Improvements Of Harmony Search Algorithm With Its Applications In Optimization

Global optimization problems widely exist in different fields of scientific research and engineering applications.They are often non-differentiable,non-convex and multimodal so that the exact optimization methods such as gradient descent method and newton method are not efficient in these cases.The metaheuristic algorithms are well known efficient optimization methods.Although they can not guarantee to obtain the exact global best solution,they can obtain near global best solution within reasonable time.Therefore,they are more suitable for real world complex optimization problems and become one of the research hots in the area of intelligent optimization.Harmony search(HS)algorithm is a recently developed metaheuristic algorithm.It has characteristics like simple in concept,easy for implementation and with a few parameters.HS has been widely applied to different area.However,HS has drawbacks such as low solution accuracy and sensitive to its parameters.Hence,it is worth to further to improve its performance.What's more,due to the 'No free lunch' theory,there is no a general method that suitable for solving all types of problems.The aim of this dissertation is to study the optimization mechanism of HS,and develop modified HS variants that can solve complex problems such as nested experiments design,expensive optimization problems and the optimal tool path planning problem in five-axis flank milling.Firstly,a modified HS with intersect mutation operator and cellular local search is proposed.the MHS algorithm divides all harmonies in harmony memory into two different groups according to their fitness.The novel intersect mutation operation has been developed to generate new-harmony vectors,so that the diversity of the harmonies in harmony memory pool are well maintained.Furthermore,a cellular local search also has been developed in MHS,that helps to improve the exploitation ability by exploiting the space near current harmonies.Therefore,the exploration and exploitation capabilities are well balanced.Two different benchmark function sets are used to validate the performance of MHS.The results demonstrate that not only the MHS outperforms other HS variants but also competitive with other metaheuristic algorithms.At last,the MHS is successfully applied to the size optimization of a PV/wind hybrid renewable energy system.Secondly,a novel nested maximin Latin hypercube design method is developed based on successive local enumeration and a modified novel global harmony search algorithm.In the proposed method,the successive local enumeration is employed to select sample points for a low-fidelity model,whereas the modified novel global harmony search algorithm is employed to select sample points for a high-fidelity model.A mathematical model is also formulated for selecting the high-fidelity sampling points from the low-fidelity sampling points set.The numerical experiment results show that the proposed method has very good properties in space-filling,computation cost and prediction accuracy.At last,a long cylindrical pressure vessel design case is used to further validate the performance of the proposed nested experiment design method.Thirdly,a variable fidelity surrogate assisted HS algorithm(VFS-MHS)is developed for solving expensive optimization problems in complex products design.Owing to the advantages of both the low-fidelity model and the high fidelity model,VFS-MHS can build accurate surrogate model and obtain satisfactory solutions within limited computation budget.To balance the exploration and exploitation,a novel infilling criterion is proposed based on the non-dominate sorting concept in multi-objective optimization.The well known CEC2014 benchmark set of expensive optimization problems is used to verify the performance of VFS-MHS.The results demonstrate that VFS-MHS is superior to other compared methods.And a case study of long cylindrical pressure vessel design employing the VFS-MHS method is also discussed.Fourthly,a novel optimal tool path planning method is developed for 5-axis flank finishing cut based on a geometric decomposition strategy and multi-population harmony search.The proposed approach geometrically divides the surface to be machined into a number of segments.The tool paths on those sub-surfaces are independently optimized by the multi-population harmony search algorithm.The test results on eight representative surfaces show that the proposed approach produces higher machining precision with less computational time than the previous methods.Finally,the above work is summarized and the future research directions are discussed.