Managing product transitions with learning and congestion effects

The introduction of a new product into an operating factory can have significant adverse impacts on the throughput and cycle time of all products produced in the factory, and thus needs to be managed carefully. In previous work we proposed a production planning model for new product introductions that captures the impact of additional variability caused by the new product and of learning as experience producing the new product is gained. This paper extends the earlier work by incorporating learning through deliberate experimentation using engineering lots and the impact of cycle time on line yield due to delays in detecting adverse events. We formulate a non-convex nonlinear optimization model to determine the mix of production and engineering lots to be processed, and obtain approximate solutions using a genetic algorithm. Numerical experiments with different scenarios show the importance of carefully managing the releases of production and engineering lots and of accelerating learning early in the time horizon through judicious use of engineering lots.