“…However, the neural network is often limited by the distribution and amount of training data (Lee et al, 2004), the rough set has a crude way of knowledge expression and reasoning strategy, and the expert systems cannot be applied universally owing to the poor robustness (Tony and Liang, 2011). In recent research, hidden Markov model (HMM) (Xie et al, 2014; Zong et al, 2014) has been proven effective diagnosis model to address the problem of system fault diagnosis in different industrial processes, such as the incipient faults of gear (Kang and Zhang, 2011), the chemical precess diagnosis (Li et al, 2014), the bearing diagnosis (Xu et al, 2015), and so forth. Specifically, the HMM demonstrates its rationality and practicability on its model structure.…”