一种并行工程约束分解方法

在并行工程产品开发过程中，往往按照问题的结构特点将较大规模的问题分解成一些子问题，并希望通过求解子问题来获得原问题的解。实际中，分解得到的子问题之间往往不是完全独立的，一般的简单分解方法只能有限地降低求解难度和简化问题规模。如何进一步分解各个子问题间的关系，使各个子问题的设计结果不但满足原问题的总体要求而且还能由此获得优化的总体设计结果是一个重要问题。该文给出了分解的意义，提出了基于约束的优化分解方法。

A Concurrent Engineering Constraint Decomposition Method

In Concurrent Engineering, complex problems involving complex constraints are often met in the product development process. It is a general method that decomposing a big problem into some small sub problems according to its structure characteristics and then obtaining the problem's solution through resolving the sub problems. Actually, the sub problems are not independent each other, so that general decomposition methods can only solve the problem partially. It is important to further decompose the relationship among the sub problems so that optimal solutions can be got through resolving the sub problems separately. To minimize couplings among sub problems, couplings will be further decomposed during design process. New constraints are introduced and constraints on multiple variables are converted to constraints on each individual variable (a range of values). This kind of decomposition is called constraint decomposition. Constraint decomposition is necessary especially in conceptual design period. This paper gives the meaning of constraint decomposition and proposes a constraint based optimal decomposition method. Constraint decomposition reduces the size of a big problem and lowers the degree of difficulty in searching for solutions. To facilitate constraint decomposition, the concepts of strong satisfaction and strong satisfaction variables are proposed, and the upper level and lower level of variables are defined. By our proposed constraint decomposition method, firstly, value ranges of variables in the upper level are derived by bottom up approach. Then by limiting value ranges of variables in the lower level by top down approach, strong satisfaction is guaranteed for the lower level variables and Pareto solutions for the problem are achieved. The bottom up and top down modifications for constraints can guarantee strong satisfaction and Pareto solutions for the problem.This constraint decomposition method is effective for propagating upper level function related variables to lower level design variables, and at the same time it guarantees to satisfy the overall requirements by considering design variables alone. Since variables in the upper level account for the entire system, relative constraints and autonomy between design teams can be taken into account during the information propagation. This will reduce some unnecessary information exchange among design teams and facilitate coordination during product development process.