Abstract:Real-time fault detection via sensor signal analyses is widely performed in electromechanical systems to prevent sudden stops and the resulting losses, which requires a sufficient amount of fault data. Because it is difficult to obtain sufficient fault data with actual physical systems, fault detection is usually conducted using unsupervised detection models with datasets collected under health operations. Supervised models with few fault datasets of limited types have been developed, often with sophisticatedly augmented fault datasets. However, actual faults occur cumulatively with machine use, and the information contained in the actual data is very valuable. Therefore, this study proposes a feedback loop-based fault detection method that can be adjusted depending on the fault datasets during operation. Unsupervised fault detection is conducted using a long short-term memory autoencoder-based deep learning model and statistical process control chart with a derived anomaly score before observing any fault state. After obtaining even a small amount of fault datasets, fully connected layers are added for supervised classification, such as transfer learning techniques, to determine whether the current state is healthy or faulty. The performance of the proposed continuous learning method for fault detection is verified using datasets from a fused filament fabrication three-dimensional printer under several scenarios: no fault, one type of fault, and two types of faults. The detection performance improves as informative feedback is provided by the newly added layers. It was observed that even small fault datasets have the potential to improve the performance of trained unsupervised fault detection model. Since it is possible to detect fault occurrence from the initial state of the product life cycle, and the performance would be automatically updated as the printer is used, it is expected that it can be installed in a commercial three-dimensional printer and used directly for maintenance, without analyzing other machine's datasets.

Fault Classification of 3D-Printing Operations Using Different Types of Machine and Deep Learning Techniques

A Low-Cost Multi-Sensor Data Acquisition System for Fault Detection in Fused Deposition Modelling

Machine Learning Based Fault Anticipation for 3D Printing*

Defect Analysis of 3D Printed Cylinder Object Using Transfer Learning Approaches

Visual Detection of Surface Defects Based on Self-Feature Comparison in Robot 3-D Printing

Multi-Axis 3D Printing Defect Detecting by Machine Vision with Convolutional Neural Networks

A CNN-Based Adaptive Surface Monitoring System for Fused Deposition Modeling

Defect Localization Using Region of Interest and Histogram-Based Enhancement Approaches in 3D-Printing

Qualify assessment for extrusion-based additive manufacturing with 3D scan and machine learning

Detection of Material Extrusion In-Process Failures via Deep Learning

Multi-view Online Vision Detection Based on Robot Fused Deposit Modeling 3D Printing Technology

Multivariate sensor data analysis for fault detection toward feedback loop-based continuous learning in FFF 3D printer

Enhancing Fused Deposition Modeling Precision with Serial Communication-Driven Closed-Loop Control and Image Analysis for Fault Diagnosis-Correction

Development and comparison of machine-learning algorithms for anomaly detection in 3D printing using vibration data

Automated fault detection for additive manufacturing using vibration sensors

Empirical analysis of sensor type importance for data preparation of real-time operational status monitoring in fused deposition modeling 3D printers

An unsupervised end-to-end approach to fault detection in delta 3D printers using deep support vector data description

AutoML-driven diagnostics of the feeder motor in fused filament fabrication machines from direct current signals

An Evaluation of Classification Methods for 3D Printing Time-Series Data

In-Situ Laser-Based Process Monitoring and In-Plane Surface Anomaly Identification for Additive Manufacturing Using Point Cloud and Machine Learning

A Deep Learning approach to Defect Detection in Additive Manufacturing of Titanium Alloys