An Integrated, Evolutionary Approach to Facility LAyout and Detailed Design
The unequal-area, shape constrained facility layout problem is a NP-hard combinatorial optimization problem concerned with minimizing material handling costs. An integrated methodology that incorporates a genetic algorithm and a constructive heuristic is developed to simultaneously solve the traditional block layout problem of locating and shaping departments and the detailed design problem of locating the input/output stations of departments. These problems have received much attention over the past half-century with the majority of research focused on solving them individually or sequentially. This thesis aims to show that an integrated methodology which combines the problems and solves them in parallel is preferable to sequential approaches.
The complexity of the integrated layout problem is reduced through a Flexbay formulation and through pre-assigned intra-departmental flow types. A genetic algorithm with a two-tiered solution structure generates and maintains a population of block layout solutions throughout an evolutionary process. Genetic operators reproduce and alter solutions in order to generate better solutions, find new search directions, and prevent premature convergence of the algorithm. An adaptive penalty mechanism guides the search process and reduces the computational overhead of the algorithm. Through the placement of input/output stations, the optimization of a block layouts material flow network is implemented as a subroutine to the genetic algorithm. A contour distance metric is used to evaluate the costs associated with material movement between the input/output stations of departments and aids in constructing practical aisle structures. A constructive placement heuristic places the input/output stations and perturbs them until no further improvement to a layout can be realized.
The integrated approach is applied to several well known problems over a comprehensive test plan. The results from the integrated approach indicate moderate variability in the solutions and considerable computational expense. To compare the integrated methodology to prior methodologies, some of the best results from the unequal-area facility layout problem are selected from prior research and the I/O optimization heuristic is applied to them. The results of the integrated approach uniformly and significantly outperform the results obtained through sequential optimization. The integrated methodology demonstrates the value of a simultaneous approach to the unequal-area facility layout problem.
Advisor:Harvey Wolfe; Bryan A. Norman; Jayant Rajgopal
School:University of Pittsburgh
School Location:USA - Pennsylvania
Source Type:Master's Thesis
Date of Publication:09/13/2004