Novel Response Relation Method for Sensor Data Analysis of Complex Engineering Systems

Abstract

Technology of Internet of Things (IoT) offers extensive applications for industrial productivity and safety improvement. Advanced miniature sensors are available for monitoring multiple process parameters in a complex industrial or an engineering system. An industrial plant's overall operational status is captured using a network of sensors and stored on a cloud storage platform, where it is evaluated using the machine learning algorithms to produce valuable insights. Finding the correlation among these sensor variables is essential before feeding the same to machine learning algorithms. The present study proposes a novel approach to choose a few critical sensors out of numerous sensors based on the Response Relationship methodology. The Response Relationship method enables the system to be fully autonomous and helps find the interrelation among variables. The Response Relationship among variables is quantified and used for calculating the Remaining Useful Life of a complex engineering system. The proposed methodology is also applied to binary and multi-class classification to demonstrate the efficiency of the Response Relationship method. The results obtained are compared with standard methods of prediction and classification in terms of suitable metrics.