给定地震场景下的随机地震动降维模拟?

建立了一种可根据地震场景预测和模拟随机地震动加速度过程的降维模型。首先，挑选了1 766 条实测强震记录，根据断层类型和场地类别进行了分组，并识别了各组地震动的演变功率谱参数；然后，基于地震场景参数和演变功率谱参数，训练了两者的高斯过程回归模型（GPRM），并采用K-fold 交叉验证法，验证GPRM 预测的有效性和精确性；最后，基于非平稳随机过程的谱表示法，通过引入随机函数的降维思想，实现了在给定地震场景下的随机地震动降维模拟。数值算例表明，预测的样本在频谱、峰值、强震持时等方面均与实测记录保持一致，体现了本文方法良好的工程适用性。这为地震动目标地区提供较为合理的人工地震动数据以及工程结构的随机地震反应分析与可靠性评价奠定了基础。

Dimension-reduction Simulation of Stochastic Ground Motion underPredefined Earthquake Scenarios

Based on the earthquake scenarios, a dimension-reduction model capable of predicting andsimulating the stochastic ground motion acceleration process was developed. Firstly, 1766 strong mo-tion records were selected and grouped according to fault types and site classification. Parameters ofevolutionary power spectrum (EPS) for each group were identified. Secondly, based on the parametersfor earthquake scenarios and the EPS, a Gaussian process regression model (GPRM) was trained.Meanwhile, the K-fold cross-validation method was adopted to verify its prediction effectiveness andaccuracy. Finally, using the spectral representation method of non-stationary random processes and in-corporating the concept of dimension reduction for random functions, the dimension reduction simula-tion of stochastic ground motion was achieved under predefined earthquake scenarios. Numerical examples showed that the predicted samples were consistent with the measured records in terms of fre-quency spectrum, peak values, and duration of strong motion, which verified the suitability of the pro-posed methodology in engineering applications. The research provides reasonable artificial ground mo-tion data for target areas, and lays a foundation for random seismic response analysis and reliabilityevaluations of engineering structures.