324 IEEE TRANSACTIONS ON SEMICONDUCTOR MANUFACTURING, VOL. 17, NO. 3, AUGUST 2004 Neural-Network-Based Delivery Time Estimates for Prioritized 300-mm Automatic Material Handling Operations Da-Yin Liao, Member, IEEE, and Chia-Nan Wang Abstract—This paper deals with lot delivery estimates in a 300-mm automatic material handling system (AMHS), which is composed of several intrabay loops. We adopt a neural network approach to estimate the delivery times for both priority and regular lots. A network model is developed for each intrabay loop. Inputs to the proposed neural network model are the combination of transport requirements, automatic material handling resources, and ratios of priority lots against regular ones. A discrete-event simulation model based on the AMHS in a local 300-mm fab is built. Its outputs are adopted for training the neural network model with the back propagation method. The outputs of the neural network model are the expected delivery times of priority and regular lots in the loop, respectively. For a lot to be trans- ported, its expected delivery time along a potential delivery path is estimated by the summation of all the loop delivery times along the path. A shortest path algorithm is used to find the path with the shortest delivery time among all the possible delivery paths. Numerical experiments based on realistic data from a 300-mm fab indicate that this neural network approach is sound and effective for the prediction of average delivery times. Both the delivery times for priority and regular lots get improved. Specially, for the cases of regular lots, our approach dynamically routes the lots according to the traffic conditions so that the potential blockings in busy loops can be avoided. This neural network approach is applicable to implementing a transport time estimator in dynamic lot dispatching and fab scheduling functions in realizing fully automated 300-mm manufacturing. Index Terms—AMHS, neural network, prioritized service, 300-mm semiconductor manufacturing. I. INTRODUCTION C HIP makers have been challenged with the potential ad- vantages and uncertainties of migration to 300-mm wafer fabrication. The positive side suggests attractive cost benefits, more reliable product quality, and higher productivity. On the down side, there are thousands of unknowns to be clarified or new paradigms yet to be developed before 300-mm manufac- turing gets ready for mass production. Highly automated mate- rial handling is one of the biggest concerns to the practitioners. Manuscript received March 31, 2003; revised March 15, 2004. This work is supported in part by National Science Council, R.O.C., under grants of NSC91- 2212-E-260-003 and NSC92-2213-E-260-030. D.-Y. Liao is with the Department of Information Management, National Chi-Nan University, Puli, Nantou 545, Taiwan, R.O.C. (e-mail: dyliao@ ncnu.edu.tw). C.-N. Wang is with the Institute of Industrial Engineering, National Chiao- Tung University, Hsinchu 300, Taiwan, R.O.C. (e-mail: merlinwang90g@ nctu.edu.tw). Digital Object Identifier 10.1109/TSM.2004.831533 Comparing to the operations in 200-mm semiconductor man- ufacturing, a cost-effective 300-mm fab demands highly auto- mated operations in both processing and material transfer in order to optimize equipment utilization and product cycle times. Due to the increased number of chips from a 300-mm wafer, the required number of wafers is reduced by a factor of 2.25. Therefore, a high-mix 300-mm fab has to suffer from higher varieties of products than a 200-mm fab does. High product mix leads to more frequent process changes and fine tunes on process and metrology equipment. Also, it results in frequent process experiments and inspections as well as frequent pilot or risk production. In a wafer fab, a lot will be granted high priority, named Hot Lot or Super Hot Lot, if either it is going to execute several crit- ical operations for experiments or inspections on process con- ditions, or it was born as a pilot or risk lot for process character- ization or design validation before a new product is released to production. Hot lots are very important to both fab operations and product development of IC (Integrated Circuits) designers. Operations of hot lots can be either preemptive against normal operations, or capacity-reserved for no-wait manufacturing. In contemporary 200-mm semiconductor manufacturing, hot lots are specially handled by human operators in order to reduce the transport delay between distant processing equipment. It be- comes very challenging to reduce such delays in a 300-mm auto- matic transport environment. The dynamics of a 300-mm fab are very complicated when incorporating automatic material han- dling systems (AMHS) into the shop floor. Manufacturing of high priority lots has well-known signifi- cant impacts on production cycle times as well as throughputs of regular production [8], [11], [15]. Such an effect is usually believed to become worse in 300-mm semiconductor manufac- turing due to highly automated material handling operations in- volved. Ehteshami et al. [8] conduct object-oriented simulation experiments of a wafer fabrication model to investigate the im- pact of hot lots on the cycle time of other lots in the system. Their simulation results show that as the proportion of hot lots in the work-in-process (WIP) increases, both the average cycle time and the corresponding standard deviation for all other lot types increase as well. They conclude that hot lots induce either worse services for regular lots or an increase in inventory costs. Fronckowiak et al. [11] use a simulation tool, ManSim/X, to analyze the impact for different hot lot distributions for two dif- ferent products. Narahari and Khan [15] model semiconductor manufacturing systems as re-entrant lines and study the effect of hot lots through an approximate analysis of the re-entrant line 0894-6507/04$20.00 © 2004 IEEE