基于模型的诊断推理又称为基于深知识的诊断推理，它利用了系统结构和行为等方面的深层知识，克服了传统故障诊断专家系统中过分依赖于专家经验的固有缺陷，从而引起了研究者的广泛兴趣。基于模型的故障诊断通常分两步进行，第一步是与领域相关的冲突识别，第二步是与领域无关的候选产生。本文研究了基于模型故障诊断中的候选产生方法，提出了拟ｈｉｔｔｉｎｇ集的概念，给出一种由拟 ｈｉｔｔｉｎｇ 集求解最小ｈｉｔｔｉｎｇ 集的方法，并对该方法的正确性给予了证明。在此基础上，开发了一种基于拟ｈｉｔｔｉｎｇ 集的候选产生的递推算法。由于在求解过程中仅仅涉及到两个集合之间的运算操作，因而该算法简单实用，极大地减少了诊断的计算量，且程序容易实现。特别是对于复杂的被诊断对象系统，该算法可以明显地提高诊断效率，以满足实时性的要求 Model-based diagnostic reasoning, also known as deep knowledge-based diagnostic reasoning, takes advantage of deep knowledge of system structure and behavior to overcome the inherent flaws in traditional fault diagnosis expert systems that rely excessively on expert experience, causing researchers The widespread interest. Model-based fault diagnosis is usually carried out in two steps. The first step is domain-related conflict identification and the second step is domain-independent candidate generation. This paper studies the method of candidate generation based on model fault diagnosis, proposes the concept of quasi hitting set, gives a method of solving minimal hitting set by quasi hitting set, and proves the correctness of the method. On this basis, a recursive algorithm based on candidate hitting set was developed. Because the solution involves only the operation between the two sets of operations, the algorithm is simple and practical, which greatly reduces the computational complexity of the diagnosis, and the program is easy to implement. Especially for complex systems of diagnosed objects, the algorithm can significantly improve the diagnostic efficiency to meet the real-time requirements