Abstract: The minimal conflict set cluster is constructed by all the minimal conflict sets. The minimal hitting sets of all the minimal conflict sets are the faults of the system to be diagnosed. Therefore, computing all the minimal hitting sets according to the minimal conflict set cluster is an important step in model based diagnosis. In this paper, a new algorithm is proposed to obtain all the minimal hitting sets by using dynamic maximum element coverage value. The algorithm processes elements in turn according to the descending order of the element coverage value. The search space can be greatly reduced by joining the corresponding heuristic strategy and pruning strategy in the process of solving the hitting sets. Adjacency list is used to store the minimal conflict set cluster in the algorithm. Among the basic storage structures, adjacency list has better space cost than matrix. Besides, the adjacent pointer can quickly find the elements in the minimal conflict set cluster which can be covered by an element. When a hitting set is obtained, the minimal hitting set is obtained by using the minimal hitting set decision rule. Therefore, the algorithm can generate and only generate all the minimal hitting sets. Experimental results show that the proposed algorithm has better computational efficiency than other algorithms.