Machine Learning Based Chi-Square Improved Binary Cuckoo Search Algorithm for Condition Monitoring System in IIoT

Industrial Internet of Things (IIoT) provides reliable solutions for real time monitoring of the organizations by notifying them before any critical failure occurs. In order to detect the faulty bearings before it reaches the critical level during the rotation, an IIoT based Condition Monitoring System (CMS) is proposed in this research. Initially, the data from the Case Western Reserve University (CWRU) is collected and pre-processed using Normalization to improve the quality of the data. The relevant features are then extracted using vibration analysis techniques. To remove irrelevant features, a Machine Learning (ML) based Chi-Square (CS) Improved Binary Cuckoo Search Algorithm (CS-IBCSA) is proposed in this research. The optimal features obtained through the Chi-square IBCSA are fed to the Support Vector Machine (SVM) for accurate fault detection of bearings. The aim of this framework is to provide a reliable CMS that is capable of Predictive Maintenance (PdM) and identifying faults in real time applications. The proposed SVM-CS-IBCSA achieved an accurate detection of fault bearings at 99.56%, precision of 98.82%, and f1-score of 98.14% when compared to the existing methods AdaBoost-Grid Search Cross Validation (AdaBoost-GSCV) and coach and player learned neural network - cuckoo search (CP-LNN-CS).