Neural Network Applications to Buffer Allocation Problems in Automated Transfer Lines

1College of Engineering and Petroleum, Kuwait University

2Faculty of Engineering Fayoum University, Fayoum, Egypt

Efficiency of serial production flow lines is severely affected by equipment unreliability,

particularly if the stages of production are rigidly linked and sufficient storage spaces are not

provided. However, excessive storage space increases the initial system costs as well as workin-

process levels while moderate storage space may not meet the production requirements.

Therefore, in the design and operation of serial production lines, determination of the amount

of storage space to be allocated to each station is particularly important. This paper presents a

procedure, which combines simulation and neural network modeling approach to determine

the optimum buffer capacities between the stages of an unreliable production flow line based

on such line parameters as operation times, expected equipment failure rates, and the repair

rates. Case problems are solved and the results are compared to previously reported

algorithms. Study results indicate that the proposed procedure could be a useful tool in

designing serial production lines with intermediate storages.