Open Oral Presentation
Speaker: Mr. Dhaval Pujara
.
Title:
“Development of Solution Methodologies for Dynamic and Real-Time Scheduling of Single Burn-In Oven and Non-identical Parallel Burn-in Ovens with Machine Eligibility Restrictions in Semiconductor Manufacturing “
Advisor: Dr. M. Mathirajan
Day & Date: Monday, 11th March 2024
Time: 10:00 AM
Venue: Online M S Team Link:
https://teams.microsoft.com/l/meetup-join/19%3ameeting_M2RmNmNkOWYtZDY5Mi00MmU2LTgwODMtMTdiMmM2NDk3MDM5%40thread.v2/0?context=%7b%22Tid%22%3a%226f15cd97-f6a7-41e3-b2c5-ad4193976476%22%2c%22Oid%22%3a%2244805bd2-0703-4c91-93aa-a976ddb8a045%22%7d
Join conversation
teams.microsoft.com
Abstract:
This study addresses dynamic and real-time scheduling of a bottleneck batch processing machine, particularly Burn-in Oven (BO) used in the testing stage of Semiconductor Manufacturing (SM) process to perform Burn-in operation – which is a quality check operation to test the semiconductor chips against immature failure. Based on the detailed analysis of the literature, this study defines a new research problem configuration on dynamic scheduling (DS) of (a) a Single Burn-in oven (SBO), and (b) multiple Non-identical Parallel Burn-in Oven(s) (NPBO) with Machine Eligibility Restriction (MER) to minimize Total Weighted Earliness/Lateness (TWE/L). Further, this study aims to understand the impact of job and resource related real-time events on the performance of the DS algorithms developed for (a) SBO and (b) NPBO with MER to minimize TWE/L.
Initially, (0-1) Mixed Integer Linear Programming (MILP) models are developed for DS-SBO, DS-NPBO, and DS-NPBO with MER with an objective of minimizing TWE/L. Due to computational intractable nature of the proposed (0-1) MILP models, this study has proposed 25 variants of Dispatching Rule (DR) based Greedy Heuristic Algorithm (DR-GHA) for (a) DS-SBO, and (b) DS-NPBO with MER to solve the large-scale real-life instances. The main reason for constructing GHA using DR is due to the fact that DRs are very popular and heavily used in SM industry. The performance evaluation of each of the 25 variants of DR-GHA proposed for (a) DS-SBO, and (b) DS-NPBO with MER are analysed both empirically and statistically by developing a suitable computational experiment. Based on the performance analyses, 5 relatively well performing variants of DR-GHA are identified for each of the problem on DS-SBO and DS-NPBO with MER considered in this study.
As second part of the contribution, considering the BO operating environment in SM, this study proposes a research hypothesis that “appropriate modification in the work-in-process (WIP) data in case of the occurrence of J-RTE: Job related Real Time Events (such as due-date change, early/late arrival of jobs, change in job priority, new job addition, job cancellation) and updating the available time of BO in case of the occurrence of R-RTE: Resource related RTE (such as breakdown, operator illness, tool failure, shortage of material, defective material ) is sufficient when there is an efficient algorithm available / proposed for (a) DS-SBO, and (b) DS-NPBO with MER. That is, there is no need of developing any rescheduling algorithm or modifying the current algorithm whenever any types J-RTE, R-RTE, or both types of RTE (called as JR-RTE) occur while scheduling the Burn-in Oven(s). This proposed hypothesis is proved empirically and statistically by developing a suitable computational experiment. That is, the performances analyses carried out w.r.t. the proposed hypothesis indicated that the ranking position of each of the 25 variants of DR-GHA for (a) DS-SBO and (b) DS-NPBO with MER remains the same when these are applied to occurrences of RTE and non-occurrences of RTE scenarios.
Finally, Artificial Neural Network (ANN) integrated hybrid Meta-Heuristic (MH) Algorithms: Simulated Annealing (SA) algorithm, and Genetic Algorithm (GA) are proposed for (a) DS-SBO and (b) DS-NPBO with MER. The purpose of ANN model is to predict an instance-specific unique set of MH-parameters’ values for given instance. Based on the detailed performance evaluation of these two ANN integrated hybrid MH-algorithms, it is observed that the ANN integrated hybrid SA algorithm is found to perform better than the other MH algorithm considered in this study. Further, the impact of JR-RTE on ANN integrated hybrid MH algorithms is also studied following the similar research processes carried out in understanding the impact of JR-RTE on (a) DS-SBO, and (b) DS-NPBO with MER and observed that efficiency of each of the MH algorithms considered in this study remains the same when these are applied to occurrences of RTE and non-occurrences of RTE scenarios.
There are certain limitations: (i) study only considers Dispatching Rules (DRs) used by earlier studies on scheduling of BO, and (ii) while studying the impact of JR-RTE this study considers equal chances of occurring of each of these types of JR-RTE. Apart from overcoming these limitations, there are several future research directions: (i) development of DS of NPBO with MER considering other due-date and/or completion time-based scheduling objectives, and (ii) development of new DRs while developing schedule for Burn-in Oven(s).
ALL ARE WELCOME