Simultaneous-fault diagnosis considering time series with a deep learning transformer architecture for air handling units

•A simultaneous fault diagnosis model is developed with a transformer architecture.•Time-series data are utilized to diagnose simultaneous faults in the early stage.•A novel multi-head attention mechanism attending time steps and features is adopted.•The proposed model is verified by experiments on an on-site air handling unit.•The model outperforms GRU model with less training time under the same accuracy.