Abstract:
Durability tests are important to ensure the safety and reliability of a ground vehicle and involve frequently driving a vehicle through a series of events that simulate different road conditions or obstacles encountered during actual driving. Since durability tests are costly in-terms of time and money, accelerated durability lab tests can be used to spot failures before actual road tests. Signals of different events of the actual durability road tests generate three continuous time series data, that can be used to conduct accelerated durability lab tests. The actual analysis of these time series is very challenging because they are (i) of high frequency (ii) very noisy and (iii) inconsistent.
The purpose of this study was to identify the patterns of signals from the noisy and inconsistent time series data collected from the field tests. The Box-Jenkins methodology was used to identify models corresponding to different events. Due to complex structures of the real data, ARMA modelling was considered after testing stationarity of the given time series. While the time series data in vertical direction was used to identify the first three events, the time series in vertical, longitudinal and lateral directions were used to identify other four events.
