Research Of Data Mining Applied To Fault Diagnosis

Nowadays the equipments become more and more integrated and complicated, so it is of great research meaning to find and forecast the fault in time and to ensure the safe, effective and reliable function of equipments when it's working. As the conventional diagnosis methods still have some vices such as the simplex object, the hard-to-established model, the reliability to subjective experiences and the hard-to-gained rules, sometimes the diversity, complexity and concealment of faults can't be treated successfully. To such vices, some data mining technologies are studied for fault diagnosis in this paper, which achieve fault diagnosis of complex equipment universally, fleetly and out of subjective experience.This thesis mainly presents the application of fault attributes reduction based on rough set theory and the application of fault classification based on decision tree algorithm and association rule mining.An improved algorithm of rough sets is presented by finding attributes of the same classification ability, which avoids the establishment of recognition matrix and the acquirement of core attributes during attributes reducing and improves the reduction efficiency. An improved ID3 algorithm is presented by quering and processing database directly and efficiently using the embedded SQL, which improves the efficiency and feasibility and has a good expansibility with the increasing samples.The Apriori algorithm is improved by database query, which satisfies not only transactional database but also relational database and improves the efficiency of rule acquiring, then it can be properly used in actual diagnosis. The three algorithms are combined to fault diagnosis of complicatied equipments: at first, fault attributes reduction is performed by rough set theory for the predigest of fault database, then the high-efficiency of decision tree and the comprehensive character of association rule mining are combined to achieve rule's mining fleetly and entirely。At last a fault diagnosis system is designed, which achieves a series of function of data acquirement, storage, pretreatment and rules mining, fault matching. It is proved that the system can find the corresponding relations between the attribute and the class of fault,and then the database of diagnosis rules is established and fault matching is executed automatically, which satisfy the actual need of fault diagnosis.