| Summary: | 碩士 === 國立臺灣大學 === 電機工程學研究所 === 88 === The competitiveness of semiconductor foundry manufacturing is rooted on both real and virtual products. The latter includes information, business processes, and services for customers. Recently, For meeting the highly competitive market requirements, the idea of “virtual fab” (VF) has been proposed by the semi-conductor industry. The service concept is introduced to manufacturing and the business strategy becomes to provide quality products as well as services. To achieve competitiveness, it is essential to develop manufacturing service mechanism and enablers.
In this thesis, the creation of ATP (Available to Promise) service is used to convey our general ideas of systematic and automatic service creation. An ATP service is essentially an order delivery commitment service. It takes customer orders as inputs and bind information resources of necessary procedures. Then ATP evaluates the capacity situation and calculates due dates by using a resource reservation algorithm. The marketing department finally confirms the due dates with individual customers to complete the ATP service. A good ATP service must provide customers with quick response, transparent business process and credible delivery schedule.
A Binding mechanism is adopted and extended for creating manufacturing services. The binding model consists of four layers: physical and logical layer, binding interface base, programming and operation model, and manufacturing service. Binding mechanism consists of four service creation procedures. The first is "name and resource mapping", which converts an external order into internal work orders. The second step, “business and operation process binding,” transfers internal work orders into work and business process flows. The third step, “resource reservation binding,” reserves resources such as machine capacities, human resources, and raw materials of the factory and calculates possible due dates. Finally, “service management binding” binds the service plan to make it a managed service.
The “resource reservation binding” step requires an efficient decision engine. In this thesis, resource reservation algorithms are developed to allocate resource efficiently. There are four functional blocks for resource reservation: lot allocation, lot capacity requirement calculation, residual capacity calculation, and capacity allocation. The “lot allocation” allocates lots belonging to firm orders with the same product specification. The ”capacity allocation” functions allocates the capacity and evaluates the due date and wafer start date by first using a PULL procedure to see if a lot can be delivered on the customer demanded due date. If not due to insufficient capacity, a PUSH procedure estimates the delivery date by starting the lot as early as possible.
ATP service creation process depends on both order and fab resource management, including software, hardware, human resource and business process. Our binding model facilitates such service creation. It exploits object-oriented (O.O.) abstraction of a fab and is service process driven. Input data and/or information are mostly available. In this thesis, both the ATP binding service creation process and the ATP model are implemented using a Computer-Aided Software Engineering tool -- Rational Rose 98. The tool generates the object oriented C ++ code of a model expressed in UML (Unified Modeling Language) format. The resource reservation algorithm is coded in visual C++ program and tested. Numerical experimentation results demonstrate the ideas and the practicability of binding mechanism for ATP service creation.
|
|---|
