Identifier

etd-04042013-163026

Degree

Doctor of Philosophy (PhD)

Department

Chemical Engineering

Document Type

Dissertation

Abstract

Particulate processes have been widely involved in various industries and most products in the chemical industry today are manufactured as particulates. Previous research and practise illustrate that the final product quality can be influenced by particle properties such as size and shape which are related to operating conditions. Online characterization of these particles is an important step for maintaining desired product quality in particulate processes. Image-based characterization method for the purpose of monitoring and control particulate processes is very promising and attractive. The development of a digital image-based framework, in the context of this research, can be envisioned in two parts. One is performing image analysis and designing advanced algorithms for segmentation and texture analysis. The other is formulating and implementing modern predictive tools to establish the correlations between the texture features and the particle characteristics. According to the extent of touching and overlapping between particles in images, two image analysis methods were developed and tested. For slight touching problems, image segmentation algorithms were developed by introducing Wavelet Transform de-noising and Fuzzy C-means Clustering detecting the touching regions, and by adopting the intensity and geometry characteristics of touching areas. Since individual particles can be identified through image segmentation, particle number, particle equivalent diameter, and size distribution were used as the features. For severe touching and overlapping problems, texture analysis was carried out through the estimation of wavelet energy signature and fractal dimension based on wavelet decomposition on the objects. Predictive models for monitoring and control for particulate processes were formulated and implemented. Building on the feature extraction properties of the wavelet decomposition, a projection technique such as principal component analysis (PCA) was used to detect off-specification conditions which generate particle mean size deviates the target value. Furthermore, linear and nonlinear predictive models based on partial least squares (PLS) and artificial neural networks (ANN) were formulated, implemented and tested on an experimental facility to predict particle characteristics (mean size and standard deviation) from the image texture analysis.

Date

2013

Document Availability at the Time of Submission

Secure the entire work for patent and/or proprietary purposes for a period of one year. Student has submitted appropriate documentation which states: During this period the copyright owner also agrees not to exercise her/his ownership rights, including public use in works, without prior authorization from LSU. At the end of the one year period, either we or LSU may request an automatic extension for one additional year. At the end of the one year secure period (or its extension, if such is requested), the work will be released for access worldwide.

Committee Chair

Romagnoli, Jose

DOI

10.31390/gradschool_dissertations.2561

Share

COinS