| Mechanical structures are mostly connected by multiple components.Connection positions and distributions affect the overall performance of the mechanical structure of components,such as static characteristics,dynamic characteristics,accuracy and life.Welding is a widely used connection.For large-scale structures,there may be a lot of spot welds,and the locations of connections between components are huge,which must be determined by the optimization method of the system.Using models with large-scale number of degrees of freedom(such as finite element models)can ensure that the simulation has sufficient computational accuracy.However,for the optimization of structural connection locations and other engineering problems requiring repeated evaluation of the overall performance of the structure,this will consume a lot of computational costs,even seriously affecting the feasibility of optimization,.In order to reduce the computational cost of reanalysis of the structure in the optimization iteration process,the proper orthogonal decomposition(POD)and radial basis function neural network(RBF)are introduced to optimize the distribution of the spot welds in the "double-hat" structure.Specific research work includes:(1)Finite element analysis of "double-hat" benchmark structure is carried out in ABAQUS.The parametric finite element model of the structure is established by using Python language to realize the automation of modeling,simulation calculation and post-processing of the results.The positions of spot welds are sampled using optimal Latin hypercube design in Isight.The nodal displacement of the model is extracted,generating the snapshot array of nodal displacement.A suitable number of orthogonal bases are extracted by POD method to obtain reduced order model(ROM)which can ensure that the model contains sufficient performance information.(2)The RBF based on POD is introduced.The POD-RBF-ROMs of the model are established according to the coordinates of spot welds and the orthogonal basis of the corresponding groups at the sampling point.By training the model,a set of orthogonal basis coefficients at arbitrary spot welds positions are obtained by interpolation method in the approximate model,and then the structural displacement response corresponding to the spot welds positions is solved..(3)The simulation results of the POD-RBF-ROMs are compared with those of the corresponding finite element model.The error analysis of the POD-RBF-ROMs is carried out,and the validity of the modeling method is verified.(4)Based on POD-RBF-ROMs,taking the maximum displacement response as the objective function,the optimization model of structural spot welds location constraints is established.The optimal distribution of structural spot welds is obtained by using the nonlinear quadratic programming method(NLPQL).The validity of the method is proved by the finite element analysis of the optimized structure.Finally,the advantages of POD-RBF approximation model are analyzed.Combining POD with RBF,this paper successfully establishes a ROM for spot welds distribution of a "double-hat" structure.The model can quickly and effectively carry out approximate reanalysis of welded structures,solve the problem of excessive calculation cost caused by repeated valuation in optimization problem,and realize the optimal design of solder joint distribution mode of structure.The research results have certain theoretical guiding significance for related engineering problems. |
