hinge connections and sliding connections and the friction coefficient and damping coefficient of the joints were added to accurately define the connection relationship between the components. The telescopic motion of the hydraulic cylinder was considered to play a driving role. The motion characteristics of the loader working device under real working conditions were simulated by setting the stroke and motion speed curve of the hydraulic cylinder.

[0017] Optionally, S2 is specifically:

[0018] According to the common material types in the actual operation of the loader, the shape and size distribution characteristics of the material particles are determined, the appropriate particle shape is selected on the EDEM software, and the physical properties of the material are defined. For sand and gravel materials, a spherical particle model is used, and for irregular materials, one of the multi-sphere or polyhedron particle models is used, and the particle diameter is set to be randomly distributed within a certain range to simulate the unevenness of the actual material; considering the contact relationship between particles and between particles and the working device, the contact model between particles and between particles and the working device components is defined. The contact model includes a normal contact force model and a tangential contact force model. The contact stiffness, restitution coefficient, static friction coefficient and dynamic friction coefficient are set to accurately simulate the collision, accumulation and sliding behavior of material particles during movement and the interaction with the working device.

[0019] Optional, S3 is:

[0020] The coupling interface is started on the EDEM software, and a data interaction window is established between the MBD model and the DEM model to realize information transmission between the two models. During the coupling simulation process, the motion state of the working device in the multi-body dynamics model is transmitted to the discrete element model in real time, affecting the movement and distribution of material particles; at the same time, the force exerted by the material on the working device in the discrete element model is fed back to the multi-body dynamics model, changing the force state of the working device and determining the time step of the coupling simulation.

[0021] Optionally, S4 is specifically:

[0022] Based on the time series of the boom cylinder, bucket cylinder and steering cylinder as one of the working condition variables, different cylinder movement time series are designed, and each time series includes the start time of the cylinder, the movement speed change curve, and the stop time; considering the particle size factor, different material particle average diameters or diameter ranges are set as one of the working condition variables; the load conditions of the working device when processing large particle materials, medium particle materials and small particle materials are simulated respectively; the working condition is changed according to the particle type, and a variety of materials with different properties are selected for simulation, and the particle contact parameters are changed as the working condition. For the set working condition combination, the MBD-DEM coupling simulation program is run. During the simulation process, the force information of the key parts of the working device is monitored and collected in real time, including the magnitude, direction and action time of the force, to form load history data and save it as a data file.

[0023] Optionally, S5 is specifically:

[0024] The load history data and the corresponding working condition data are cleaned, and the obviously abnormal data outliers and erroneous data are removed by setting the data threshold. Then the data is normalized and all the data are mapped to a specific interval [0,1] to improve the training efficiency and stability of the neural network. The normalization formula uses the linear transformation formula. For a certain data x, the normalized result is:

[0025]

[0026] In the formula, x max and x m i n are the minimum and maximum values ​​of the data in the dataset, respectively.

[0027] Optionally, S6 is specifically:

[0028] Construct an LSTM neural network, including an input layer, multiple hidden layers, and an output layer; divide the preprocessed data set into a training set, a validation set, and a test set, and use the training set to train the LSTM neural network. The root mean square error function RMSE loss function is used to measure the error between the network prediction value and the actual value, that is:

[0029]

[0030] In the formula, n is the number of samples, y is i is the actual value, is the predicted value;

[0031] The Adam optimization algorithm is selected to adjust the weights and biases of the LSTM neural network to minimize the loss function. The validation set is used to evaluate the performance of the network after each round of training. When the validation set loss no longer decreases or reaches the preset training accuracy requirement, the training is stopped and the parameters of the network model with the best performance are saved.

[0032] It can be seen from the above technical solutions that, compared with the prior art, the present invention provides a loader load history prediction method based on MBD-DEM coupled simulation data drive, which has the following beneficial effects:

[0033] 1. Improve the accuracy of load prediction. Through MBD-DEM coupling, the mechanical kinematics and dynamics characteristics of the loader working device and the discrete characteristics of the material can be considered at the same time. Compared with the method of considering the mechanical structure or material effect alone, the actual working conditions of the working device during the working process are more comprehensively restored;

[0034] 2. Better adapt to complex working conditions. The data-driven approach enables the prediction method to learn the load variation rules under various complex working conditions. By collecting and analyzing the data under these different working conditions, the model can effectively adapt to the complex and changing working environment;

[0035] 3. Improve equipment safety and reliability. By accurately predicting the load history, the stress conditions of the working device during operation can be evaluated in advance, and then early warning of possible dangerous situations such as overload can be issued. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0037] Figure 1 It is a flow chart of the method for predicting the load history of a loader of the present invention;

[0038] Figure 2 It is a schematic diagram of the loader dynamics setting of the present invention;

[0039] Figure 3 It is a schematic diagram of the dynamic model of the loader working device of the present invention;

[0040] Figure 4 It is a schematic diagram of the shovel loading pile arrangement of the present invention;

[0041] Figure 5 It is a schematic diagram of the gravel particle model of the present invention;

[0042] Figure 6 This is a schematic diagram of a coal particle model of the present invention;

[0043] Figure 7 A schematic diagram of the coupling arrangement of the present invention;

[0044] Figure 8 Establishing a schematic diagram for the data set of the present invention;

[0045] Fig. 9 It is the LSTM neural network prediction flow chart of the present invention;

[0046] Fig.10 It is a schematic diagram of the prediction results of the present invention;

[0047] Fig.11 It is a schematic diagram of the prediction results of the present invention. DETAILED DESCRIPTION

[0048] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0049] The embodiment of the present invention discloses a method for predicting the load history of a loader based on MBD-DEM coupled simulation data drive, such as Figure 1 As shown, the following steps are included:

[0050] S1. Establish a multi-rigid body dynamics model of the loader working device in the multi-body dynamics modeling software;

[0051] S2. Use discrete element simulation software to construct a discrete element model of material particles;

[0052] S3, realize MBD-DEM coupling based on multi-rigid body dynamics model and discrete element model;

[0053] S4. Set different working condition combinations and obtain load history data based on MBD-DEM coupling simulation;

[0054] S5. preprocessing the load history data and corresponding working condition data obtained from the MBD-DEM coupling simulation;

[0055] S6. Build and train LSTM neural network;

[0056] S7. Predict the load history of the loader based on the trained LSTM neural network.

[0057] Furthermore, S1 is specifically:

[0058] like Figure 2 and Figure 3 As shown in the figure, according to the CAD drawing of the 5t loader, a 3D model is drawn on Soildworks in a simplified manner without affecting the motion process and calculation results. The boom, rocker arm, connecting rod and bucket are precisely drawn and then imported into Adams to define the material properties and mass distribution, add kinematic constraints such as hinge connection and sliding connection, and the friction coefficient and damping coefficient of the joint, accurately define the connection relationship between the components, consider the telescopic movement of the hydraulic cylinder to play a driving role, and set the stroke and motion speed curve of the hydraulic cylinder to simulate the motion characteristics of the loader working device under real working conditions, such as material insertion, material shoveling, boom lifting and unloading reset.

[0059] Furthermore, S2 is specifically:

[0060] like Figure 4 As shown, according to the common material types in the actual operation of the loader (such as sand, coal, crushed stone, etc., see Figure 5-6 ) Determine the shape and size distribution characteristics of the material particles, select the appropriate particle shape (such as sphere, multi-sphere, polyhedron, etc.) on the EDEM software, and define the physical properties of the material. For sand and gravel materials, a spherical particle model is used, and for irregular materials, one of the multi-sphere or polyhedron particle models is used, and the particle diameter is set to be randomly distributed within a certain range to simulate the unevenness of the actual material; considering the contact relationship between particles and between particles and working devices, define the contact model between particles and between particles and working device components. The contact model includes a normal contact force model (such as the Hertz-Mindlin model) and a tangential contact force model (such as the Coulomb friction model). Set the contact stiffness, restitution coefficient, static friction coefficient and dynamic friction coefficient to accurately simulate the collision, accumulation and sliding behavior of material particles in the process of movement and the interaction with the working device.

[0061] Furthermore, S3 is specifically:

[0062] like Figure 7 As shown in the figure, the coupling interface is started on the EDEM software, and a data interaction window is established between the MBD model and the DEM model to realize the information transmission between the two models. During the coupling simulation, the motion state of the working device in the multi-body dynamics model (such as the position, speed, acceleration, etc. of the bucket) is transmitted to the discrete element model in real time, affecting the movement and distribution of material particles; at the same time, the force of the material on the working device in the discrete element model (such as the impact force and friction force of the particles on the bucket) is fed back to the multi-body dynamics model to change the force state of the working device and determine the time step of the coupling simulation.

[0063] In the embodiment of the present invention, the selection of the time step needs to comprehensively consider the stability and calculation accuracy requirements of the multi-body dynamics model and the discrete element model. The time step should be small enough to capture the rapid dynamic changes of the working device and material particles, but too small a time step will increase the amount of calculation.

[0064] Furthermore, S4 is specifically:

[0065] Based on the time series of the boom cylinder, bucket cylinder, and steering cylinder as one of the working condition variables, different cylinder movement time series are designed, each time series includes the start time of the cylinder, the movement speed change curve, and the stop time; considering the particle size factor, different material particle average diameters or diameter ranges are set as one of the working condition variables; the load conditions of the working device when processing large particle materials, medium particle materials, and small particle materials are simulated respectively; the working condition is changed according to the particle type, and a variety of materials with different properties are selected for simulation, such as the coal, gravel, fine sand, etc. mentioned above, each material has different physical parameters; the particle contact parameters are changed as the working condition conditions, which in the embodiment of the present invention include adjusting the contact stiffness, friction coefficient and other parameters between particles and between particles and the working device; for the set working condition combination, the MBD-DEM coupling simulation program is run, and during the simulation process, the force information of the key parts of the working device (such as the boom root, bucket connection point, etc.) is monitored and collected in real time, including the magnitude, direction and action time of the force, to form load history data, and save it as a data file, which is convenient for the basic preparation of the data set construction for the later LSTM neural network training. The saved data is as follows Figure 8 shown.

[0066] In an embodiment of the present invention, the cylinder movement time sequence can be: under fast shoveling conditions, the boom cylinder first extends quickly, and then the bucket cylinder slowly retracts; under slow lifting conditions, the boom cylinder extends at a lower speed and at a uniform speed, and other different time sequences simulate the cylinder movement under different working rhythms; large particle materials are coal with a diameter of 50-100 mm, medium particle materials are gravel with a diameter of 10-50 mm, and small particle materials are fine sand particles with a diameter of 0.1-10 mm.

[0067] Furthermore, S5 is specifically:

[0068] The load history data and the corresponding working condition data are cleaned, and the obviously abnormal data outliers and erroneous data are removed by setting the data threshold. Then the data is normalized and all the data are mapped to a specific interval [0,1] to improve the training efficiency and stability of the neural network. The normalization formula uses the linear transformation formula. For a certain data x, the normalized result is:

[0069]

[0070] In the formula, x max and x m i n are the minimum and maximum values ​​of the data in the dataset, respectively.

[0071] Furthermore, S6 is specifically:

[0072] Construct an LSTM neural network, including an input layer, multiple hidden layers and an output layer; divide the preprocessed data set into a training set, a validation set and a test set, and in the embodiment of the present invention, divide them in a ratio of 70%:15%:15%; use the training set to train the LSTM neural network, and use the root mean square error function RMSE loss function to measure the error between the network prediction value and the actual value, that is:

[0073]

[0074] In the formula, n is the number of samples, y is i is the actual value, is the predicted value;

[0075] The Adam optimization algorithm is selected to adjust the weights and biases of the LSTM neural network to minimize the loss function. The Adam optimization algorithm combines the first-order moment estimation and second-order moment estimation of the gradient, and can adaptively adjust the learning rate to accelerate the convergence of the network. During the training process, the appropriate number of training rounds and learning rate are set. After each round of training, the validation set is used to evaluate the performance of the network. When the validation set loss no longer decreases or reaches the preset training accuracy requirement, the training is stopped and the network model parameters with the best performance are saved.

[0076] In the embodiment of the present invention, the above-mentioned LSTM neural network prediction model is used for load history prediction of unknown working conditions, specifically:

[0077] For unknown working conditions, the data such as the three-cylinder time series, particle size, particle type, and particle contact parameters are obtained and processed in the same preprocessing method as the training data, including data cleaning and normalization.

[0078] 2. Prediction calculation:

[0079] The processed unknown working condition data is input into the trained LSTM neural network, and the network will calculate and output the predicted load history data based on the learned mapping relationship.

[0080] 3. Post-processing of results:

[0081] The predicted load history data is denormalized, and the interval [0,1] is used for normalization. The denormalization formula is:

[0082] x=x'*(x max -x min )+x min

[0083] Among them, x' is the predicted normalized data, and x is the actual data after denormalization. It is restored to the actual physical quantity range to obtain the load prediction results that can be directly used for loader design analysis. The results are visualized and compared with the actual values ​​to intuitively analyze the prediction results. The flow chart is as follows Fig. 9 The results are shown in Fig.10 and Fig.11 As shown, Fig.10 is the y-axis force comparison result, Fig.11 Comparison results of z-axis force.

[0084] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0085] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for predicting load history of a loader based on MBD-DEM coupled simulation data driven, characterized in that: The following steps are involved: S1. Establish a multi-rigid body dynamics model of the loader working device in the multi-body dynamics modeling software; S2. Use discrete element simulation software to construct a discrete element model of material particles; S3, realize MBD-DEM coupling based on multi-rigid body dynamics model and discrete element model; S4. Set different working condition combinations and obtain load history data based on MBD-DEM coupling simulation; S5. preprocessing the load history data and corresponding working condition data obtained from the MBD-DEM coupling simulation; S6. Build and train LSTM neural network; S7. Predict the load history of the loader based on the trained LSTM neural network.

2. The method for predicting load history of a loader based on MBD-DEM coupled simulation data drive according to claim 1, characterized in that: S1 is specifically: According to the CAD drawings of the 5t loader, a 3D model was drawn on Soildworks in a simplified manner without affecting the motion process and calculation results. The boom, rocker arm, connecting rod and bucket were precisely drawn and then imported into Adams to define the material properties and mass distribution. Kinematic constraints such as hinge connections and sliding connections and the friction coefficient and damping coefficient of the joints were added to accurately define the connection relationship between the components. The telescopic motion of the hydraulic cylinder was considered to play a driving role. The motion characteristics of the loader working device under real working conditions were simulated by setting the stroke and motion speed curve of the hydraulic cylinder.

3. The method for predicting load history of a loader based on MBD-DEM coupled simulation data drive according to claim 1, characterized in that: S2 is specifically: According to the common material types in the actual operation of the loader, the shape and size distribution characteristics of the material particles are determined, the appropriate particle shape is selected on the EDEM software, and the physical properties of the material are defined. For sand and gravel materials, a spherical particle model is used, and for irregular materials, one of the multi-sphere or polyhedron particle models is used, and the particle diameter is set to be randomly distributed within a certain range to simulate the unevenness of the actual material; considering the contact relationship between particles and between particles and the working device, the contact model between particles and between particles and the working device components is defined. The contact model includes a normal contact force model and a tangential contact force model. The contact stiffness, restitution coefficient, static friction coefficient and dynamic friction coefficient are set to accurately simulate the collision, accumulation and sliding behavior of material particles during movement and the interaction with the working device.

4. The method for predicting load history of a loader based on MBD-DEM coupled simulation data drive according to claim 1, characterized in that: S3 is specifically: The coupling interface is started on the EDEM software, and a data interaction window is established between the MBD model and the DEM model to realize information transmission between the two models. During the coupling simulation process, the motion state of the working device in the multi-body dynamics model is transmitted to the discrete element model in real time, affecting the movement and distribution of material particles; at the same time, the force exerted by the material on the working device in the discrete element model is fed back to the multi-body dynamics model, changing the force state of the working device and determining the time step of the coupling simulation.

5. The method for predicting loader load history based on MBD-DEM coupled simulation data drive according to claim 1, characterized in that: S4 is specifically: Based on the time series of the boom cylinder, bucket cylinder and steering cylinder as one of the working condition variables, different cylinder movement time series are designed, and each time series includes the start time of the cylinder, the movement speed change curve, and the stop time; considering the particle size factor, different material particle average diameters or diameter ranges are set as one of the working condition variables; the load conditions of the working device when processing large particle materials, medium particle materials and small particle materials are simulated respectively; the working condition is changed according to the particle type, and a variety of materials with different properties are selected for simulation, and the particle contact parameters are changed as the working condition. For the set working condition combination, the MBD-DEM coupling simulation program is run. During the simulation process, the force information of the key parts of the working device is monitored and collected in real time, including the magnitude, direction and action time of the force, to form load history data and save it as a data file.

6. The method for predicting load history of a loader based on MBD-DEM coupled simulation data drive according to claim 1, characterized in that: S5 is as follows: The load history data and the corresponding working condition data are cleaned, and the obviously abnormal data outliers and erroneous data are removed by setting the data threshold. Then the data is normalized and all the data are mapped to a specific interval [0,1] to improve the training efficiency and stability of the neural network. The normalization formula uses the linear transformation formula. For a certain data x, the normalized result is: In the formula, x max and x m i n are the minimum and maximum values ​​of the data in the dataset, respectively.

7. The method for predicting load history of a loader based on MBD-DEM coupled simulation data drive according to claim 1, characterized in that: S6 is specifically: Construct an LSTM neural network, including an input layer, multiple hidden layers, and an output layer; divide the preprocessed data set into a training set, a validation set, and a test set, and use the training set to train the LSTM neural network. The root mean square error function RMSE loss function is used to measure the error between the network prediction value and the actual value, that is: In the formula, n is the number of samples, y is i is the actual value, is the predicted value; The Adam optimization algorithm is selected to adjust the weights and biases of the LSTM neural network to minimize the loss function. The validation set is used to evaluate the performance of the network after each round of training. When the validation set loss no longer decreases or reaches the preset training accuracy requirement, the training is stopped and the parameters of the network model with the best performance are saved.

Images