| This thesis concentrates on reduction of cost of contract manufacturing for semiconductor companies. It describes a case study undertaken at an Original Equipment Manufacturing company. The contract manufacturers (S1, S2, and S3) in this case provide Assembly and Test services for the company. All the products that are manufactured by the company can be clustered within different process families (defined by factor A, B and C). The capability of the suppliers is also defined by process family. That is, if a supplier has the capability to manufacture (Assemble and Test) a particular process family, then all parts within that process family can be outsourced to that supplier, provided it satisfies all the necessary constraints. Constraints can be categorized as part level (capability constraint) or operational constraints (assembly and test capacity constraints). To better manage overall business relationship with suppliers and to mitigate the risk exposure to any one supplier, the company has defined a few business rules that were also modeled as constraints within this project.;A Linear Programming (LP) model with 50 process families, 1500 parts, and 3 suppliers is formulated and solved by LINGO. This computational model allows the company to provide an optimal mapping solution for its products based on inputs (price for assembly and test and transportation fee) and all the constraints associated with manufacturing. A small example with dummy, but reasonable, data is used to illustrate the optimal sourcing strategy and the result is validated by numerous test cases as well as the company's historical data. The value of the tool is not only to optimize manufacturing spend but also to monetize and quantify some of its constraints. In addition, LINGO is able to retrieve data from EXCEL, which is the data handling interface of many companies. Based on EXCEL's input, this LP model provides convenience by avoiding writing explicit LP equations. The remaining part of the thesis concentrates on accommodating this outsourcing model to be capable of treating two most commonly utilized quantity discount schemes, all-unit quantity discounts and incremental quantity discounts. Once again, a small example is adopted for demonstrating both pricing discount schemes, and the results show that with proper modification, the LINGO model is also able to deal with large sized non-linear problems with great efficiency. |
