Mobile machine anomaly detection in container handling operations

Abstract

The objective of this Master’s thesis is to research Deep Learning (DL) based anomaly detection methods for unlabeled time series data in container handling operations. Detected anomalies can be the sign of a defect in the container handling equipment. Predictive maintenance aims to detect and prevent failures in industrial equipment by analyzing Key Performance Indicator (KPI) data and identifying anomalies that indicate potential issues or malfunctions.

This Master’s thesis examines forecasting based deep learning methods with convolutional and long short-term memory layers for detecting anomalies from onboard control system data. The final goal of the work is to obtain a method to identify signs of abnormal steering behavior.

The thesis presents the background of the problem in the introduction with the research questions. The thesis also covers the test setup of the experiments along with the training and testing of the selected deep learning method.

The results show that it is possible to detect anomalies by using the selected method.