Comparative Research On Optimization Of Crawler Crane Complex Truss With Multi-Surrogate Models

With the increasing complexity of large construction machinery structures,it is extremely difficult to obtain the relationship between objective functions and constraint functions of optimization functions and many optimization variables for most complex practical engineering structure optimization problems.At present,more researchers rely on finite element software to optimize engineering problems,but the solution of finite element model of complex engineering problems often takes a lot of time,and the multi-variable optimization process often takes thousands of iterations,which makes the finite element solution workload increase geometrically.Therefore,a method that can effectively reduce the calculation amount of optimization process-surrogate model comes into being.The optimization method based on the surrgate model can greatly improve the optimization efficiency of engineering problems.However,there are many kinds of surrogate model algorithms,and different algorithms have different application scopes.In view of this,this paper compares and analyzes the application of three common surrogate model algorithms in complex truss structure optimization,and finally proposes a combination of surrogate model algorithm with higher accuracy,and verifies the effectiveness and applicability of the proposed method with examples.The main research content of this paper includes the following aspects:(1)In terms of surrogate model,this paper studies the modeling mechanism of three surrogate models,namely polynomial response surface,radial basis neural network and Kriging model,and analyzes their applicability from the principle level.In addition,the sampling methods and error criteria of surrogate model are introduced.Secondly,by using four different types of functions,Griewank,Schaffer,Sumsqares and Schwefel’s problem 12,the effects of three kinds of surrogate model algorithms on the accuracy of the surrogate model are compared and analyzed respectively from the types of test functions,dimensions,number of sample points and order of polynomial response surface.(2)In the aspect of finite element modeling,based on the combination of the crawler crane boom structure as optimization object,uses the built-in APDL language of ANSYS software to carry out the whole machine modeling,and carries out nonlinear static analysis of the whole machine model under the specified load combination.In order to get the simulation results more in line with the actual situation,on the basis of the statics model,the hoisting wire rope modeling is added,and the transient dynamics analysis of the finite element model under the two conditions of lifting and sudden unloading is carried out.The dynamic response of the finite element model under the two dynamic schemes is researched.(3)In terms of optimization methods,aiming at the optimization problem of the surrogate model,an algorithm combining global optimization algorithm and gradient algorithm(MIGA-MMFD)is proposd,and the strong convergence of the proposed algorithm is verified by the test function.The optimization problems of a simple plane truss and a multi-bar space truss are analyzed using radial basis neural function and polynomial response surface respectively.Finally,PRS,RBF and Kriging models are used to fit the crawler crane boom system.In order to get better surrogate accuracy,the research method of partition is adopted.According to the advantages and disadvantages of different surrogate models,a combined surrogate model of RBF-PRS is proposed,which is optimized by MIGA-MMFD algorithm.The effectiveness and applicability of this method are verified by the engineering practice of crawler crane complex truss optimization.