Data Filtering in Vehicle–Bridge Impact Simulations: Evaluation of Different Force Filtering Methods and Recommendations

Abstract The impact force time history obtained from a crash test or high-fidelity simulation is usually contaminated by numerical noise. The noise signal usually appears as a series of short duration, high amplitude spikes in the time domain that complicate the interpretation of the test or simulation results. The moving average method has been widely used to remove noise from the original data by smoothing the force time history over a moving time window of fixed size, typically 25 or 50 ms. Many researchers use the peak impact force from the smoothed signal as the equivalent static force for structural design purposes. However, there is no theoretical basis for this practice. High fidelity finite-element simulations are used to investigate the adequacy of this empirical assumption by comparing the dynamic response of typical bridge piers and barriers to their responses under the equivalent static force obtained by the moving window method. It is shown that this practice is unconservative for the cases considered. The empirical mode decomposition (EMD) method is used to decompose the original impact force data into noisy modes and useful dynamic signals. A typical impact force signal consists of high-frequency noise, pulselike signals, and monotonic loading components, the last two of which should be considered in the impact design of bridge structures.