Detection of Plastic Hinges in Inelastic Nonlinear Steel Frames Using Deep Learning

This study presents a numerical method based on deep learning techniques for detecting plastic hinges in inelastic nonlinear steel frames under static loadings. In which, a dataset randomly generated by an advanced analysis method is utilized for the training and testing processes of deep learning (DL). The geometric nonlinearity including the P-δ and P-Δ effects is considered by the stability function, whilst the material one is taken account of Column Research Council (CRC) tangent modulus concept and Orbison yield surface based on the refined plastic hinge approach. Accordingly, the plastic hinge information can be easily traced via the Orbison yield surface. A benchmark two-story frame is exhibited to demonstrate the reliability of the proposed methodology. Obtained outcomes indicate that the plastic hinges in inelastic nonlinear steel frames can be directly diagnosed by DL techniques such as Deep neuron network (DNN) and Extreme Gradient Boosting (XGBoost) without using incremental-iterative algorithms. All investigated cases are programmed by Python.