WTC-AutoFormer-based additive manufacturing failure prediction in digital twins

Additive Manufacturing (AM) has been widely used in the industrial field. Failures in the manufacturing process may lead to a decrease in product quality and resource waste. Existing research has focused on predicting failures in additive manufacturing. But most of them lack accurate modeling of complex dynamic processes and fail to effectively integrate real-time data and simulation models. To address this challenge, we proposed a novel failure prediction method WTCAutoFormer for additive manufacturing with the additive manufacturing digital twin platform. Firstly, we utilize Wavelet Transform Convolution (WTC) to preprocess the multidimensional sensor data collected during the additive manufacturing process. WTC extracts high-frequency and low-frequency features with wavelet transform to more accurately capture abnormal fluctuations in sensor sequences. Subsequently, we adopt AutoFormer to extract hidden features in sensor sequences. The experimental results show that the proposed method has significant advantages in the accuracy of failure prediction. Compared to traditional prediction methods, WTC-AutoFormer can more accurately identify potential problems and provide warnings, thereby reducing risks and costs in the production process.