Investigation of Structural Response under Human-Induced Excitations Using Noise-Assisted and Adaptively Transformed Multivariate Empirical Mode Decomposition

Abstract

Flexible structures such as footbridges, stadiums, and large slabs are prone to narrowband excitations generated by occupants. Both the excitation and response are of a nonstationary nature that is often not suitable for conventional time-based and frequency-based analyses. In this regard, an investigation of a hybrid technique that makes use of multivariate empirical mode decomposition (MEMD) for the human–structure interaction problem is presented. The method uses a noise-assisted version of the MEMD along with an adaptive projection algorithm to extract monocomponent intrinsic mode functions (IMFs), to reduce mode misalignment and mixing, and to account for power imbalances among channels as well as benefiting from correlations at an intrinsic level. To overcome persistent mode mixing, especially for closely spaced modes, a method called complete EMD with adaptive noise (CEEMDAN) is used. The time-frequency representation of the response is provided by incorporating the Hilbert-Huang spectrum (HHS). Using the HHS may reveal the instantaneous changes in the energy, frequency, phase, and amplitude in dynamic systems. An extensive explanation and the conditions that are required for this observation can be found in previous work. The performance of the application along with its limitations are reported in light of two case studies conducted on a laboratory grandstand and on a footbridge.