A manufacturing layout reasoning architecture based on automated integration of linear objective optimization and nonlinear qualitative analysis
Abstract
A combined interactive and automated reasoning system which can rectify a representative set of the commonly encountered qualitative anomalies in layouts (such as undesirable flow path configuration, unutilized regions between cells) is addressed. Local neighborhood manipulations of the qualitative layout anomalies are attempted by heuristic guidelines and their global impacts are assessed by a flow travel based linear objective optimization subject to linearized design constraints. Such anomalies are presently rectified by human layout designers on an ad hoc basis. The system is currently based on two useful layout design philosophies: (a) the desire to generate the most compact layout given the building and other structural constraints and (b) the desire to avoid direct intersection of inter-cell flow paths with the cell configurations. An object-oriented system design using Smalltalk-80 has been adopted. Such a design permits a very modular representation of entities and provides benefits like high level user interface and openness of the solution states to the user. For example, the layout and material flow entities, whether physical (like cell, flow) or abstract (like local qualitative layout problem, automated layout analyzer), are clearly partitioned in the software as objects. A system entitled QLAARP (Qualitative Layout Analysis using Automated Recognition of Patterns) is introduced. An LP solver package is used for linear objective optimization. The automated reasoning is modeled by an analyzer object which, by default, adopts a hill climbing solution search strategy for selecting the best quantitative solution from a collection of solutions obtained through a series of generate and test operations involving local neighborhood manipulations on each targeted layout anomaly. Better results have been obtained for a set of previously published layout design situations. The methodology has also been applied at workstation level of details by treating the workstations as cells. The system can be applied to alter existing layouts of factories.
Degree
Ph.D.
Advisors
Kashyap, Purdue University.
Subject Area
Industrial engineering|Operations research|Artificial intelligence
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.