Ai for Detecting Variations in the Oee Data Reception Rate in the Manufacturing Industry

Abstract

Industry 4.0 provides a favourable environment for processing and analysing massive amounts of data by implementing and integrating innovative Internet of Things (IoT), artificial intelligence (AI) and machine learning (ML) techniques. Their use permits to identify opportunities for improvement and increases the efficiency of the equipment and overall manufacturing process, which industrial companies strive for. One of the most common measures of efficiency is the Overall Equipment Effectiveness (OEE), which quantifies the efficiency of a manufacturing system or piece of equipment. It considers factors such as availability, performance, and quality of the equipment. This work presents a methodology to detect, in a robotic antenna manufacturing line, variations in the OEE data reception rate leveraging IoT and ML techniques. The line consists of a set of robotic cells and machines that sequentially execute each one of the production processes required for the manufacture of antennas. The study aims at identifying patterns in the reception of data from machines, propagating alerts when data that do not follow these patterns are found. Firstly, a dataset containing the OEE historical information for each machine is created. Secondly, the dataset is split into train and test sets. Lastly, different ML algorithms are trained to create a model capable of detecting unusual variations in the OEE data reception rate, and later on compared. Fluctuations in the rate at which data is received could indicate issues on the production line, and detecting these anomalies automatically enables proactive maintenance to be performed without human involvement. The analysis results show that more accurate results are obtained by using supervised ML classification techniques instead of unsupervised learning for anomaly detection.