LOADING CONCEPTS FOR THE CELLULAR MANUFACTURING SYSTEM

TIMOTHY J GREENE, Purdue University

Abstract

The advancement of Group Technology and Cellular Manufacturing in the past two decades has created a new problem. That problem is how to load jobs into the cellularly divided manufacturing system so as to balance the machine loads within a cell and the cell loads between cells. The research objective was to develop methodologies to load jobs into the cellularly divided job shop. The methodologies developed maintain the work load balance within and between cells, and for the facility's performance measures, perform significantly better than baseline techniques. The structure of the problem consisted of a queue of jobs where each job had to be assigned to one cell of a cellular manufacturing system. A job could only be assigned to a cell that contained all the machine types necessary to complete the job. Such a cell was called a feasible cell. It was assumed that each job had at least one and usually two or more feasible cells. A mixed integer program was developed that, if trackable, would optimally assign and schedule the jobs into and through the cells. The MIP used one of three objective functions which minimized mean flowtime, makespan, or lateness. The MIP was untrackable for all but the smallest problems. The assignment of a job to a cell requires two tasks; determine which job, and determine which cell. The two tasks can be performed in either order. If the job is identified first, and then the cell selected; the assignment procedure has JOB PRIORITY. If the cell is identified first and then the job selected; the assignment procedure has MACHINE PRIORITY. After a thorough analysis of the system characteristics, job and cell attributes were employed in different forms as criteria to identify and select the jobs and cells. Over thirty different assignment techniques were developed and examined. Two assignment techniques, JPL, and MPL, were determined to perform better than the other techniques examined. JPL is a Job Priority Loading heuristic and MPL is a Machine Priority Loading heuristic. JPL and MPL were evaluated across a wide range of conditions. Random cell assignment and a basic cell balancing procedure were used as baselines for comparison. The evaluation was performed using a SLAM simulation model. The performance measures evaluated were mean flowtime, mean tardiness and mean machine utilization. Shortest operation time truncated (SOT/T) was employed as the sequencing heuristic. From the evaluation it was determined that JPL and MPL were significantly better than the two baseline loading techniques for the range of variables examined. In addition, it was determined that several of the variables examined, (number of cells, remainder cell, job density, and preload level) had an effect on the performance measures but did not effect the relative ranking of JPL and MPL versus the two baseline techniques. Four common sequencing heuristics were evaluated in a posttest; and it was determined that for JPL and MPL, SOT/T performed as well or better than the other three sequencing heuristics. In summary, the problem of how to efficiently and effectively load the cellular manufacturing system was examined with the objective being the development of techniques to load the cellular manufacturing system. Two loading heuristics were developed that performed significantly better over the range of variables examined, than did existing techniques.

Degree

Ph.D.

Subject Area

Industrial engineering

COinS