Height consistency compensation in laser-directed energy deposition of thin-walled parts

•A framework of geometric defect compensation is built for LDED forming of thin-walled parts.•The morphology of the deposited layer and its height value are extracted using machine vision.•Flatness rate is introduced to quantitatively characterize layer morphology and height consistency.•Interval-specific scanning speed as the output via parametric feedback is used for high consistency compensation of subsequent deposition. Geometric defects, especially for the abnormal height variability of the layers, tend to occur in the laser-directed energy deposition (LDED) forming of thin-walled parts, which can severely affect the final accuracy and quality. Facing this problem, this work proposes a framework to implement the real-time layer height prediction and parameter feedback for the compensation on the height inconsistency during the thin-walled part forming. First, the deposited layer contours are extracted in succession and the corresponding height values are recorded using machine vision, including melt pool recognition, spatter removal and melt pool edge localization, etc. The height consistency of the deposited layer, as a data basis, is evaluated by calculating the newly introduced flatness rate. In addition, the mapping relationship between layer height and scanning speed (h-v) is established as the quantitative foundation of the parameter feedback algorithm. By iterating the h-v function with the height data of the previous deposition, the prediction of the next layer height can be achieved. Meanwhile, the parameter feedback can be performed to output a set of deposition interval specific scanning speeds, to be adopted for the height consistency compensation in the next layer deposition. Finally, through experimental validation, it is demonstrated that the proposed method has practicality and applicability in real-time height consistency compensation, and the flatness rate of the thin-walled part was improved by a maximum of 22.8 %. This study presents an in-situ instant process monitoring and defect repair approach to ensure geometric accuracy, aiming to inspire more stable and intelligent LDED forming. [Display omitted]