Abstract: Extraction of simple and effective decision rules from the numerous data containing inconsistent and redundant in formation is one of the most important issues needed to be solved in fault diagnosis.Firstly,According to complete information system and incomplete information system,the corresponding fusion algorithms are shown,which provide an effective method to deal with the overloading data of sensors and information fusion for incomplete sensors.Secondly,a multi-sensor information fusion fault diagnosis model based on rough set theory is presented.From original fault data containing inconsistent and redundant information,the fault symptom attributes are reduced using the method based on attribute significance.Then,a set of maximal generalized decision rules are generated by using a proposed value reduction algorithm,and a decision rule base for fault diagnosis is established.Finally,when the presented model is applied for fault diagnosis,the discretized fault symptom attributes are matched with the rules in the decision rule base,and the returned diagnostic decision rules are evaluated by using certainty factor,coverage factor and support factor,then diagnostic results are obtained.A proposed diagnostic example proves the method is feasible and available.