Detection and Classification of Faults in Dynamically Controlled Systems using Gaussian Process Regression with Hidden Markov Model

Nowadays, cyber security has evolved over decades in response to increasing complexity and frequency of cyber threats. Traditional protection methods are related to firewalls, anti-virus programs, and intrusion detection systems, rely on pattern matching to counter known attacks. However, these methods are limited in detecting unknown threats, highlighting the need for advanced cyber security measures. In addressing emerging and unknown threats, emphasizing the need for more proactive and adaptive security approaches. To overcome these drawbacks, a Gaussian Process Regression with Hidden Markov Model (GPR-HMM) is proposed for fault detection and classification of faults in dynamically controlled systems using Electrical Fault Detection and Classification (EFDCD) dataset. Initially, for preprocessing, data cleaning, normalization, and scaling are performed using Short-Time Fourier Transform (STFT). Then, for feature extraction Multi Scale Entropy (MSE) is considered to quantify the complexity of time series data at multiple scales. At last, GPR-HMM is employed as classifier for detection and classification of faults. From the results, it clearly shows that the proposed GPR-HMM obtained accuracy value in the context faults detection when compared to existing Data Entry and Office Automation (DEOA).