Modeling and simulation of three-component ground motion intensity envelope function based on conditional generative adversarial network

Artificially simulated ground motions are critical to seismic analysis of engineered structures. The ground motion intensity envelope function (IEF) defines the non-stationary characteristics and controls its duration. The three-stage intensity envelope model has been widely used in engineering applications. However, its limitations in terms of consistency of trends in horizontal and vertical changes, as well as the complexity of parameterization, hinder its further application. This study proposes a three-component ground motion intensity envelope function deep generative model (IEF-DGM) based on conditional generative adversarial network (CGAN), capable of producing high-quality IEF by considering five key condition labels: moment magnitude (M_W), Joyner-Boore distance (R_JB), shear wave velocity (V_S30), and significant durations (D_S5-75, D_S5-95). Five evaluation indicators are introduced to assess the model’s predictive accuracy. Results indicate that the proposed model effectively captures the stochastic and temporal non-stationarity of the IEF and accurately predicts the vertical component, a rarely studied area. Additionally, the model is used to simulate three-component artificial ground motion records, which closely matches the target response spectra, validating its accuracy and applicability. In conclusion, the proposed ground motion IEF-DGM provides a novel approach to simulation and prediction of ground motions with significant implications.