Abstract: The matching process between the antibody and the antigen is investigated by their near neighbor relationship in the shape space,and the relationship between the pruning threshold and the probabilities of the mismatch and the misdiagnosis is studied for fault diagnosis algorithm based on immune model(aiNet).The fault model boundaries and the containing relationship of these fault models are defined through the rough set theory,and observation criterion and concrete algorithm for self-adaptive adjustment pruning threshold are presented.The simulation result shows that the fault diagnosis algorithm is excellent,the correct diagnosis probabilities of all the fault types and the rate of new fault detection are improved.