Approach to diagnostics and prognostics based on evolutionary feature parameters

Time series analysis methods are used to prognostics the feature parameters evolution that can indicate the system's fault. Considering the uncertainty of feature parameters, a method for fault diagnostics and prognostics are presented. The relationship between feature parameter and fault criterion is first discussed. Then, the fault criterion is summarized to two types: general strength/fault threshold and space distribution of feature parameters for fault mode. The evolution of feature parameters along product lifetime is a stochastic process under the influence of product work conditions and environment conditions. Based on monitoring data of feature parameters, the time series analysis methods can be used to prognostics the future conditions of systems. A quadric exponential smoothing model is presented in a case study. For a certain time, the conditions of systems can be diagnosed according to the quantificational relationship between feature parameter and fault criterion. Based on the consideration of feature parameters distribution, fault probability and fault index are two kinds of results that can use to assist the decision for maintenance.