Modelling and rescheduling of a re-entrant wafer fabrication line involving machine unreliability

Modelling and rescheduling of a re-entrant wafer fabrication line involving machine unreliability

0.00 Avg rating0 Votes
Article ID: iaor20051745
Country: United Kingdom
Volume: 42
Issue: 21
Start Page Number: 4431
End Page Number: 4455
Publication Date: Jan 2004
Journal: International Journal of Production Research
Authors: , ,
Keywords: electronics industry
Abstract:

A methodology for rescheduling a semiconductor fabrication line in the presence of predicted and unpredicted machine unreliability and job demand is presented. The proposed methodology consists of four major steps. (1) Representation of the fabrication line: the fabrication line is represented using a Neuro-Expert Petri net (NEPN) model. The entire fabrication line is decomposed into a ‘two machine and one buffer’ subsystem and represented using the NEPN model. (2) Estimation of failure and repair rates: the probabilistic machine failure and repair rates are determined using the method proposed by Jeong and Kim. (3) Activation of the rescheduling algorithm: the rescheduling algorithm is activated after the planned mean time to reschedule the fabrication line. This is calculated by taking into account the estimated machine failure and repair rates, and the threshold queue limits of the waiting jobs. The rescheduling algorithm is also activated in cases of unpredictable machine breakdown or repair, and where the job queue length exceeds a certain threshold. (4) Execution of the rescheduling algorithm: the rescheduling algorithm determines whether or not to reschedule a job onto another machine by taking into account factors such as due dates and profits generated by the jobs. The proposed solution methodology is explained using an illustrative example. The computational results developed for the makespan criteria reveal that the proposed methodology is reliable and performs relatively well.

Reviews

Required fields are marked *. Your email address will not be published.