Infrared thermographic imaging based real-time layer height estimation during directed energy deposition

Monitoring the layer height during the directed energy deposition (DED) process is challenging but essential for the assurance and control of the dimensions and quality of additively manufactured products. In this study, a real-time layer height estimation system was developed, particularly for the DED process based on infrared thermographic imaging. The proposed system estimates the layer height using the measured melt pool properties and printing process parameters as inputs to an artificial neural network (ANN). The unique advantages of the proposed technique include the following: (1) an ANN is developed and optimized for real-time estimation of the layer height during the DED process; (2) the relationship between the layer height and melt pool properties is numerically and experimentally investigated under various printing conditions; (3) because of the coaxial design of the infrared camera installed, the proposed system can estimate the layer height even under the complex movement of the printing nozzle, making the system attractive for additive manufacturing of complex geometries; and (4) when combined with the developed melt pool depth estimation technique developed by the authors, a single infrared imaging system can estimate the melt pool width, length, depth, and layer height simultaneously. The layer height estimation performance was examined by printing multilayer and multitrack stainless steel 316 L materials under varying printing conditions. The overall root mean squared error and absolute percentage errors of the height estimation were 25.44 mu m and 12.62%, respectively, for printing an average layer height of 200 mu m.

Automated summary

In this study, a real-time layer height estimation system based on infrared thermographic imaging was developed for the directed energy deposition (DED) process. The system utilizes an artificial neural network (ANN) to estimate the layer height using melt pool properties and printing process parameters as inputs. The proposed system has unique advantages, such as its ability to estimate layer height even under complex movement of the printing nozzle and its simultaneous estimation of melt pool width, length, depth, and layer height. The performance of the layer height estimation was examined through printing experiments, with an overall root mean squared error of 25.44 μm and an absolute percentage error of 12.62% for an average layer height of 200 μm.