A structure-control integrated optimization method for unmanned mining excavator
Through the Latin hypercube sampling and Kriging proxy model combined with the expected super-volume improvement method of the NSGA-III algorithm, the problem of structure and control target heterogeneity in structure-control optimization of unmanned mining excavator is solved, and better optimization is achieved As a result, the performance of the excavator is improved.
Patent Information
- Application Number
- CN202410465455.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-04-18
- Publication Date
- 2024-11-08
- Estimated Expiration
- 2044-04-18
AI Technical Summary
The structure-control optimization method of existing unmanned mining excavators ignores the relationship between structure and control, resulting in suboptimal optimization results, and the heterogeneous properties between the structure and the control target cannot be effectively handled, and the excavation cannot be effectively improved. performance of the machine.
The Latin hypercube sampling method is used to generate the initial sample point set, and the Kriging agent model is constructed. Combined with the NSGA-III algorithm and the expected super-volume improvement method, iteratively optimizes the structure-control integrated optimization problem of the unmanned mining excavator, and uses low-cost The difference between objective function and expensive objective function, obtaining the optimal Pareto solution through multiple iterations and evaluations.
The heterogeneous properties between the structure and the control target are effectively dealt with, the mechanism and control performance of the unmanned mining excavator are improved, and better optimization results are obtained.
Smart Images

Figure CN118378518B_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the technical field of optimization design and relates to a structure-control integrated optimization method for an unmanned mining excavator. Background Art
[0002] Mining excavators are the core and source equipment for open-pit mining. They integrate the two major functions of "excavation" and "loading". Their performance determines the production efficiency of the entire mine. The unmanned and intelligent development of mining excavators is of great significance to improving the economic benefits of the mining industry, improving the working environment of employees, and protecting the life and health of personnel. The optimization design of unmanned mining excavators includes two important aspects: structural optimization and control optimization. The existing optimization methods are mostly serial in solving the optimization process involving structure and control system, that is, the structure is optimized first, and then the control optimization of excavation is carried out on the basis of the optimized structure. Then, due to the neglect of the relationship between structure and control, the optimization results are often suboptimal, which makes the performance prompt of the equipment insufficient. In addition, some related studies comprehensively consider the parameters related to structure and control with the optimization objectives, and optimize the unmanned mining excavator in an integrated manner. In this type of method, the structural objectives and control objectives of the unmanned mining excavator are regarded as isomorphic, that is, they both need to obtain results through computationally expensive simulations. Therefore, when solving this type of problem, the existing methods need to use a proxy model basis to fit and replace all objective functions and constraints. However, the control performance parameters of a real unmanned mining excavator are not computationally expensive, so there is no need to fit them through an expensive proxy model. Therefore, the structure-control integrated optimization problem of an unmanned mining excavator is a type of heterogeneous problem, and existing methods cannot solve this type of problem. The present invention aims at the structure-control integrated optimization problem of a heterogeneous unmanned mining excavator and proposes a structure-control integrated optimization method for an unmanned mining excavator. Summary of the invention
[0003] In view of the structure-control integrated optimization problem of heterogeneous unmanned mining excavators in the prior art, the present invention provides a structure-control integrated optimization method for unmanned mining excavators, which is used to perform integrated optimization of unmanned mining excavators to improve the performance of the excavators.
[0004] In order to achieve the above object, the technical solution adopted by the present invention is:
[0005] A structure-control integrated optimization method for an unmanned mining excavator comprises the following steps:
[0006] Step 1: First, determine the parameter type, quantity and sampling range of the structure-control integrated optimization problem of the unmanned mining excavator. The parameter types include the structural parameters of the excavation components and the control parameters of the excavation trajectory.
[0007] Then use the Latin hypercube sampling method to generate N s The input sample point set X of the initial sample is then evaluated. The expensive excavator boom mass structure objective function and the excavator boom stress constraint conditions are obtained through simulation. The inexpensive excavation energy consumption and excavation volume control objective functions can be directly calculated through explicit functions. By substituting the input sample point set X into the objective function and constraint conditions, the output response Y is obtained. The sample point set X is used as the initial population P.
[0008] Step 2: For the expensive excavator boom mass structure objective function and excavator boom stress constraint conditions in the excavator optimization problem, a Kriging proxy model is constructed through the sample set (X, Y). For the inexpensive excavation energy consumption and excavation volume control objective functions in the excavator optimization problem, no proxy model needs to be constructed, and the current Pareto frontier is calculated.
[0009] Step 3: Use the population iteration method in the NSGA-III algorithm to iterate the initial population P of the excavator optimization problem. max times, and obtain the population after iteration
[0010] Step 4: Get the population The expected hypervolume improvement calculation is performed on the points in the design space using a nearest neighbor approximation method. First, a sample size of N is E Uniform sampling of , and calculating the dominance relationship between all samples, and then obtaining the hypervolume improvement of each sample relative to the current Pareto frontier. The Kriging model predicts each point in as a high-dimensional normal distribution. The high-dimensional normal distribution of each point is calculated with a sample size of N. normal Sampling is performed, and for the jth normal distribution sample, the nearest uniform distribution sample is expressed as:
[0011]
[0012] Among them, ‖‖2 represents the 2-norm, p j is the jth normally distributed sample, q is a uniformly distributed sample point, q j is the distance p in the uniformly distributed sample j The nearest point. j Hypervolume Improvement HVI(p j ) is expressed as:
[0013] HVI(p j )=HVI(q j )#(2)
[0014] By taking the average of all normally distributed sample points, the expected hypervolume improvement EHVI(x) is obtained as:
[0015]
[0016] Finally, we get the population The expected hypervolume improvement for all samples in .
[0017] Step 5: Find the population The individual with the largest expected hypervolume is expressed as:
[0018]
[0019] Evaluate the new point x using the expensive excavator boom mass structure objective function and excavator boom stress constraints and the cheap excavation energy consumption and excavation volume control objective functions new , and get its response y new . new ,y new ) is added to the training sample set (X, Y), and a new round of initial population is obtained through the population selection method in the NSGA-III algorithm. Steps 2 to 5 are repeated until the total number of evaluations of the structure-control integrated optimization problem of the unmanned mining excavator reaches N. max . From the number N max The optimal Pareto solution is obtained from the sample set (X, Y) as the solution to the structure-control integrated optimization problem of the unmanned mining excavator.
[0020] The effective gain effect of the present invention is: the present invention proposes an integrated optimization method that considers the heterogeneous properties between the objectives and constraints of the structural system and the objectives and constraints of the control system in the structure-control integrated optimization of the unmanned mining excavator. Compared with the optimization method that considers the structure and control of the unmanned mining excavator as isomorphic, the method of the present invention can handle the differences between heterogeneous objective functions well, effectively utilize the uncertainty of the objective prediction based on the proxy model and the certainty of the cheap objective function, and can obtain better optimization results, effectively improving the structure and control performance of the unmanned mining excavator. BRIEF DESCRIPTION OF THE DRAWINGS
[0021] Figure 1 This is a flow chart of the structure-control integrated optimization method of the present invention.
[0022] Figure 2 This is a schematic diagram of the structural modeling of an unmanned mining excavator according to an embodiment of the present invention; in the figure, 1 is a boom, 2 is an arm, and 3 is a bucket.
[0023] Figure 3 It is a schematic diagram of finite element analysis of the boom of an unmanned mining excavator according to an embodiment of the present invention.
[0024] Figure 4 This is a schematic diagram of modeling a control system for an unmanned mining excavator according to an embodiment of the present invention.
[0025] Figure 5 This is the Pareto front result of the embodiment of the present invention. DETAILED DESCRIPTION
[0026] The present invention is further described below in conjunction with specific embodiments.
[0027] like Figure 1 As shown, the present invention provides a structure-control integrated optimization method for an unmanned mining excavator. First, the scope, objectives and constraints of the structure-control integrated optimization problem of the unmanned mining excavator are obtained, and then the initial sample set is obtained by Latin hypercube sampling, and the response of the sample set is evaluated. Then, a Kriging proxy model is established for the computationally expensive mechanism objective function and constraints. The population is subjected to multiple evolutionary iterations, and the expected hypervolume improvement value is calculated for the points in the population obtained after the iteration. The point with the largest expected hypervolume in the population is obtained, evaluated and added to the training set, and a new population is selected for the next cycle iteration through the environmental selection method until the sample size condition is met.
[0028] Figure 2 Schematic diagram of modeling of the structure system of the unmanned mining excavator in the embodiment. The rope connection between the bucket 3 and the boom 1 is simplified by a spring, a fixed support is applied to the end of the boom 1, and a rotating support is applied to the bottom of the boom 1. In addition to gravity, the equivalent force of the excavated material and the excavation resistance are applied to the bucket 3, totaling 300,000N. Figure 3 The structural analysis results of the boom of an unmanned mining excavator obtained using finite element analysis software. The upper figure is the meshing result, and the lower figure is the stress distribution result. Figure 4 This is a schematic diagram of modeling a control system for an unmanned mining excavator according to an embodiment of the present invention, wherein the excavation trajectory is represented by a polynomial.
[0029] The three structural parameters of the embodiment of the present invention are as follows: Figure 4 As shown, the ranges are [6.5, 8], [7.8, 8.3] and [40, 50] respectively. In addition, there are 5 control parameters, which are the x-direction position of the bucket tooth tip and the final position of the excavation trajectory. and in the y direction The sixth polynomial coefficient and And the final mining time t fThe structural optimization target is the mass of the boom. The structural constraint function needs to ensure that its maximum stress is less than the allowable stress of the structure. The boom material is Q235 steel, the yield strength is 235MPa, and the safety factor is set to 2. The structural constraint is expressed as:
[0030] g1(x)=max_stress-235 / 2#(6)
[0031] The objective function of the control system is the energy consumption and excavation volume under the excavation trajectory, expressed as:
[0032] f2(x)=E total #(7)
[0033] and
[0034] f3(x)=-V p #(8)
[0035] The structure-control integration optimization process of the unmanned mining excavator according to the embodiment of the present invention is as follows:
[0036] First, all design variables are sampled using the Latin hypercube sampling method. The dimension of the structural variables is 3 and the dimension of the control variables is 5. Therefore, the sampling dimension m is set to 8 and the initial sample size N is set to s After obtaining the sample X, the sample point is evaluated, and the expensive unmanned mining excavator structure objective function and constraint conditions are calculated by Figure 3 The finite element simulation shown in the figure obtains that the computationally inexpensive control objective function and constraints can be directly calculated by explicit functions to obtain the output Y of sample X. The initial sample set is taken as the initial population P.
[0037] After that, for the computationally expensive structural objective functions and constraints, a Kriging proxy model is constructed using the sample set (X, Y). The current Pareto front is calculated. Set the maximum number of iterations t max The population iteration method in the NSGA-III algorithm is used to iterate the initial population 10 times to obtain the iterated population P 10 .
[0038] Then, for the obtained population P 10 The expected hypervolume improvement calculation is performed on the points in the graph, and the uniform sample size N is set. E The sample size N of the normal distribution is 10000. normal Set to 500. Calculate the population P 10 The expected hypervolume improvement for all points in .
[0039] Finally: Find the population P 10 The individual x with the largest expected hypervolume new , evaluate the expensive function at new points xnew , and get its response y new . new ,y new ) is added to the training sample set (X, Y), and the new round of initial population is obtained through the population selection method of NSGA-III. Steps 2 to 5 are repeated, and the number of expensive evaluations N is set. max The optimization stops when the number of expensive samples is equal to or exceeds 50. After that, the optimization results are sorted non-dominated to obtain the sample distribution with the highest ranking. Figure 5 This is the Pareto front result of the embodiment of the present invention. It can be seen that all sample distributions form a Pareto surface, and the result is reasonable.
Claims
1. A structure-control integrated optimization method for unmanned mining excavators, It is characterized in that The following steps are involved: Step 1: First, determine the parameter type, quantity and sampling range of the structure-control integrated optimization problem of the unmanned mining excavator. The parameter types include the structural parameters of the excavation components and the control parameters of the excavation trajectory. Then use the Latin hypercube sampling method to generate N s The input sample point set X of the initial sample is then evaluated. The mass structure objective function of the excavator boom and the stress constraint conditions of the excavator boom are obtained through simulation. The excavation energy consumption and excavation volume control objective functions are directly calculated through explicit functions. The output response Y is obtained by bringing the input sample point set X into the objective function and constraint conditions. The sample point set X is used as the initial population P. Step 2: For the objective function of the excavator boom mass structure and the stress constraint condition of the excavator boom, a Kriging proxy model is constructed through the sample set (X, Y); for the excavation energy consumption and excavation volume control objective functions, no proxy model needs to be constructed, and the current Pareto frontier is calculated; Step 3: Use the population iteration method in the NSGA-III algorithm to iterate the initial population p of the excavator optimization problem t max times, and obtain the population after iteration Step 4: Get the population The expected hypervolume improvement calculation is performed on the points in the design space using a nearest neighbor approximation method. First, a sample size of N is calculated in the entire design space. E uniform sampling of , and calculate the dominance relationship between all samples, and then obtain the hypervolume improvement of each sample relative to the current Pareto frontier; population The Kriging model predicts each point in as a high-dimensional normal distribution. The high-dimensional normal distribution of each point is calculated with a sample size of N. normal Sampling is performed, and for the jth normal distribution sample, the nearest uniform distribution sample is expressed as: in,‖ ‖ 2 represents the 2-norm, p j is the jth normally distributed sample, q is a uniformly distributed sample point, q j is the distance p in the uniformly distributed sample j The nearest point; p j Hypervolume Improvement HVI(p i ) is expressed as: HVI(p j )=HVI(q j ) (2) By taking the average of all normally distributed sample points, the expected hypervolume improvement EHVI(x) is obtained as: Finally, we get the population The expected hypervolume improvement for all samples in ; Step 5: Find the population The individual with the largest expected hypervolume is expressed as: The new point x is evaluated using the excavator boom mass structure objective function, the excavator boom stress constraint condition, and the excavation energy consumption and excavation volume control objective function. new , and get its response y new ; will (x new ,y new ) is added to the training sample set (X, Y), and a new round of initial population is obtained through the population selection method in the NSGA-III algorithm. Steps 2 to 5 are repeated until the total number of evaluations of the structure-control integrated optimization problem of the unmanned mining excavator reaches N. max ; From the number N max The optimal Pareto solution is obtained from the sample set (X, Y) as the solution to the structure-control integrated optimization problem of the unmanned mining excavator.
