How yield process misspecification affects the solution of disassemble-to-order problems

Random yields from production are often present in manufacturing systems and there are several ways that this can be modeled. In disassembly planning for product recovery, e.g., we face high yield uncertainty in harvesting parts from cores. In this case, yield randomness often is modeled as either stochastically proportional or binomial. A statistical analysis of data from engine remanufacturing of a major car producer fails to provide conclusive evidence on which kind of yield randomness might prevail. In order to gain insight into the importance of this yield assumption, the impact of possible yield misspecification on the solution of the disassemble-to-order problem is investigated. Our results show that the penalty for misspecifying the yield method can be substantial, and provide insight on when the penalty would likely be problematic. The results also indicate that in the absence of conclusive information on which alternative should be chosen, presuming binomial yields generally leads to lower cost penalties and therefore preferable results.