Theoretical testing of an empirical mode decomposition damage-detection approach using a spatial vehicle-bridge interaction model

Empirical mode decomposition (EMD) is used to detect and locate damage in a bridge using its acceleration response to the crossing of a vehicle. EMD is a technique that converts the measured signal into a number of basic functions that make up the original signal. These functions are obtained purely from the original signal in a sequential procedure, where lower order basic functions contain a range of high frequency components of the signal and higher order basic functions contain the low frequency components. Damage is identified through a distinctive peak in the decomposed signal resulting from applying EMD. Recent studies have shown the potential of this tool to detect single and multiple damages when using the response of a one-dimensional beam model to the crossing of a constant load. In this paper, the technique is further developed using simulations from a quarter-car vehicle-bridge dynamic interaction finite element model. The vehicle model is composed of mass elements, which represent the tyre and body masses, and stiffness and damping elements, which represent tyre and suspension systems. The bridge deck is modelled using plate elements with typical properties found on site and the road profile is generated stochastically from power spectral density functions based on ISO standard guidelines. Different levels of damage are simulated as localised losses of stiffness at random locations within the bridge and a number of longitudinal and transverse locations are used as observation points. The ability of the EMD algorithm to detect damage is analysed for a variety of scenarios including two vehicle configurations (light and heavy), a range of speeds between 5 and 15 m/s, and smooth and rough road surfaces. The influence of the distance from the simulated acceleration points to the damage locations, on the accuracy of the predicted damage, is also discussed.