Convolutional neural network assisted infrared imaging technology: An enhanced online processing state monitoring method for laser powder bed fusion

Abstract

Laser powder bed fusion (LPBF) manufacturing technology is regarded as one of the most promising additive manufacturing technologies. However, the quality and properties of parts manufactured by LPBF are usually affected by different processing states in the LPBF process, which seriously limits the application of this technology. In this present study, an enhanced online processing state monitoring method for LPBF is proposed, which combines convolutional neural network (CNN) and infrared imaging technology. The LPBF process of 316L stainless steel samples with different parameters was recorded by an infrared thermal imager integrated into a commercial LPBF system. An improved YOLOV5 target recognition method for macro defect monitoring in LPBF manufacturing process is proposed. This method has the ability to enhance the small target detection capability by adding a transformer encoder structure and small target monitoring framework in the global scope. Double FPN structure is used to shape the neck network to obtain more effective multi-scale fusion. The CBAM structure block was inserted into the neck network to improve the fuzzy feature recognition capability. The experimental results demonstrated that compared with the original algorithm, the Mean Average Precision (mAP) of the improved model increased from 0.351 to 0.856, indicating that the improved model has a strong ability to recognize the macro processing state in LPBF process.