Fault Diagnosis of Delta 3D Printers Using Transfer Support Vector Machine With Attitude Signals

In traditional machine learning algorithms, more labeled data are required in model training for better classification accuracy. However, collecting labeled data is an expensive task for real applications. In this paper, a transfer support vector machine (TSVM) technique is introduced to tackle this problem with application to the fault diagnosis of delta 3D printers. Transfer component analysis is first proposed to capture shared features for representing the source and target domains, by cross-domain feature extraction from less labeled data in the source domain and massive unlabeled data in the target domain. Support vector machine is then applied to the fault diagnosis using the transferred features of attitude signals. In the experiments, fault classification rate achieves 83.79% using only 6.7% of the dataset for model training. Compared with peer methods, the results show that the present TSVM exhibits the best performance in extracting the cross-domain feature and improves the accuracy of the 3D printer fault diagnosis.