Research On Structural Optimization Method Of Ensemble Learning Model Based On Mapping Samples

Structural optimization can not only meet the functional requirements of the structure,but also save material consumption and effectively reduce the construction cost.For a long time,it has been a problem of great concern to designers.However,with the development of engineering construction technology,the trend of civil engineering structure towards large-scale,complex and diversified development is becoming more and more obvious,which makes the cost of structural analysis higher and higher,but it also seriously restricts the application of structural optimization method relying on a large number of iterative calculations in practical engineering.Therefore,improving the calculation efficiency has become one of the key problems of structural optimization design.Based on the improved Deb rule(IDeb)and Mapping Strategy(MS),a hybrid constraint handling technique is proposed to improve the computational efficiency of structural optimization method based on intelligent optimization algorithm;In order to reduce the number of finite element analysis in the process of optimization iteration,an ensemble learning model of virtual sample points based on MS is further proposed and used to identify the boundary of feasible region in the process of structural optimization.The main contents and research work of this paper are as follows:(1)According to the characteristics of structural optimization problem,a parameter free constraint handling technique based on IDeb and MS is proposed.Through the IDeb,this method avoids the redundant structural analysis and calculation in the iterative process,and maps the valuable design points to the boundary of the feasible region to improve the search accuracy of the optimization algorithm.The test results of optimization examples show that this method has the same calculation accuracy as the MS,but the calculation efficiency is more than doubled.(2)In this paper,the feasibility discrimination of solution in structural optimization is regarded as a two classification problem in machine learning problem.According to the mechanical relationship between design variables,structural response and constrained state variables,a virtual sample generation method based on MS is proposed.By introducing a mapping constant,the existing finite element analysis results are filled into the optimization design space,so as to generate a large number of low-cost high-precision virtual samples without finite element analysis from a small number of real samples through the mapping method for machine learning.(3)Combined with support vector machine(SVM)in machine learning and bagging algorithm in ensemble learning,a structure optimization agent model based on virtual samples is proposed.This method uses the ensemble learning method to combine multiple SVM models trained by mapping virtual sample points,constructs a high-precision strong classifier,and uses the obtained ensemble learning strong classifier to identify the boundary of feasible region of structure optimization,To replace the constraint evaluation method based on structural analysis in the optimization process.The test results show that this method can improve the computational efficiency by about three times compared with the traditional constraint handling technique.(4)The shape and size of six groups of benchmark structural optimization examples are optimized by using five algorithms: HS,PSO,TLBO,CS and MRFO to test the optimization performance of the method proposed in this paper.The test results show that the two methods proposed in this paper can be effectively combined with the intelligent optimization algorithm,and can obtain similar optimization results based on the general optimization constraint handling technique,which has good universality and robustness.The hybrid constraint handling technique can double or more the computational efficiency of the four groups of optimization examples;The optimization method of virtual sample agent model can improve the efficiency by three times or more.When the two optimization methods are applied to the engineering optimization design,the results save about 20% of the materials compared with the initial design,and save about 58% and 89% of the calculation cost compared with the general constraint handling technique.The optimization results show that the proposed method has good universality and efficiency,and provides a structural optimization scheme.