This paper proposedan improved set pair analysis method for intermittent faults with strong concealment and ambiguity. Firstly, by establishing a relationships between the fault type and the symptom of the fault by using the support degree and confidence degree calculated by the association rule, the weight of each fault symptom corresponding to the faulthas been established;The ivalue is calculated by the method based on minimizing the expected loss, and the fuzzy uncertainty between the intermittent fault and the permanent fault is characterized, and the subjectivity of the difference coefficient selection in the set pair analysis is avoided. And then the improved setpair analysis is applied to thediagnosis of intermittent faults of ground air conditioner. The example proves that the method has a good effect on intermittent fault diagnosis.