Master of Engineering Science
Electrical and Computer Engineering
The smart manufacturing evolution enables financial and operational improvements across the manufacturing industry. However, smart manufacturing encompasses complex, interconnected systems which can fail at any time. To address this challenge, a novel, two-part anomaly detection system for robotic processes, with an application focus on robotic surface finishing, is presented. The first part proposes an unsupervised Attention-based Convolutional Long Short-Term Memory Autoencoder with Dynamic Thresholding (ACLAE-DT) framework for anomaly detection and diagnosis in multivariate time series of robotic surface finishing components. The second part proposes a deep residual Convolutional Neural Network-based triplet model for anomaly detection in the produced robotic surface finishes, where defective training samples are synthesized exclusively from non-defective samples via random erasing data augmentation. Evaluation results demonstrate the performance strength of ACLAE-DT over state-of-the-art time series methods and the performance strength of the proposed triplet model in detecting anomalies for known and novel surfaces.
Summary for Lay Audience
Manufacturing is a critical part of any industry and is considered the essence of the secondary sector of the economy. To further increase revenue and reduce operational risks in manufacturing, smart manufacturing was born. Smart manufacturing aims to add intelligence and autonomy to manufacturing by utilizing collaborative robotics, interconnected sensors, and a range of advanced technologies in an effort to optimize the entire manufacturing process. However, the complex network of systems utilized in smart manufacturing increase the possibility of system failures, which can cause disruptive difficulties for the manufacturer. As a result, detecting anomalies, which are irregular observations that differ from the norm of the data, accurately and rapidly across manufacturing processes is vital to enabling the operators in solving the underlying issues.
In this work, a novel, two-part anomaly detection system for robotic processes, with an application focus on robotic surface finishing, is presented. The entire system leverages Deep Learning (DL), which is a collection of theories and methods that enable computers to learn from data and make predictions without being explicitly programmed. More specifically, DL allows complex features to be learned from the data autonomously without the need for manual feature development, thus creating an end-to-end learning structure with minimal human interference. The first part of the presented system proposes a DL-based framework for anomaly detection and diagnosis in multivariate time series of robotic surface finishing components. Moreover, the second part proposes a DL-based computer vision framework for anomaly detection in the produced robotic surface finishes. The entire system eliminates the use of any real-life anomalous samples during training, as there is often an abundance of non-anomalous data and a relatively small or no amount of anomalous data in well-optimized processes. Evaluation results, conducted on real-life manufacturing data sets, demonstrate the performance strengths of the proposed system over existing methods in detecting anomalies in robotic time series processes and in industrial surface images.
Tayeh, Tareq, "An Anomaly Detection System for Smart Manufacturing Using Deep Learning" (2021). Electronic Thesis and Dissertation Repository. 7961.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.