VIBRATION ANALYSIS AND CONDITION FORECASTING FOR ROTATING MACHINERY USING LOCAL MODELING NEURAL NETWORKS

摘要

In recent years, scientific and technological interest in the predictive maintenance of industrial machinery has steadily risen. This interest is explained by the huge economic savings that can be obtained with an early diagnosis of machinery faults. In this area, vibration analysis has become almost the universal method to assess the state of a machine. Although there are many general known rules that can help to assess the state of a machine, it is still necessary to know and evaluate each machine individually to obtain an accurate diagnosis and forecasting of its future condition. Furthermore, this evaluation must be done by experts in the area. This leads to the necessity of automatizing this process.rnTo build a complete automatic predictive maintenance system based on vibration analysis, a module capable of establishing the baseline state and forecasting the future condition of a machine must be constructed as the base of the overall monitoring system. In this paper, a scheme to use neural networks for this purpose is presented. Also, the results achieved by a local modeling neural network applied to this problem are presented to empirically prove the effectiveness of the proposed method.