In the previous paper, we presented a model-based diagnosissystem augmented with domain-dependent constraints. Compared to previous approaches this framework has the advantage of making valuable constraints explicit to reduce the candidate diagnosis and embedding constraints seamlessly into diagnosis generation. We show that, the superset of the faulty components in a minimal constrained diagnosis cannot always compose a constrained diagnosis, namely minimal diagnosis assumption is not correct for constrained diagnosis system. Then a natural question is how to characterize constrained diagnosis space completely. Argument-based systems analyze defeasible in terms of the interactions between arguments for alternative conclusions. Not only does the notion of an argument naturally point at possible proof theories, but also do notions like argument, attack and defeat have natural counterparts in the way people think. Inspired by these, we draw an analog between the process of getting model-based constrained diagnosis and argumentation: some components might be charged of fault by other components, and they can defend themselves. We show that any constrained diagnosis is the cogent extension of a minimal constrained diagnosis within our argumentation framework. In this way the constrained diagnosis space is characterized completely by all minimal constrained diagnosis and their cogent extension. Compared with other methods that address the problem of minimal diagnosis assumption, our method has the advantage in computation because it is easy to check whether an extension is cogent or not.%在含约束的极小诊断中故障部件的每个超集并非总能构成一个含约束的诊断.作者把基于模型的含约束的诊断比喻成一个辩论过程：一些部件可以指责某些部件出了故障；反过来这些部件也可以为自己辩护.在作者所建立的辩论框架下含约束的诊断空间得到了紧致的刻画.与相关工作相比，这种方法具有计算上的优势，它也可以解决极小诊断假设问题.
展开▼
