Automatic performance degradation prediction by use of field data with noise

This paper proposes an automatic performance degradation prediction approach for equipment by use of field data with noise. In this method, the auto-regression model is employed for making performance degradation prediction by means of field data in order to fit the change process of equipment. The order determination for auto-regression model is according to the FSIC criteria which not only consider the covariance of prediction error by use of different auto-regression estimation algorithms, but also consider the covariance of model parameter estimation. It therefore can modify the over fitness quite well. In the parameter identification for auto-regression model, the conventional least square method has been improved. It not only modifies the influence from white noise to the parameter estimation, but also can obtain improved auto-regression model by means of the selection of optimal value. In this research, the field data of the temperature of gas engine are adopted to analyze the performance degradation of gas engine. Through the simulation, the effectiveness of the proposed method has been verified, which also proves that the improved least square method (ILS) has great application value.