MULTIVARIATE TIME SERIES DEEP NEURAL NETWORK PREDICTION OF SEISMIC HYSTERETIC PERFORMANCE OF REINFORCED CONCRETE SHEAR WALLS

The reinforced concrete shear wall structure has superior seismic performance and reasonable cost, and is widely used in regions with high seismic intensity. Accurately predicting the hysteretic performance and skeleton curve of shear walls directly determines the accuracy and reliability of structural design and analysis. A deep learning-based prediction method for the hysteretic performance of shear wall structures is proposed, which can directly predict the bearing capacity index according to the basic design parameters of the structure (such as material properties, geometric dimensions, load conditions, etc.). The hysteresis curves of one group of structures are predicted by the hysteresis tests of three groups of shear walls. The results show that the deep learning method has high prediction accuracy by comparing the characteristics of the time domain. Compared with the finite element simulation results, deep learning can quickly predict the hysteresis curve only by inputting different parameters, which has the advantage of high computational efficiency; while the finite element simulation requires geometric modeling, constitutive model selection, material property input and load case definition, and is time-consuming and labor-intensive compared with deep learning.