Fusing convolutional generative adversarial encoders for 3D printer fault detection with only normal condition signals

Collecting data from mechanical systems in abnormal conditions is expensive and time consuming. Consequently, fault detection approaches based on classical supervised learning working with both normal are not applicable some condition-based maintenance tasks. To address this problem, paper proposes Fusing Convolutional Generative Adversarial Encoders (fCGAE) method to create models only data. Firstly, obtain an adequate deep feature space, encoder 1D convolutional neural networks created. Then, these encoders optimized unsupervised way through Bidirectional Networks. Finally, the multi-channel features collected system merged One-Class Support Vector Machine. fCGAE applied 3D printers, where experimental results two cases show excellent generalization capabilities better performance compared peer methods.