As shown in Fig. \ref{970718}, the model development starts with the data collection process. Healthy operational data in August 2016 are collected, which consist of 267804 samples. The samples are divided into a training dataset (50%), testing dataset (30%) and validation dataset (20%).