Fused Deposition Modeling 3D Printing Fault Diagnosis using Temporal Convolutional Network

초록

Fused Deposition Modeling (FDM) is used in quick prototyping and industrial manufacturing in many emerging industries. However, the process is imperfect, and printing failures often occur, thus leading to wasted time and resources. To address these issues of FDM-type 3D printers, developing a smart monitoring device designed to predict abnormal activities during the printing process accurately is proposed. In the proposed method, a set of collaborative sensors accumulates time-series data for forecasting. The proposed method uses Multi-Head Encoder-Decoder Temporal Convolutional Network (MH-ED-TCN) to extract data features and interpret the relationship between sensor readings occurring during the printing process to identify factors contributing to printing errors. Compared with other known models such as LSTM, SVM, and CNN, the proposed MH-ED-TCN model performs better in error detection. The experimental tests done yields an accuracy rate of 97.2% in the ability of the system to predict future 3D printing errors.