Improvement Of Tree-Seed Algorithms And Their Applications

Swarm intelligence is an optimization technology,that solves the management field optimization problems.Mathematical or statistical methods are difficult to solve complex optimization problems with high-dimension,multi-mode,and non-linear.Swarm intelligence algorithms inspired by nature can obtain approximate solutions to meet practical requirements.As a popular algorithm in swarm intelligence,Tree-Seed optimization algorithm(TSA)has attracted much attention due to its few parameters and simple operation.However,in solving complex problems,there are still some deficiencies,including the imbalance between exploration and exploitation,local stagnation,premature convergence,and so on.To solve the above problems,this study proposes the following three algorithms variants and evaluates them by practical engineering design problems.Improved algorithm 1.fb＿TSA: Tree-Seed Optimization Algorithm based on feedback mechanism.The design of classical TSA lacks the feedback mechanism,which is mainly manifested as follows.First,the fixed search trend parameter ST leads the self-adaptive imbalance between exporation and exploitation.Second,the seed generation mechanism with fixed number of seeds does not realize the personalized setting among trees.Therefore,based on the feedback mechanisms,firstly,the dynamic feedback mechanism of search trend parameter ST is introduced to propose st＿TSA.Secondly,the dynamic feedback mechanism of seed number(ns)is designed to proposed the ns＿TSA algorithm.Then,the seed migration mechanism is introduced based on the above feedback mechanisms to propose fb＿TSA algorithm which can realize the adaptive balance between exploration and exploitation,and avoid the premature convergence.Improved algorithm 2.TSA-BOA: a hybrid optimization algorithm based on tree-seed and butterfly optimization algorithm.TSA has endogenous defects in seed generation mechanism and tree updating mechanism,which are as follows.Firstly,there is a many-to-one relationship between seeds and trees,resulting in the local stagnation.Second,the renewal of trees is only depended on the replacement of seed offspring.If there are no outstanding offspring,trees will face the dilemma of failing renew.Therefore,firstly,this thesis introduces the aroma propagation mechanism inspired by the butterfly optimization algorithm.Secondly,the thesis designs the new optimization idea to select out the high-quality trees,which can solve the above defects of the TSA for tree and seed generation to improve the exploration ability.Improved algorithm 3.sinh TSA: tree-seed optimization algorithm based on candidate and adjustment mechanisms.The seed position of the TSA update based on uniform distribution,which is difficult to obtain the rapid convergence speed of high-dimensional optimization problems.Moreover,retaining a single best tree position leads to the insufficient diversity.Therefore,firstly,a candidate mechanism is introduced to ensure the existence of the optimal solution.Secondly,based on the sinh function,we redesign the seed position through the step size and direction to update,and optimize the adjustment mechanism,to make up for deficiencies in solving high-dimensional problems.To evaluate the performance of the proposed three algorithms variants,this thesis provides comparative experiments with the baselines including three improved algorithms,several TSA variants,and the current classical swarm intelligence algorithm based on the IEEE CEC 2014 and IEEE CEC 2017 standard functions suites.Furthermore,the thesis applies three algorithm variants to solve six practical engineering design problems.The experimental results verify the innovation effectiveness of the three variants and the broad application prospects for engineering design problems.