Anomaly Detection by Using Artificial Intelligence

Chong, Qi Tat (2019) Anomaly Detection by Using Artificial Intelligence. Final Year Project (Bachelor), Tunku Abdul Rahman University College.

Text ChongQiTat.pdf Restricted to Registered users only Download (2MB)

Abstract

Nowadays, anomaly detection is a well-known and widely utilized computational technique used to detect novelty, anomalous and unusual behavior for a statistical problem. It possesses the capability to capture error by pre-indicating a threshold based on the condition, standard and requirement of the system. According to various researches, the technique of anomaly detection is used in numerous field which includes medical illness detection, detecting bank fraud, and network system log monitoring and telecommunication fraud detection to analyze the pattern and behavior of the data in order to detect the occurrence of novelties. Similar as the application of the other examples from the research, the anomaly detection technique is also suitable to be used on part failure detection of machine in the production line of manufacturing company. In this project, there is a collaboration with a global leading electronic manufacturing service providing company which used machine with high precision to operate the component mounting and assembling process. It is a fact that not all machines are perfect and there is an occurrence of certain amount of error in the machines. Hence, the anomaly detection technique is suggested to be applied on this situation to enhance the situation in the production line. The main objective of the system development is to ensure the detection of faulty events that will occur in the machines. The secondary objective of the research is to create an anomaly detection system to forecast the faulty event occurrence via machine learning of AI while the other objective is to smoothen the operation of the production line by reducing the time to detect error for trouble shooting of the machine. This anomaly detection system will be developed based on the method of machine learning. In this research project, raw datasets will be obtained from reliable source as the data to be analyze will be the data log file of the machine production line of the company. A few algorithms will be written in codes to learn the pattern of the training data and will be able to determine and identify the occurrence of anomalous events when more new data is fed.