In situ monitoring for numerical controlled manufacturing of large conceptual prototype based on multi-view stitching fusion

• Published in: International journal of advanced manufacturing technology 2023-10, Vol.128 (7-8), p.3197-3214
• Main Authors: Xu, Jinghua, Wang, Linxuan, Zhang, Shuyou, Tan, Jianrong
• Format: Article
• Language: English
• Subjects: Boundary representation; CAE) and Design; Computer-Aided Engineering (CAD; Design defects; Digital twins; Engine blocks; Engineering; Finite element method; Flaw detection; Industrial and Production Engineering; Machine vision; Manufacturing; Mechanical Engineering; Media Management; Monitoring; Original Article; Prototypes; Stitching; Thin walled shells; Vision systems
• ISSN: 0268-3768; 1433-3015
• DOI: 10.1007/s00170-023-12053-1

This paper presents an in situ monitoring method for numerical controlled manufacturing of large conceptual prototype based on multi-view stitching fusion (MSF). With the extensive application of numerical controlled manufacturing, like additive and subtractive manufacturing (ASM), the forming size tends to increase. Nevertheless, limited by optical lens and affiliated machine vision parameters, it is difficult to avoid the existence of blind areas when detecting defects including scratch, crack, pore, and thermal deformation by machine vision. To overcome the obstacles, multi-view information, including visible and depth sequences, were captured during the manufacturing progresses to expand monitoring scope. The overlapped regions of various frames were confirmed by stitching fusion of multi-views with homography matrix. The flaws were detected by superposing anomalous regions of multi-views which were extracted via multi-dimension synthetical anomalous degree map. The finite element analysis (FEA) was conducted to extract susceptible regions and locally adjust threshold. Inspired from digital twin visualization of Boundary representation (B-Rep), the numerical controlled manufacturing process of large textureless conceptual prototype can be quantized and visualized synchronously. The numerical outcomes were evaluated and verified that the MSF was effective in distinguishing wide-range defect regions of in situ large parts, thereby generating feedbacks. The physical experiment was implemented on W12 engine upper block and thin-walled shell of engineering machinery with laser stereo lithography appearance (SLA). The physical experiment prove that it is of great significance for fabricating large-scale high reflective objects with low defect and high efficiency, further for prototype verification of conceptual design.