Dynamic production plan policies under random yield: New models

Many make-to-order companies faces random yield in their production areas. Due to this randomness, deviations from the production plans and underproduction occur. In order to solve the random yield issue, overproduction with multiple lot releases is proposed in the random yield literature. Note that the problem of multiple lot releases is classified as reject allowance or multiple lot sizing production to order (MLPO). The optimal lot size policy under random yield has been extensively studied in the reject allowance - MLPO literature. Nevertheless, there has been no direct study on the optimal production policy. Therefore, in this thesis, the structure of the optimal production policy under random yield is investigated. Additionally, the yield rate models of unknown yield rate and improving yield rate are included. The structural changes of the optimal production policy under these yield rate models are analyzed. First, we formulate a stochastic dynamic programming (SDP) model to understand the base optimal policy. Then, the SDP model is extended to a data-driven case for different yield models. In our analysis, we have derived some general properties of the base optimal production policy and discovered that the optimal policy of the data-driven models consists layers of the base optimal policy. In addition, various approximate methods are derived to understand the threshold equation of "when to produce". Finally, a simulation study on the data-driven case is conducted in order to assess the value of information. Consequently, an extensive study on the optimal production policy under random yield is presented.