Skip lightweight design method based on static and dynamic analysis and multi-fidelity proxy model

Abstract

The invention discloses a skip lightweight design method based on static and dynamic analysis and a multi-fidelity agent model. The method comprises the following steps: firstly, performing modal analysis and static analysis on a skip based on finite element simulation; then, discrete elements and finite elements are coupled, and transient mechanical characteristics of the skip bucket are analyzed; secondly, a lightweight design strategy of the skip bucket is designed, specifically, lightweight design comprises two-stage optimization, the first stage is topological optimization, and the second stage is structural size optimization based on a multi-fidelity agent model; finally, the optimal structure parameter combination of the lifting skip bucket is obtained through reinforcement learning solving, and therefore the aim of lightweight design is achieved. According to the method, the weight, the static characteristic and the dynamic characteristic of the skip bucket can be balanced and optimized, and compared with a traditional optimization algorithm, the method has higher optimization efficiency and optimization precision.

Description

Technical Field

[0001] The invention belongs to the technical field of mining products, and in particular relates to a lightweight design method for a skip in a lifting system based on static and dynamic combined analysis. Background Art

[0002] The skip in the hoisting system is the main equipment for mine hoisting. It is widely used in the main shaft vertical hoisting and is mainly responsible for the transportation of coal blocks, workers and equipment. As an important part of the mine hoisting system, the mine skip has a large workload, which causes the hoist to consume a lot of energy. The lightweight design can reduce the weight of the skip, reduce the required power, reduce energy consumption, and improve economic benefits.

[0003] At present, the methods for lightweight design of skips are mostly based on the optimization of structural parameters according to the results of static analysis, ignoring the transient mechanical characteristics of the skip during loading and unloading, resulting in the skip design being difficult to meet the performance requirements during actual operation. Summary of the invention

[0004] Purpose of the invention: In order to solve the problems existing in the existing technologies on the market, the present invention provides a lightweight design method for the bucket in the lifting system based on static and dynamic combined analysis.

[0005] Technical solution: The present invention discloses a lightweight design method for a bucket in a lifting system based on static and dynamic joint analysis, which specifically includes the following steps:

[0006] Step 1: Establish a three-dimensional geometric model of the bucket in the lifting system and perform modal analysis and static analysis on the bucket;

[0007] Step 2: Simulate the material loading and unloading process to obtain the transient mechanical characteristics of the lifting bucket;

[0008] Step 3: Based on static transient analysis, a lightweight strategy is designed. The strategy includes first-stage optimization and second-stage optimization. The first-stage optimization is topology optimization, and the second-stage optimization is structural size optimization.

[0009] Step 4: Build a multi-fidelity proxy model;

[0010] Step 5: Combine the multi-fidelity agent model with the optimization algorithm of reinforcement learning, and solve the optimization objective of the second stage to obtain the optimal design parameters of the bucket.

[0011] Furthermore, the step 1 is specifically as follows:

[0012] S1-1: Use SolidWorks software to build the three-dimensional geometric model of the bucket;

[0013] S1-2: Import the 3D geometric model into ANSYS Workbench software, set the material properties of the bucket, and perform finite element meshing;

[0014] S1-3: Apply the working load of the bucket;

[0015] S1-4: Call the Model command of ANSYS Workbench to perform modal analysis and obtain the natural frequencies and dynamic characteristic indicators of the multi-order modes of the box structure;

[0016] S1-5: Call the Static Analysis module of ANSYS Workbench to perform static analysis on the box to obtain corresponding static characteristic indicators, which include static stress and strain.

[0017] Furthermore, the step 2 is specifically as follows:

[0018] S2-1: Establishment of the discrete element model of the bucket: Save the three-dimensional model of the bucket as Parasolid- * xt format, and import it into EDEM software with the help of EDEM data import interface; set the mesh size, material, particle and contact property parameters and working condition parameters;

[0019] S2-2: Coal block model establishment: select n coal particles with different shapes for 3D model scanning, import the 3D scanned image of the coal particles into EDEM, fill the 3D scanned image with spheres in EDEM, and establish corresponding models of n coal particles with different shapes;

[0020] S2-3: Use a rock mechanics testing machine to conduct petrological experiments on n coal particles with different shapes to obtain the macroscopic physical parameters of the coal particles, then use EDEM software to simulate the mechanical experiment on the coal particle model, continuously adjust the mesoscopic parameter values ​​set in the EDEM software, and compare the difference between the stress-strain curve obtained by the simulation and the stress-strain curve obtained by the rock mechanics test until the difference between the two curves is less than the preset threshold, thereby completing the calibration of the mesoscopic parameters in the EDEM software;

[0021] S2-4: The Hertz-Mindlin model is used to calculate the interaction forces between the coal particle models and between the coal particle models and the boundary surface to describe the dynamic behavior of the coal particle model;

[0022] S2-5: Evenly mix n coal particle models with different shapes to generate filling materials in the bucket, simulate the loading and unloading process of the bucket, and obtain the material distribution characteristics at the hopper mouth, front wall, and side wall;

[0023] S2-6: Based on the distribution characteristics of the materials at the hopper mouth, front wall and side wall, the corresponding load conditions at the hopper mouth, front wall and side wall are calculated, and the load conditions are imported into ANSYS Workbench as boundary conditions. The finite element method is coupled to perform calculations, and the stress field and strain field of the bucket are calculated. The distribution state of dynamic stress and strain is analyzed.

[0024] Furthermore, the first stage of topology optimization selects the natural frequencies of each order of the box to establish the objective function Λ:

[0025]

[0026] Among them, λ i represents the weight coefficient of the i-th order natural frequency, f i represents the i-th order natural frequency;

[0027] By maximizing Λ, we can obtain the structural topological density cloud diagram of the distribution position and distribution form of the skip box material in space;

[0028] The geometric parameter characteristics of the bucket in the design domain after topology optimization are extracted according to the structural topology density cloud map. Based on the geometric parameter characteristics, the parameters of the three-dimensional model established in step 1 are adjusted to obtain the conceptual model of the bucket after the first stage of optimization.

[0029] Furthermore, in the first stage of optimization, the rib structure of the original three-dimensional geometric model is ignored, and an abstract box model is made. The area around the suspension rope in the abstract box model is set as a non-design domain, and the other parts are set as a design domain.

[0030] Furthermore, in the second stage of optimization and construction of the objective function, the performance indicators obtained by static analysis and transient mechanical analysis are first compared. If the performance indicator obtained by transient mechanical analysis is higher than that by static analysis, an optimization objective function with transient performance indicators and improved bucket quality as optimization targets is constructed; otherwise, an optimization objective function with static performance indicators and improved bucket quality as optimization targets is constructed.

[0031] Furthermore, the second stage optimization selects the parameters to be optimized as optimization variables with the goal of lightweighting the bucket mass, and constructs the following objective function:

[0032] minM(x)

[0033] stσ(x)≤σ r (x)≤[σ(x)]

[0034] x i-min ≤x i ≤x i-max

[0035] Among them, minM(x) is the objective function, M(.) is the function for calculating the bucket quality, σ r (x) represents the ultimate stress of the bucket structure, [σ(x)] is the allowable safety stress of the bucket structure, and σ(x) represents the stress of the bucket structure; x i-min Represents the optimization variable x i The minimum value of x i-max Represents the optimization variable x i The maximum value; x is the optimization variable, x={x1,x2,…,x i ,…,x M}, M represents the total number of optimization variables.

[0036] Furthermore, the step 4 is specifically as follows:

[0037] Step 4.1: Use Latin hypercube sampling method in [x i-min , x i-max ] for x i Sampling is performed, wherein the number of sampling points of the high-fidelity data sample is H, and the number of sampling points of the low-fidelity data sample is L;

[0038] Step 4.2: Use a high-density mesh finite element model to perform response analysis on the high-fidelity data sample to obtain a high-fidelity stress response value represents the high-fidelity stress corresponding to the Hth high-fidelity data sample;

[0039] Step 4.3: Use the low-density mesh finite element model to perform response analysis on the low-fidelity data sample to obtain the low-fidelity stress response value represents the low-fidelity stress corresponding to the Lth low-fidelity data sample;

[0040] Step 4.4: Build a low-fidelity proxy model f based on Kriging LF and a high-fidelity proxy model based on Kriging f HF ;

[0041] Step 4.5: Construct a multi-fidelity proxy model f based on Co-Kriging MF :

[0042]

[0043] in, is a low-fidelity data sample, ρ is a scaling factor, is a constant, f δ The deviation response model.

[0044] Furthermore, a Gaussian random process is used to represent the low-fidelity model f LF and the deviation response model f δ :

[0045] f LF =ψ LF +M LF

[0046] f δ =ψ δ +M δ

[0047] Among them, ψ LF represents the expectation of the low-fidelity model Gaussian random process, ψ δ Denotes the deviation response model f δ The expectation of a Gaussian random process, M LF represents the variance of the low-fidelity model Gaussian random process, ψ δ Represents the variance of the Gaussian random process that models the deviated response.

[0048] Furthermore, the step 5 is specifically as follows:

[0049] Integrate the multi-fidelity agent model with the reinforcement learning algorithm, build an optimization environment, and set the reward function R:

[0050] R=-W-α·[max(0,σ max -σ allow )+max(0,δ max -δ allow )]

[0051] Where W is the total mass of the bucket, σ max is the maximum stress calculated by simulation, σ allow is the material yield stress, δ max is the maximum strain calculated by simulation, δ allow is the maximum allowable strain;

[0052] In the iterative optimization process, the PPO algorithm is used, the output of the multi-fidelity model is used as the input of the intelligent agent, and the optimization variable is used as the output of the intelligent agent. The output of the intelligent agent is evaluated according to the reward function. When the reward function value or the adjustment range of the optimization variable converges, the training ends and the optimal design parameters are obtained.

[0053] Beneficial effects:

[0054] (1) The present invention is based on the combination of discrete element simulation and finite element model analysis to simulate the actual working state of the bucket. Different from a single static analysis, the discrete cloud-finite element coupling simulation analysis involved in the present invention is more comprehensive and more in line with the actual working state of the bucket, providing a basis for optimizing the working performance of the bucket.

[0055] (2) The present invention performs lightweight design of the bucket structure based on static and dynamic joint analysis, and includes two stages: topology optimization and structural parameter optimization, so that the weight, static characteristics, and dynamic characteristics of the bucket can be optimally balanced.

[0056] (3) The present invention designs a multi-fidelity proxy model that combines additive scaling and Kriging proxy models, and combines it with reinforcement learning for parameter optimization. Compared with traditional optimization algorithms, it has higher optimization efficiency and optimization accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 It is a flow chart of the present invention.

[0058] Figure 2 A process for generating a skip conceptual model for the first topology optimization of the present invention;

[0059] Figure 3 This is a flowchart of the second structural parameter optimization based on the multi-fidelity proxy model of the present invention. DETAILED DESCRIPTION

[0060] The accompanying drawings, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0061] like Figure 1 As shown, the present invention discloses a lightweight design method for a bucket in a lifting system based on static and dynamic combined analysis, specifically:

[0062] S1: Modal analysis and static analysis of bucket in lifting system based on finite element simulation:

[0063] S1-1: Use SolidWorks software to build the three-dimensional geometric model of the bucket.

[0064] S1-2: Import the 3D geometric model into ANSYS Workbench software, set the material properties of the bucket to Q355 and perform finite element meshing.

[0065] S1-3: Apply the working load of the bucket. In this embodiment, the working load of the bucket is set to 45t.

[0066] S1-4: Call the Model command of ANSYS Workbench to perform modal analysis to obtain dynamic characteristic indicators such as the natural frequencies of the multi-order modes of the box structure and the corresponding vibration modes.

[0067] S1-5: Call the Static Analysis module of ANSYS Workbench to perform static analysis on the box body based on the established finite element model to obtain the corresponding static characteristic indicators such as static stress and strain.

[0068] S2: Transient mechanical characteristics analysis based on discrete element-finite element coupling

[0069] In order to explore the interaction mechanism between materials and bucket, the material loading and unloading processes were simulated based on the discrete element simulation software EDEM; the distribution characteristics of materials at the hopper mouth, front wall and side wall were statistically analyzed.

[0070] S2-1: The discrete element model of the bucket is established, and the established three-dimensional model of the bucket is saved as Parasolid ( * xt) format, and import it into EDEM software with the help of EDEM data import interface; set the mesh size, material, particle and contact property parameters and working condition parameters.

[0071] S2-2: Coal block model establishment: 6 coal particles with different shapes were selected for 3D model scanning, and the 3D scanned image of the coal particles was imported into EDEM. In EDEM, spheres were used to fill the 3D scanned image, and 6 special-shaped coal particle models with different shapes and specifications were established respectively.

[0072] S2-3: Petrological experiments were conducted on coal particles of six different shapes using a laboratory multifunctional rock mechanics testing machine to obtain the macroscopic physical parameters of the coal particles, including density, elastic modulus, shear modulus, Poisson's ratio, cohesion, internal friction angle, tensile strength, and the ratio of tangential and normal stiffness of joints.

[0073] Then, the EDEM software is used to simulate the mechanical experiment of the coal particle model, and the mesoscopic parameter values ​​set in the EDEM software are continuously adjusted. The difference between the stress-strain curve obtained by simulation and the stress-strain curve obtained by rock mechanics test is compared until the difference between the two curves is less than the preset threshold, thereby completing the calibration of the mesoscopic parameters in the EDEM software.

[0074]

[0075]

[0076]

[0077]

[0078] c=τ-σtan(φ)

[0079] Among them, σ nis the compressive strength, MPa, F is the maximum compressive force on the coal particles, and A is the original cross-sectional area of ​​the coal particles, mm 2 , E is the elastic modulus, σ b , σ a is the stress at the end and starting point of the straight line segment in the stress-strain curve, MPa; ε b , ε a is the strain at the end and start of the straight line segment in the stress-strain curve; μ is Poisson's ratio, ε dp , ε lp is the strain at the end and starting point of the straight line segment in the stress-strain curve; σ t is the maximum tensile stress at the center of the specimen, i.e., tensile strength, MPa; p is the ultimate pressure at test failure, N; d and t are the diameter and thickness of the pressure disk, mm; c is the cohesion; φ is the internal friction angle; σ is the normal stress; and τ is the shear stress.

[0080] S2-4: The Hertz-Mindlin (no slip) model is used as the contact model between coal particle models to calculate the interaction forces between coal particle models and between coal particle models and boundary surfaces to describe the dynamic behavior of coal block particles.

[0081] S2-5: Six coal particle models with different shapes are evenly mixed to generate filling materials in the bucket. The mass of the filled material is 45t. The loading and unloading processes of the bucket are simulated to obtain the material distribution characteristics at the hopper mouth, front wall and side wall.

[0082] S2-6: Based on the distribution characteristics of the materials at the hopper mouth, front wall and side wall, the corresponding load conditions at the hopper mouth, front wall and side wall are calculated, and the load conditions are imported into ANSYS Workbench as boundary conditions. The finite element method is coupled to perform calculations, and the stress field and strain field of the bucket are calculated, and the distribution state of stress and strain is analyzed.

[0083] S3: Lightweight design strategy for coupling static and dynamic analysis

[0084] The lightweight design includes two-stage optimization. The first stage is topology optimization to obtain the conceptual model of the bucket. The second stage is structural size optimization based on a multi-fidelity proxy model.

[0085] The first topology optimization selects each order of natural frequency as the optimization objective function, and then optimizes to obtain a rough conceptual model;

[0086] During the second stage of structural size optimization, the performance indicators such as stress and strain obtained from static analysis and transient mechanical analysis are first compared. If the dynamic characteristic indicators obtained from transient analysis are higher than the static characteristic indicators obtained from static analysis, an optimization objective function with dynamic performance indicators and improved bucket quality as optimization targets is constructed. Otherwise, an optimization objective function with static performance indicators and improved bucket quality as optimization targets is constructed, and then the second stage of structural size optimization is carried out.

[0087] S4: Generation of the bucket conceptual model based on the first topology optimization

[0088] like Figure 2 As shown in the figure, the first topology optimization is as follows:

[0089] S4-1: The first optimization setting uses the Topology Optimization module of ANSYS Workbench to optimize the natural frequency of the box. To avoid falling into the local optimum, the ribs and other structures of the original three-dimensional geometric model in step S1-1 are ignored. An abstract model of the box is made, and the area near the suspension rope is set as the non-design domain, and the other parts are set as the design domain. The objective function is defined as the weighted objective function of the first 6 natural frequencies:

[0090]

[0091] Λ is the objective function, λ i is the weight coefficient of the i-th order natural frequency, f i is the i-th order natural frequency, in this embodiment, λ1=0.5, λ2=0.3, λ3=0.2, λ4=λ5=λ6=0.05, and the constraint condition of the topology optimization this time is that the retained volume fraction is not higher than 25%.

[0092] S4-2: Maximize the objective function in ANSYS Workbench to obtain the structural topological density cloud map of the distribution position and distribution form of the skip box material in space.

[0093] S4-3: Extract the geometric parameter characteristics of the bucket in the design domain after topology optimization based on the structural topology density cloud map, and modify the three-dimensional geometric model in SolidWorks to obtain the conceptual model of the bucket after the first topology optimization.

[0094] S5: Second structural parameter optimization based on multi-fidelity surrogate model

[0095] like Figure 3 As shown in the figure, the second structural parameter optimization is as follows:

[0096] S5-1: Determine the optimization variables and ranges. According to the static and dynamic analysis structure, determine the bucket wall x1, the cross-sectional area of ​​the outer reinforcement x2, the spacing of the reinforcing ribs x3, the thickness of the reinforcing ribs x4, the bucket depth x5, the bucket width x6, and the bucket side panel length x7 as design variables. Define the design variable range x1∈[10,15], unit: mm; x2∈[6.4,7.7], unit: m 2 ; x3∈[100,150], unit: mm; x4∈[15,25], unit: mm; x5∈[3.5,4], unit: m; x6∈[1.5,2], unit: m; x7∈[0.3,0.5], unit: m; define the optimization variables x=(x1,x2,x3,x4,x5,x6,x7);

[0097] S5-2 determines the objective function of the optimization model. The optimization goal is to make the bucket structure the lightest under the premise of satisfying the ultimate stress and strain of the bucket box. The mathematical description is as follows:

[0098]

[0099] In the formula, M(x), σ r (x) are the mass and ultimate stress of the bucket structure, σ(x) is the stress of the bucket, [σ(x)] is the allowable safety stress of the bucket structure, and x i-max 、x i-min is the i-th optimization variable x in x=(x1,x2,x3,x4,x5,x6,x7) i The maximum and minimum values ​​of .

[0100] S5-3: Using the Latin hypercube sampling method, in [x i-min ,x i-max ] for x i Data sampling is performed, wherein the number of sampling points of high-fidelity data samples is H, and the number of sampling points of low-fidelity data samples is L.

[0101] S5-4: Use a high-density mesh finite element model to perform response analysis on high-fidelity data samples to obtain high-fidelity stress response values represents the high-fidelity stress corresponding to the Hth high-fidelity data sample.

[0102] S5-5: Use a low-density mesh finite element model to perform response analysis on low-fidelity data samples and obtain low-fidelity stress response values represents the low-fidelity stress corresponding to the Lth low-fidelity data sample.

[0103] S5-6: Establish a low-fidelity proxy model based on Kriging LFand a high-fidelity proxy model based on Kriging f HF ;f LF The expression is:

[0104]

[0105] is a low-fidelity data sample, For The corresponding low-fidelity stress.

[0106] S5-7: Constructing a multi-fidelity proxy model based on Co-Kriging MF :

[0107]

[0108] Where ρ is the scaling factor, is a constant, f δ The deviation response model.

[0109] A Gaussian random process is used to represent the low-fidelity model f LF and the deviation response model f δ :

[0110] f LF =ψ LF +M LF

[0111] f δ =ψ δ +M δ

[0112] Among them, ψ LF represents the expectation of the low-fidelity model Gaussian random process, ψ δ Denotes the deviation response model f δ The expectation of a Gaussian random process, M LF represents the variance of the low-fidelity model Gaussian random process, ψ δ Represents the variance of the Gaussian random process that models the deviated response.

[0113] The bucket structure parameters are optimized based on the Co-Kriging multi-fidelity proxy model and the reinforcement learning optimization algorithm. This embodiment uses the reinforcement learning optimization algorithm to achieve bucket parameter optimization and obtain the optimal design parameters of the bucket to achieve the lightweight design goal.

[0114] S6-1: Taking the lightweight of the bucket as the optimization goal and meeting its static characteristics requirements, the bucket wall, the cross-sectional area of ​​the outer reinforcement, the spacing of the reinforcing ribs, the thickness of the reinforcing ribs, the bucket depth, the bucket width, and the bucket side panel length are selected as key optimization parameters, and constraints are set.

[0115] S6-2: Integrate the multi-fidelity model with the reinforcement learning algorithm and build an optimization environment, where the state is the output response of the multi-fidelity agent model, the action is to adjust the parameters, and the reward function R is set to:

[0116] R=-W-α·[max(0,σ max -σ allow )+max(0,δ max -δ allow )]

[0117] Where W is the total mass of the bucket, σ max is the maximum stress calculated by simulation, σ allow is the material yield stress, δ max is the maximum strain calculated by simulation, δ allow is the maximum allowable strain and α is the coefficient.

[0118] S6-3: In this optimization process, the PPO (Proximal Policy Optimization) algorithm is used as the core of reinforcement learning. In each iteration, the agent's input is the static performance indicators obtained through multi-fidelity model simulation and the calculated reward value. The agent generates corresponding actions based on these inputs, that is, outputs the optimized bucket geometry parameters. After each action is executed, the current reward value is calculated according to the preset reward function, and the quality of the parameter configuration is evaluated. The higher the reward value, the closer the current design is to the optimization goal. The agent updates the strategy based on the current reward signal.

[0119] It should also be noted that the various specific technical features described in the above specific embodiments can be combined in any suitable manner without contradiction. To avoid unnecessary repetition, the present invention will not further describe various possible combinations.

Claims

1. A lightweight design method for bucket based on static and dynamic analysis and multi-fidelity proxy model, characterized in that: The specific steps include: Step 1: Establish a three-dimensional geometric model of the bucket in the lifting system and perform modal analysis and static analysis on the bucket; Step 2: Simulate the material loading and unloading process to obtain the transient mechanical characteristics of the lifting bucket; Step 3: Based on static transient analysis, a lightweight strategy is designed. The strategy includes first-stage optimization and second-stage optimization. The first-stage optimization is topology optimization, and the second-stage optimization is structural size optimization. Step 4: Build a multi-fidelity proxy model; Step 5: Combine the multi-fidelity agent model with the optimization algorithm of reinforcement learning, and solve the optimization objective of the second stage to obtain the optimal design parameters of the bucket.

2. The lightweight design method for bucket based on static and dynamic analysis and multi-fidelity proxy model according to claim 1 is characterized in that: The step 1 is specifically as follows: S1-1: Use SolidWorks software to build the three-dimensional geometric model of the bucket; S1-2: Import the 3D geometric model into ANSYS Workbench software, set the material properties of the bucket, and perform finite element meshing; S1-3: Apply the working load of the bucket; S1-4: Call the Model command of ANSYS Workbench to perform modal analysis and obtain the natural frequencies and dynamic characteristic indicators of the multi-order modes of the box structure; S1-5: Call the Static Analysis module of ANSYS Workbench to perform static analysis on the box to obtain corresponding static characteristic indicators, which include static stress and strain.

3. The lightweight design method for bucket based on static and dynamic analysis and multi-fidelity proxy model according to claim 1 is characterized in that: The step 2 is specifically as follows: S2-1: Establishment of the discrete element model of the bucket: Save the three-dimensional model of the bucket as Parasolid- * xt format, and import it into EDEM software with the help of EDEM data import interface; set the mesh size, material, particle and contact property parameters and working condition parameters; S2-2: Coal block model establishment: select n coal particles with different shapes for 3D model scanning, import the 3D scanned image of the coal particles into EDEM, fill the 3D scanned image with spheres in EDEM, and establish corresponding models of n coal particles with different shapes; S2-3: Use a rock mechanics testing machine to conduct petrological experiments on n coal particles with different shapes to obtain the macroscopic physical parameters of the coal particles, then use EDEM software to simulate the mechanical experiment on the coal particle model, continuously adjust the mesoscopic parameter values ​​set in the EDEM software, and compare the difference between the stress-strain curve obtained by the simulation and the stress-strain curve obtained by the rock mechanics test until the difference between the two curves is less than the preset threshold, thereby completing the calibration of the mesoscopic parameters in the EDEM software; S2-4: The Hertz-Mindlin model is used to calculate the interaction forces between the coal particle models and between the coal particle models and the boundary surface to describe the dynamic behavior of the coal particle model; S2-5: Evenly mix n coal particle models with different shapes to generate filling materials in the bucket, simulate the loading and unloading process of the bucket, and obtain the material distribution characteristics at the hopper mouth, front wall, and side wall; S2-6: Based on the distribution characteristics of the materials at the hopper mouth, front wall and side wall, the corresponding load conditions at the hopper mouth, front wall and side wall are calculated, and the load conditions are imported into ANSYS Workbench as boundary conditions. The finite element method is coupled to perform calculations, and the stress field and strain field of the bucket are calculated. The distribution state of dynamic stress and strain is analyzed.

4. The lightweight design method for bucket based on static and dynamic analysis and multi-fidelity proxy model according to claim 1 is characterized in that: The first stage of topology optimization selects the natural frequencies of the box at each order to establish the objective function Λ: Among them, λ i represents the weight coefficient of the i-th order natural frequency, f i represents the i-th order natural frequency; By maximizing Λ, we can obtain the structural topological density cloud diagram of the distribution position and distribution form of the skip box material in space; The geometric parameter characteristics of the bucket in the design domain after topology optimization are extracted according to the structural topology density cloud map. Based on the geometric parameter characteristics, the parameters of the three-dimensional model established in step 1 are adjusted to obtain the conceptual model of the bucket after the first stage of optimization.

5. The lightweight design method for bucket based on static and dynamic analysis and multi-fidelity proxy model according to claim 1 is characterized in that: In the first stage of optimization, the rib structure of the original three-dimensional geometric model is ignored, and an abstract box model is made. The area around the suspension rope in the abstract box model is set as the non-design domain, and the other parts are set as the design domain.

6. The lightweight design method for bucket based on static and dynamic analysis and multi-fidelity proxy model according to claim 1 is characterized in that: In the second stage of optimization and construction of the objective function, the performance indicators obtained by static analysis and transient mechanical analysis are first compared. If the performance indicators obtained by transient mechanical analysis are higher than those by static analysis, an optimization objective function with transient performance indicators and improved bucket quality as optimization targets is constructed. Otherwise, an optimization objective function with static performance indicators and improved bucket quality as optimization targets is constructed.

7. The lightweight design method for bucket based on static and dynamic analysis and multi-fidelity proxy model according to claim 1 is characterized in that: The second stage of optimization is to select the parameters that need to be optimized as optimization variables with the goal of lightweighting the bucket mass, and construct the following objective function: minM(x) stσ(x)≤σ r (x)≤[σ(x)] x i-min ≤x i ≤x i-max Among them, minM(x) is the objective function, M(.) is the function for calculating the bucket quality, σ r (x) represents the ultimate stress of the bucket structure, [σ(x)] is the allowable safety stress of the bucket structure, and σ(x) represents the stress of the bucket structure; x i-min Represents the optimization variable x i The minimum value of x i-max Represents the optimization variable x i The maximum value; x is the optimization variable, x={x1,x2,…,x i ,…,x M }, M represents the total number of optimization variables.

8. The lightweight design method for bucket based on static and dynamic analysis and multi-fidelity proxy model according to claim 7 is characterized in that: The step 4 is specifically as follows: Step 4.1: Use Latin hypercube sampling method in [x i-min , x i-max ] for x i Sampling is performed, wherein the number of sampling points of the high-fidelity data sample is H, and the number of sampling points of the low-fidelity data sample is L; Step 4.2: Use the finite element model with high-density mesh to perform response analysis on the high-fidelity data sample to obtain the high-fidelity stress response value σ HF , represents the high-fidelity stress corresponding to the Hth high-fidelity data sample; Step 4.3: Use the low-density mesh finite element model to perform response analysis on the low-fidelity data sample to obtain the low-fidelity stress response value σ LF , represents the low-fidelity stress corresponding to the Lth low-fidelity data sample; Step 4.4: Build a low-fidelity proxy model f based on Kriging LF and a high-fidelity proxy model based on Kriging f HF ; Step 4.5: Construct a multi-fidelity proxy model f based on Co-Kriging MF : in, is a low-fidelity data sample, ρ is a scaling factor, is a constant, f δ The deviation response model.

9. The lightweight design method for bucket based on static and dynamic analysis and multi-fidelity proxy model according to claim 8, characterized in that: A Gaussian random process is used to represent the low-fidelity model f LF and the deviation response model f δ : f LF =ψ LF +M LF f δ =ψ δ +M δ Among them, ψ LF represents the expectation of the low-fidelity model Gaussian random process, ψ δ Denotes the deviation response model f δ The expectation of a Gaussian random process, M LF represents the variance of the low-fidelity model Gaussian random process, ψ δ Represents the variance of the Gaussian random process that models the deviated response.

10. The lightweight design method for bucket based on static and dynamic analysis and multi-fidelity proxy model according to claim 1, characterized in that: The step 5 is specifically as follows: Integrate the multi-fidelity agent model with the reinforcement learning algorithm, build an optimization environment, and set the reward function R: R=-W-α·[max(0,σ max -s allow )+max(0,δ max -d allow )] Where W is the total mass of the bucket, σ max is the maximum stress calculated by simulation, σ allow is the material yield stress, δ max is the maximum strain calculated by simulation, δ allow is the maximum allowable strain, α is the coefficient; In the iterative optimization process, the PPO algorithm is used, the output of the multi-fidelity model is used as the input of the intelligent agent, and the optimization variable is used as the output of the intelligent agent. The output of the intelligent agent is evaluated according to the reward function. When the reward function value or the adjustment range of the optimization variable converges, the training ends and the optimal design parameters are obtained.

Images