Intelligent Fault Diagnosis Based On Rough Sets And Support Vector Machine

With the development of scientific technology, such as control theory, signal processing, artificial intelligence, pattern recognition, fault diagnosis technology is pushed to the direction of intelligence. Many countries have developed this aspect of study and gained a great number of achievements. The main problems of intelligent fault diagnosis are the shortage of typical sample and the knowledge discovery, both of which are severely hampering the development of fault diagnosis. Support vector machine and rough sets theory offer a good method for solving these issues.This paper mainly according to the key problems of support vector machine need to solve in fault diagnosis, carried out a systematic and deeply study in continuous attributes discretization based on rough sets theory, attributes reduction of fault samples, the model construction of fault classifier. The main contents completed as follows:1. Rough sets theory is a data processing method based on discrete data. The paper proposed method of discretization of continuous attributes based on distribution characterization of interval data. It considers the frequency of attribute values in choice of the cut point, Use information quantity as feedback on the basis of consistency within the interval and differences between the intervals, merger the intervals. It makes less split points of attributes, to set the rules more concise.2. A heuristic algorithm for attributes reduction based on discernibility matrix was proposed. The algorithm considered of the mutual information and the distribution of attributes value, which between selected conditional attributes and decision attribute, definite a new measurement parameters of the attribute importance from the viewpoint of information theory, and the parameters was used as the heuristic information of the attributes reduction based on discernibility matrix, obtain ideal reduction results.3. General support vector machine can only solve the problem of two type's classification, but the actual situation is more often the multi-classification issues. The paper studied the implementation of two types and the multi-classification issues of support vector machine, and through simulation analysis comparing the different ways'performance and characteristics.4. In order to improve the efficiency and reliability of fault diagnosis, combine support vector machine and rough sets theory, establishment a fault diagnosis model based on rough sets and support vector machine, application in the diesel engine fault diagnosis, and obtained satisfactory results.