Health monitoring of a TriGen plant: a Big Data proposal

Abstract

Nowadays, energy efficiency is becoming a critical factor in factories all over the world. Thanks to proper and timely monitoring of the operation and performance of the factories, remarkable energy savings can be obtained. This project aims to perform health monitoring in large factories or corporations by means of data-driven techniques. Specifically several machine learning models will be developed to perform fault detection. This monitoring includes fault detection and fault prediction of any of the components of the factory. This project relies upon previous work done during an internship in the National University of Ireland at Galway in which the log files of the Boston Scientific Corporation's (BSC) tri-generation plant were studied. This work contains a Big Data architecture's proposal to store all the data from both the logs of the tri-generation plant and the simulation data obtained for the absorption chiller subsystem within the tri-generation plant (due to the lack of discriminative information about faulty behaviour in the real data), and a conceptual data model to describe the relationships, entities and attributes of that data. The Machine Learning models have been tested successfully in the absorption chiller subsystem, providing promising results.