Research On Structural Optimization Design System Based On Parallel Optimization Algorithm

Optimization design for practical engineering mechanical structures commonly faces with multidimensional design parameters and nonlinear constraints,resulting in the failure of the traditional local optimization algprithms(such as sequential quadratic programming,interior point,active-set optimization,and trust region reflective)and the low computational efficiency of the global swarm intelligence optimization algorithms.Regarding to the disadvantanges of the global swarm intelligence optimization algorithms employed in mechanical structures,this work proposes an improved genetic algorithm(GA)and an improved particle swarm optimization(PSO).Combining with the parallel algorithm,this work constructs a parallel optimization design strategy.Based on the parallel algorithm,the improved GA and the improved PSO,this work further provides a parallel-improved GA and a parallel-improved PSO.Finally,this work develop an software relevant to the previous methods,and various examples including benchmark examples and engineering examples are employed to demonstrate the advangages of the proposed approaches.The detailed contents are given as follows.This work will present some improvements to overcome the shortcomings of the traditional GA such as unstable convergence and low computational efficiency.Firstly,this work employs a normal-distribution-based crossover operator and a adaptive-degree-variation-based variation operator for improving the computational efficiency of the genetic variation process involving in searching the best.Then,this work further introduces the interior point penalty function to improve the traditional GA for dealing with nonlinear constraints,overcoming the issues of the traditional GA suchas the low computational efficiency.This work employs an external set of profiles of elite strategies,gradually decreasing inertia weights and domain variants to circumvent the advantages of the traditional PSO,such as premature maturity,lower computational efficiency,and poor particle diversity.Five benchmark examples are employed to demonstrate the effectiveness of the proposed method.Aiming at the low computational efficiency of swarm intelligence optimization algorithm.And combined with the natural parallelism of swarm intelligence optimization algorithm.We propose a multi process parallel strategy based on coarse granularity.This strategy deals with the initial population according to the characteristics of the parameters to be optimized in structural design.Then,each subpopulation is assigned a process to carry out an independent iterative optimization process.Combined with the improved genetic algorithm and improved particle swarm optimization algorithm proposed above.And based on coarse-grained multi process parallel strategy.Finally,five typical benchmark examples are used to verify the efficiency and accuracy of the proposed method.This structural optimization design system is built using Qt Creator and is capable of interacting with Python and ANSYS finite element analysis software.The entire process of structural design optimization can be done in the user interface,and the results of the optimization calculation are also displayed in the user interface for further optimization analysis.Finally,the system function is tested by using the ten-bar truss structure optimization design problem and the car side impact response surface model structure optimization problem as an example.The test applies the improved parallel genetic algorithm and the improved parallel particle swarm algorithm proposed in this paper to solve the problem,respectively.The results show that the structure optimization design system of this paper basically meets the design requirements and has certain engineering significance.In order to facilitate the application of mechanical structure optimization design.This paper uses QT creator platform and python language to develop a set of structure optimization design system.Finally,the structural optimization design of ten bar truss and the structural optimization of vehicle side impact response surface model are used to verify the engineering adaptability of the proposed algorithm and the developed system.The results show that the proposed algorithm and the developed system have a certain engineering application value.