Research And Application On Multi-objective Evolutionary Algorithms In Uncertain Environment

Evolutionary algorithm is proved to be a virtual way to solve mulit-objective optimizationproblems. But the MOPs in reality are complex, non-linear and uncertainty, especially theuncertainties in environment can’t be ignored. The capacity of algorithm to counter interference isimportant to reality application. So the research of the multi-objective evolutionary algorithm issignificant. Shop scheduling is the requirement of advanced production, the guarantee of productionsource and production cost, one of the most important aspects of production automation, and aeffective way to improve production efficiency, because of those, shop scheduling draw the attentionof deciders and researchers. Flow shop is one of the most typical scheduling models of production. Ithas a high complexity in theory and typicality in application. In most research, the flow shopscheduling is taken as a certainty problem which has present parameter. But the product environmentsare variable, so the research on fuzzy flow shop is meaningful, which can decide the validly of thescheduling project. The main work of this paper includes the following three aspects:Firstly, three typical multi-objective algorithms are taken in a comparative experiment withinterference from different dimension noisy. The effectiveness of four metrics are tested, and theMonte-Carlo integration are used as sample approach, and the feasibility are test by experiment, thescope which is appropriate for Monte Carlo integration, and the shortcoming of it is stated. The resultsof the experiment shows regression of the MOEAs for noisy fitness function, the traditionalMonte-Carlo integration is inapplicable when the noisy is large, and new metric for noisy fitnessfunction is inevitable.Secondly, no-wait flow shop scheduling problem is an important area in production scheduling,we applied estimation of distribution algorithms, to solve the NWFS problem. To compute themakespan fast, we prove a characteristic of NWFS, then proposed a algorithm with a O(n)computation complexity depend on the characteristic. We use a graph theory based neighborhoodsearching method to improve the neighborhood searching procedures, to gain a most reduction of themakespan then accelerate the convergence to good solutions,. The effect is proved by the experiences.Thirdly, for the fuzzy flow shop scheduling problem, an modified MOEAs are proposed to solveit. The mathematic model of fuzzy flow shop scheduling problem is established, expect the makespanwhich is taken as optimization object in traditional scheduling algorithms, the newly defined fuzzylevel function is taken as the second object, a preservation of population by the cut operation based oncrowding distance,5classic benchmarks are tested on the proposed algorithm.