A simulation and machine learning informed diagnosis of the severe accidents

•A simulation and machine learning informed model (SMLIM) is proposed.•An LSTM network model is developed for forecasting and regression analysis.•The model has four feature variables with dry-well pressure as the target variable.•The accident at the Fukushima Daiichi Nuclear Power Plant (FDNPP) unit 1 is chosen.•Training and test data are generated by MELCOR simulations.•Measured data at the FDNPP is analyzed with SMLIM.•SMLIM is shown to be useful for the diagnosis of the accident.