The Experience-based Learning Algorithm And Its Engineering Application

Experience-based learning algorithm(EBL)is an intelligent optimization algorithm recently proposed by the author.It is a random optimization algorithm inspired based on the mutual learning behavior of different individuals in the population.The advantage of this algorithm is that it has simple structure with no parameters to be adjusted.However,as a swarm intelligence algorithm,EBL algorithm is difficult to avoid the shortcomings of such kind of method.For example,the algorithm is easy to fall into local optimal solution at the initial stage of operation,resulting in too early convergence,and thus lead to wrong solution results.Moreover,when the algorithm is applied to different engineering optimization problems,the original EBL algorithm may not be applicable and it needs to be improved according to the actual situation.Based on the analysis of the research status of EBL algorithm,this paper puts forward different improvements from the perspective of various practical projects,and applies the improved algorithm to various engineering optimization problems.The main research work and conclusions completed in this paper are summarized as follows:(1)For the delamination detection problem of composite laminates,a two-step delamination detection method based on different modeling methods is proposed.In the first step of the two-step method,the one-dimensional through-thickness equivalent beam element is used to model the composite laminated beam,and the potential delamination damage location in the structure is predicted by EBL algorithm;In the second step of the two-step method,the typical three-dimensional solid element is used to model the composite beam,and the EBL algorithm is used to detect the specific delamination damage,including the delamination position,the size of the delamination and the number of interface layers.The two-step method combines the advantages of two different modeling methods and can significantly reduce the computational cost without sacrificing the detection accuracy.In order to verify the effectiveness and robustness of the proposed two-step method,the layered damage identification of eight layer quasi Isotropic Symmetric Composite Beams with different delaminations is carried out.The identification results are compared with one-step method and other three population intelligent algorithms including cloud model based fruit fly optimization algorithm(CMFOA),particle swarm optimization(PSO)and squirrel search algorithm(SSA),and the influence of noise on the detection accuracy is studied.Numerical and experimental results show the superiority of the proposed two-step method in detecting delamination in composite laminates.(2)For the damage problem of functionally graded materials,an objective function based on sparse regularization is proposed,and the EBL algorithm is used to identify the damage of functionally graded materials.The proposed new objective function can improve the accuracy and robustness of optimization search,so as to obtain better damage identification performance.The functionally graded beam specimens with different damage conditions are made of poly(methyl methacrylate)(PMMA)polymer materials.The damage identification of the beam specimens is carried out to verify the effectiveness and robustness of the proposed new objective function.The identification results are compared with the traditional objective function.In addition,the influence of noise on the identification performance is also studied.The numerical and experimental results of functionally graded beam samples show the superiority of the objective function based on sparse regularization in the damage identification of functionally graded materials.(3)For the problem of soil parameter identification in the construction stage of rockfill dam,an improved EBL algorithm based on the concept of quasi-opposition learning(OEBL)is proposed.The algorithm inherits the advantages of the original EBL algorithm,such as no need to set parameters,simple structure,and fast convergence speed.In order to improve the exploration ability of the algorithm,the quasi opposition learning principle is introduced,and the OEBL algorithm is used to identify the soil parameters of monkey rock face rockfill dam to verify the performance of the proposed algorithm.The identification results show that the settlement value of rockfill dam calculated by the parameters identified by OEBL algorithm is very consistent with the field measured value.The soil parameters in each construction stage are very different,and are affected by the field construction speed.The recognition results are compared with the original EBL,PSO,CMFOA and SSA algorithms.The comparison results show that OEBL algorithm has good performance for the soil parameter identification during dam construction.(4)For the identification of mechanical parameters of soil layer,the improved OEBL algorithm based on quasi opposition learning principle is also used to verify the performance of the algorithm in soil layer parameter identification again.Based on the characterization of soil dissipation curve,the mechanical parameters of soil layer are identified by OEBL algorithm.The identification results show that OEBL algorithm has good performance in the identification of mechanical parameters of soil layer.(5)For the problem of nonlinear pattern updating of concrete model,an improved EBL algorithm(MEBL)is proposed to improve the ability of individual parameter updating of concrete model.The single degree of freedom system of bouc-Wen model and Bouc–Wen–Baber–Noori(BWBN)model is numerically simulated to study the effectiveness and robustness of MEBL algorithm,and four precast concrete infilled walls are tested in the laboratory.BWBN model is used to describe the hysteretic behavior of infilled walls,and the MEBL algorithm is used to identify the parameters of BWBN model.The identification results are compared with the original EBL,CMFOA,Jaya,PSO and SSA algorithms.The numerical and laboratory results show that the proposed MEBL algorithm can successfully identify the nonlinear hysteretic parameters of BWBN model,and the proposed MEBL algorithm is superior to other algorithms.(6)For the problem of parameter identification of nonlinear model of shield tunnel nodes,an EBL algorithm combining the concept of quasi-opposition and a new update mode(IEBL)is proposed.The improved algorithm not only refers to the advantages of quasi opposition principle,but also makes use of the empirical information of more individuals in the population,so as to enhance the global optimization ability of the algorithm.Taking the shield tunnel of Guangzhou Metro Line 7 as the prototype,three scaled model specimens are designed for low cycle reciprocating load test.The hysteretic behavior of shield tunnel model nodes is described by asymmetric bouc Wen model,and the parameters of the model are identified by the proposed IEBL algorithm.The identification results are compared with the original EBL,CMFOA,Jaya,PSO and SSA algorithms.The identification results show that IEBL algorithm can successfully identify the nonlinear hysteretic parameters of asymmetric Bouc-Wen model with better the identification effect than other algorithms.