Using learning theory in assembly lines for new products

This paper presents a planning model for the assembly of new products, generally required in small quantities, and whose basic task times are relatively long. As a consequence, some consideration of Industrial Learning is incorporated into the planning model. An algorithm based on this model was successfully implemented for on-line planning and scheduling of a new production line for the assembly of aircraft engine housings in which the plant had limited experience. The jigs used for production are very expensive and would not be duplicated. Also, the arrival time of parts needed for the assembly could not be guaranteed, and updating of due dates needs to be considered by the planning model. Initial estimates of the learning parameters are updated continuously by the model, and as a result, line rebalancing is undertaken to account for task times reduction and modification due to both learning and correction of initial estimates. The powerful MUST algorithm was selected to generate the station balances since their requirement is for frequent rebalancing of the production line due to changes in task-times. Trial runs yield good solutions for rebalancing the line, while satisfying ‘zoning’ constraints, and keeping the maximum number of activities in their original stations, so as to minimize potential learning losses.