A machine vision system for in-situ quality inspection in metal powder-bed additive manufacturing

Additive manufacturing;Quality inspection and control;Machine vision;Bayesian;Defect detection and identification;Statistical modeling;Image processing;Visual camera;Metal powder-bed fusion;Laser powder-bed fusion;Selective laser melting;Automated defect detection;Automated visual inspection;Geometry inspection;Porosity inspection,

Aminzadeh, Masoumeh ; Kurfess, Thomas Mechanical Engineering Rosen, David Cowan, Richard Qi, Jerry Zhou, Chen ; Kurfess, Thomas

University:Georgia Institute of Technology

Department:Mechanical Engineering

关键词: Additive manufacturing;    Quality inspection and control;    Machine vision;    Bayesian;    Defect detection and identification;    Statistical modeling;    Image processing;    Visual camera;    Metal powder-bed fusion;    Laser powder-bed fusion;    Selective laser melting;    Automated defect detection;    Automated visual inspection;    Geometry inspection;    Porosity inspection,;   

Others  :  https://smartech.gatech.edu/bitstream/1853/56291/1/AMINZADEH-DISSERTATION-2016.pdf

美国|英语

来源: SMARTech Repository

【 摘 要 】

In-situ inspection of geometric accuracy and porosity in metal powder-bed additive manufacturing (AM) using visual camera images was proposed and addressed. For the first time, an imaging setup and the required machine vision (MV) algorithms were developed, implemented, and evaluated to inspect the part cross sectional geometry and to provide an assessment of porosity, in metal powder-bed AM, by direct visualization of porosity from in-situ visual camera images. A visual imaging setup was developed to produce images in-situ from each layer that visualize the fused objects in the layer of powder and the detailed surface quality in terms of formed porosity. Appropriate image processing algorithms were designed and implemented for detection of fused geometric objects and estimation of the geometric parameters. Geometric objects were detected with a boundary point-to-point (root-mean-square) error of 81 microns. Formed porosity was directly detected from the camera images of the layers using machine vision algorithms. In addition to detection of individual pores, an intelligent approach was proposed and implemented to identify defective regions in the part layer that provides a qualitative assessment of porosity. For this purpose, a statistical Bayesian framework was developed and trained based on texture features. The Bayesian network was designed to maximize the Figure of Merit and finally led to sensitivity of 89% and specificity of 82%. In addition to offering an efficient MV-based inspection system, this work also provides an infrastructure for developing more precise and confident imaging and detection systems for powder-bed AM for visible-light camera images as well as other sources of 2D measurements such as height maps, microscopic images, and stereo imaging.