Engineering Approaches to Developing Process

Many data-driven models for fault detection and Smart Manufacturing (SM) applications have been proposed. Unfortunately, only a few models have been deployed for real-time usage and practical applications. This study demonstrates the deployment and practical application of fault detection models using a counter-current, shell and tube heat exchanger and a Smart Manufacturing Innovation Platform (SMIP). Experimental data from laboratory-scale heat exchanger equipment trains several state-of-the-art fault detection models, including fully connected, convolutional, and recurrent neural networks. We also present tutorials on how these models can be deployed for practical real-time applications using the SMIP. We evaluate and compare these models' fault detection and computational performances during deployment and real-time usage. Although some models require more time for inferences, our comparative study shows that these models give similar performances. We provide an easily customizable pipeline for SM applications and deployments through these implementations.

Keywords: Fault Detection, Smart Manufacturing and Smart Manufacturing Innovation Platform.