A homeostatic approach to adaptive ambient control in smart factories

Chong, Shao Yang (2024) A homeostatic approach to adaptive ambient control in smart factories. Final Year Project, UTAR.

Preview

PDF Download (18Mb) | Preview

Abstract

This project aims to develop a system using Artificial Intelligence (AI) and Internet of Things (IoT) to regulate and adjust environmental conditions in smart factories, focusing on the production of feed pellets, particularly during the drying stage. This paper addresses this gap by exploring existing technologies within the framework of Industrial Revolution 4.0. We've examined various factors affecting feed pellet quality by examining moisture level and climate setting using near-infrared (NIR), microwave, and capacitance sensors to detect ambient conditions. These data are visualized on a Grafana Dashboard for real-time data monitoring, a predictive model for dryer setup, and a computer vision system for quality control assessment. Testing will be conducted in a simulated environment to achieve a minimal working product (MVP). Lastly, the project has successfully implemented both a LSTM predictive modelling and a Siamese Network with CNN as base model for computer vision task. The LSTM predictive model is used to identify the optimal ambient setting to produce the highest quality pellet based on parametric optimisation. It is then integrated with the machine to manipulate and scale the environment conditions to best fit the pellet requirements. With the implemented Siamese Network with CNN has also successfully classify the pellets quality based on the appearance and colour of the given pellet.

Actions (login required)