Online image processing and defect detection in layered manufacturing using process signature

Layered Manufacturing (LM) is one such technology which has been rapidly growing in various applications of manufacturing. It is an additive manufacturing process where the parts are built layer by layer based on the geometry provided by a CAD system. Our research will only focus on Fused Deposition of Ceramic (FDC), which is one of LM techniques and is used for the fabrication of structural and functional parts using polymer, wax and ceramic. There may be many reasons that can cause defects, such as incorrect design parameter settings, position tracking control, flow control and disturbances, etc. Due to additive nature of LM process, the defects in a layer have to be identified and removed prior to deposition of the next layer in order to achieve high product quality. Therefore, a closed-loop machine vision system scheme is proposed for online process monitoring and process control. The machine vision system monitors the surface quality of each layer and feeds back this information to a closed-loop control system for online removal of possible defects.; In this dissertation, we introduce the concept of process signature for simply representing a uniform region within each layer, and propose an efficient and robust methodology of signature analysis for defect detection of single material ceramic parts and multi-material ceramic parts. It was shown that this signature analysis method is superior to the methods based on histogram and 1-dimentional discrete wavelet packet transform. Also, a statistical feedback control architecture integration SPC and APC is presented to adjust the parameter, such as flow rate or positioning speed, to minimize the possible defects in the subsequent layers. In some cases, tool path information is not available or it is too complicated to be used. Therefore, a general methodology for defect detection is proposed based on 2-dimentional discrete wavelet packet frame. In addition, an algorithm for feature space reduction is presented in order to improve the computational efficiency.; The proposed signature analysis methodology has been implemented to the current FDC machine and the FDC image processing software package FipSoft 1.0 is developed.