In many real-life applications, three-dimensional (3D) surface topography contains rich information about products and manufacturing processes. Different faults commonly appear on the topography of finished products in local and global patterns during manufacturing. Such faults are likely to cause changes in the variance or autocorrelation of the topographic values. Monitoring such changes is challenging due to the unique properties of the topographic surfaces. In particular, the topographic values are spatially autocorrelated with their neighbors and their locations are randomly changing from one surface to another under normal process behavior. The existing online monitoring approaches for the 3D surface topography lack the detection and diagnoses of changes in the topographic surfaces. In this dissertation, we investigate and develop four different online monitoring approaches to accurately characterize, detect, and diagnose various changes in topographic surfaces.
In the first approach, we develop a multi-level spatial randomness approach for online monitoring of global changes the surfaces. We propose a multi-level surface thresholding algorithm for improving the representation of surface characteristics in which an observed surface topography is sliced into different levels in reference to the characteristics of normal surfaces. The spatial statistical dependencies of surface characteristics at each surface level are accurately captured through a proposed spatial randomness (SR) profile. We then develop an effective monitoring statistic based on the functional principal component analysis for identifying anomaly surfaces with global changes based on their SR profiles.
In the second approach, we propose a multi-label separation-deviation surface model for effective monitoring of local variance changes in 3D topographic surfaces. The approach improves the representation of local topographic changes through a developed multi-label separation-deviation surface (MSS) model, which labels the important surface characteristics and smoothes out the noisy characteristics. We also propose two effective features for monitoring changes in surface characteristics. The MSS feature is introduced for capturing deviations within the label assignments, and the generalized spatial randomness feature is derived for quantifying deviations between the label assignments. These two features are integrated into a single monitoring statistic to detect local variations in topographic surfaces.
In the third approach, we develop a novel approach based on graph theory for accurate monitoring of local autocorrelation changes in 3D topographic surfaces. We enhance the representation of surface characteristics by proposing an in-control multi-region surface segmentation algorithm, which segments the observed surface pixels into clusters according to the information learned from in-control surfaces. The local and spatial topographic characteristics are accurately described through a developed maximum local spatial randomness feature. After representing the surface as a spatially weighted graph, we monitor its connectivity through the developed spatial graph connectivity statistic for accurate detection of local autocorrelation changes in topographic surfaces.
In the fourth and final approach, we investigate a generalized spatially weighted autocorrelation approach for fault detection and diagnosis in 3D topographic surfaces. We develop two algorithms to identify and assign spatial weights to the suspicious topographic regions. The normal surface “hard” thresholding algorithm initially enhances the representation of surface characteristics through binarization, followed by the normal surface connected-component labeling algorithm, which utilizes the obtained binary results to identify and assign spatial weights to the regions with suspicious characteristics. We also develop a generalized spatially weighted Moran (GSWM) index, which exploits the assigned weights to effectively monitor and detect changes in the spatial autocorrelation of each identified region. After an anomaly surface is detected based on its GSWM index, we accurately extract different fault information such as fault size, type, location, magnitude, and the number of faults.
The proposed approaches are validated for their effectiveness, efficiencies, and performance for online monitoring and diagnosis of various changes in 3D topographic surfaces.
|Advisor:||Elsayed, Elsayed A., Jeong, Myong K.|
|Commitee:||Pham, Hoang, Guo, Weihong, Lee, Howon|
|School:||Rutgers The State University of New Jersey, School of Graduate Studies|
|Department:||Industrial and Systems Engineering|
|School Location:||United States -- New Jersey|
|Source:||DAI-B 82/7(E), Dissertation Abstracts International|
|Keywords:||Anomaly detection, Graph theory, Image segmentation, Spatial randomness test, Surface monitoring, Surface topography|
Copyright in each Dissertation and Thesis is retained by the author. All Rights Reserved
The supplemental file or files you are about to download were provided to ProQuest by the author as part of a
dissertation or thesis. The supplemental files are provided "AS IS" without warranty. ProQuest is not responsible for the
content, format or impact on the supplemental file(s) on our system. in some cases, the file type may be unknown or
may be a .exe file. We recommend caution as you open such files.
Copyright of the original materials contained in the supplemental file is retained by the author and your access to the
supplemental files is subject to the ProQuest Terms and Conditions of use.
Depending on the size of the file(s) you are downloading, the system may take some time to download them. Please be