Time-Series Identification of Fatigue Strain Data Using Decomposition Method.

This paper presents the time-series identification a variable-amplitude (VA) strain signal on lower arm suspension component in terms of time-series component analysing and correlate to the fatigue damage properties. The identification technique was used is called classical decomposition method, to classify the strain data into trend, cyclical, seasonal and irregular components. The time history plot of a study case showed the fatigue data contains high and low amplitude events and has resulted the highest amplitude for a pavé, highway and campus are 224 µε, 321 µε and 619 µε, respectively. The trend pattern of a fatigue strain data is a nonstationary series in variance and mean, where a campus data produced highest slope of 31.2Χ10^-4 compared to the others. By observing the cyclic movement of the moving average plot, the fatigue strain data contained expansion, contraction and random background. The autocorrelation plot is weak in identifying seasonal pattern, but the autocorrelation coefficient, r1 values are statistically significant and show a positive serial correlation. Based on residual plot in irregular analysis, the residuals pattern is considered random. As to correlate the fatigue characteristic and time-series component, it was found a campus data produced highest value of fatigue damage. This study discovered a slope of a linear trend pattern could be affected to the fatigue damage properties because the fatigue strain data are nonstationary, VA time-series data and have a random background. Thus, the findings of these characteristics are expected for a nonstationary signal. [ABSTRACT FROM AUTHOR]