This paper proposes a concept of an intelligent optimal design approach, which is organized by a kind of compound knowledge model. The compound knowledge consists of modularized quantitative knowledge, inclusive experience knowledge and case-based sample knowledge. By using this compound knowledge model, the abundant mathematical information of mathematical programming and the symbolic knowledge of artificial intelligence can be united together in this model. The intelligent optimal design model based on such a compound knowledge and the automatically generated decomposition principals based on it are also presented. This paper realized the object-orientated based decomposition of optimal model and integration of compound knowledge. Practically, our approach has been applied to the production planning, process schedule and optimization of production process of a refining & chemical work and a great profit was achieved. Specially, the methods and principles presented in this paper are adaptable not only to continuous process industry, but also to discrete manufacturing one.