Steel plates fault diagnosis on the basis of support vector machines

Fault diagnosis is always a big concern in industry production. As industrial technology has developed a lot, new fault diagnosis methods are needed to distinguish faults with only fine distinctions. The higher quality a production is required to have, the better fault diagnosis method the factories should apply. A fault diagnosis method based on modified Support Vector Machines (SVMs) is shown in this paper. With this method, dimension of samples is effectively reduced by recursive feature elimination (RFE) algorithm, and computing time is saved at the same time. Besides, classification accuracy is improved by parameter optimizing and sample size balancing strategy. A faults dataset of steel plates is taken as a practical case. And SVMs that are modified by different algorithms are utilized to complete fault diagnosis. This combined measure shows its superiority in sorting common faults of steel plates over original SVMs. Some essential procedures in model developing, such as normalization and cross validation, are also referred to.