学术报告第201705期

Design and application of a neural network based inferential modeling approach with LASSO

Shi-Shang Jang

Department of Chemical Engineering, National Tsing-Hua University

Hsin-Chu, 30013, Taiwan

Feburary, 2017

报告时间：2017年2月28日上午10:00

报告地点：南一楼中311室

Abstract: In this study, a novel nonlinear inferential modeling approach is developed for predicting pivotal quality variables that are hard to measure in complex industrial processes. The proposed approach is an iterative backward deletion variable selection method in nature. At each iteration, the proposed approach utilizes artificial neural network (ANN) to construct a model, and then introduces the least absolute shrinkage and selection operator (LASSO) penalty into the model. The shrinkage parameter is determined by cross-validation method. After that, the shrinkage on the input weights of ANN is conducted and candidate variables whose input weights are zero are eliminated. The algorithm is repeated until there is no improvement in the model accuracy. Simulation examples as well as industrial application in a crude distillation unit were used to validate the proposed algorithm. The results showed that the proposed approach could construct more compressed model with better prediction accuracy than other existing methods.

Shi-Shang Jang

Professor and Director

Center for Energy and Environmental Research, National Tsing-Hua University

Hsin Chu, 30043, Taiwan

TEL: +886-3-571-3697; FAX:+886-3-571-5408

e-Mail: ssjang@mx.nthu.edu.tw

Education:

B.S.: National Hsin-Hua University, Hsin-Chu, Taiwan (1978)

M.S.: National Taiwan university, Taipei, Taiwan (1980)

Ph.D.: Washington University, St. Louis, M.O. (1986)

Experiences:

1986-1992: Associate Professor, Chemical Engineering Department, National Tsing-Hua University

1992 to now: Professor, Chemical Engineering Department, National Tsing-Hua University

2001 to 2004: Professor and Chairman, Chemical Engineering Department, National Tsing-Hua University

2012 to now: Director of Center for Energy and Environmental Research, National Tsing-Hua University

Research Interests:

CO2 Capturing Process Design and Optimization

Energy Integration of Manufacturing Processes

Semi-Conductor Process Control and Fault Detection and Classification

Process Control

Process Optimization

Selective Publication:

1. Yu-Jeng Lin, Tian-Hong Pan, David Shan-Hill Wong and Shi-Shang Jang *,Yu-Wen Chi, Chia-Hao Ye,” Plantwide Control of CO2 Capture by Absorption and Stripping Using Monoethanolamine Solution”, Ind. Eng. Chem. Res., 50, 3, 1338-1345, 2011.

2. Yu-Jeng Lin, David Shan-Hill Wong Shi-Shang Jang* and Jenq-Jang Ou, “Control Strategies for Flexible Operation of Power Plant with CO2 Capture Plant”, AIChE Journal, 58, 9, 2624–2949, 2012.

3. Jian-Guo Wang, Shyan-Shu Shieh , Shi-Shang Jang *, Chan-Wei Wu,” Discrete model-based operation of cooling tower based on statistical analysis”, Energy Conversion and Management 73 (2013) 226–233.

4. Jian-Guo Wang, Shyan-Shu Shieh, Shi-Shang Jang*, David Shan-Hill Wong, Chan-Wei Wu, “A two-tier approach to the data-driven modeling on thermal efficiency of a BFG/coal co-firing boiler”, Fuel, 111, 528-534, 2013.

5. Jia-Lin Kang, Kai Sun, David Shan-Hill Wong, Shi-Shang Jang*, Chung-Sung Tan, “Modeling Studies on Absorption of CO2 by Monoethanolamine in Rotating Packed Bed”, International Journal of Green House Gas Control, 25, 141-150, 2014.

6. Chun-Cheng Chang, Shyan-Shu Shieh, Shi-Shang Jang*, Chan-Wei Wu, Ying Tsou, “Energy Conservation Improvement and ON-OFF Switch Times Reduction for an Existing VFD-fan-based Cooling Tower”, Applied Energy, 154(2015), 491-499.

7. Kai Sun, Shao-hsuan Huang, David Shan-Hill Wong, Shi-Shang Jang,” Design and Application of a Variable Selection Method for Multi-layer Perceptron Neural Network with LASSO” IEEE Transactions on Neural Networks and Learning Systems, DOI: 10.1109/TNNLS.2016.2542866, 2016.