Non-stationary modeling of rail vehicle reliability: A comparative study

The non-stationary time series modeling of electric rail vehicle reliability is explored using retrospective series of Times Between Failures (TBF´s). Three modeling methods are critically assessed and compared: A Time-dependent ARMA (TARMA) method, the Recursive ARMA (RARMA) method, and the Integrated ARMA (ARIMA) method. The results of the study indicate that while all three methods are useful in modeling and predicting the TBF series, as well as in providing insight and revealing the presence of lightly damped cycles, the TARMA method offers significantly improved accuracy and analysis capabilities.