A machine-learning method for deriving state-dependent fragility curves of existing steel moment frames with masonry infills

•A machine learning-based framework for deriving state-dependent fragility curves for existing steel moment frames was presented.•Feedforward neural network classification models were utilised to predict the damage state of existing steel frames with respect to prescribed limit states.•The influence of masonry infills on the seismic response of existing steel moment frame was considered.•Four Pre-Northridge steel frames covering low- and med-rise buildings designed for low and high seismicity were adopted to demonstrated the proposed framework.