Multi Objective Evolutionary Algorithm And Its Application In Gear Reducer Of Mining Excavator Hoisting Mechanism

Many real world problems such as engineering design,portfolio optimization,resource allocation,route planning,etc.are indeed of optimization criteria.Originated from the Darwinian Evolution Theory,evolutionary algorithms(EAs)comply with the principle of Survival of the Fittest.Derived from the population-based search mechanism,EAs are able to solve complicated optimization problem in absence of the gradient information.Meanwhile,a set of solutions rather than one single solution are obtained after each iteration,the efficiency of EAs outperform the convential algorithms.Multi Objective Evolutionary Algorithms(MOEAs),a key branch of EAs,has prospered over these years,and been successively applied to solve many scientific research and practical applications.Multiobjective Evolutionary Algorithm Based on Decomposition(MOEA/D),the seminal framework of MOEAs,degenerates in some occasions.In light of this,this dissertation is devoted to intensively extend and modify MOEA/D in two scenarios(decision spapce and objective space)to improve its performance.The major work and core contribution of this dissertation include:(1)To alleviate the difference performance in convergence and diversity derived from the selection of different operators,a hybrid indicator based adaptive operator selection mechsnism is proposed.By collaboratively evaluating hybrid indicators of the solutions generated by temporary operator,it can decide which operator to select in the next generations so as to reduce unuseful iterations.Meanwhile,as CSA shows significant performance in addressing the exploration of search space,we integrated it with DE operator to formulate a new selection pool.Experiment results indicate that the selection mechanism is efficient in most cases to address the balance between exploration and exploiation in search space,so as to balance convergence and diversity.(2)To alleviate the nonuniformly distributed solutions generated by a fixed set of evenly distributed weight vectors in the presence of nonconvex and disconnected problems,an adaptive vector generation mechanism is proposed.A coevolution stragtrgy and a vector generator are synergistically cooperated to remedy the weight vectors.Optimal weight vectors are generated to replace the useless weight vectors to assure the optimal solutions distribute evenly.Experiment results indicate that this mechanism is efficient to improve the diversity of MOEA/D.(3)In light of the fact that the Decision Makers’ are more likely to generate the preferred solutions rather than a whole approximation of the true Pareto Front.An interactive satisfactory theory is introduced to articulate the preference on different objectives.The preference of the decision maker is articulated by satisfactory function.By interactively adjusting the preference,a more in-depth understanding of the MOPs is obtained to make sure that the solutons are evolved towards the preferred region of the true Pareto front.Experiment results indicate that the satisfactory optimization based interactive articulation of preference is able to explore the bounded parts of the true Pareto front to some extent and maintain a good convergence compared to efficient algorithms.(4)A safety factor based multi-objective optimization model with respect to the twostage gear reducer of a heavy mining excavator is formulated in accordance with its working condition and design’s demand.The algorithms proposed in the early sections are utilized to optimize this particular multi objective problem.The results indicate that the proposed algorithms are able to solve this particular optimization problem.