Authors: Teng, Sin Yong
Contributors: Máša, Vítězslav, Professor, Ponnambalam Sivalinga Govindarajan, Pavlas, Martin
Superior Title: TENG, S. Intelligent Energy-Savings and Process Improvement Strategies in Energy-Intensive Industries [online]. Brno: Vysoké učení technické v Brně. Fakulta strojního inženýrství. 2020.
Subject Terms: As core processing technologies in energy-intensive industries improve leaps and bounds, existing facilities gradually fall behind in terms of efficiency and productivity. Ultimately, harsh market competition and environmental legislation will force these traditional facilities to stop operations and decommission. Process improvement and retrofit projects are critical in maintaining the operational performance of these traditional facilities. Current approaches for process improvement are mainly Process Integration, Process Optimization and Process Intensification. From a high-level context, analysis from these fields consists of mathematical optimization, accumulated experience, and operational heuristics. These approaches serve good as a basis for process improvement. However, their performance can be further improved with up-to-date computational intelligence. Therefore, the purpose of this work is to apply advanced artificial intelligence and machine learning techniques into process improvement projects for energy-intensive industrial systems. The approach taken by this work is a multi-directional approach which tackles this problem from simulation to industrial systems with the following contributions: (i) Application of machine learning technique, which includes one-shot learning and neuro-evolution for data-driven single unit modelling and optimization. (ii) Application of dimension reduction (e.g. principle component analysis, deep autoencoder) for multiple-unit multiple-objective process optimization. (iii) Proposition of novel bottleneck tree analysis (BOTA) tool for the purpose of process capacity debottlenecking. An extended BOTA was also proposed to incorporate multi-dimensional problems via data-driven approach. (iv) Demonstrated effectiveness of Monte-Carlo simulations, neural network and decision trees for decision-making when integrating new process technology in existing processes. (v) Benchmarked Hierarchical Temporal Memory (HTM) and a dual-mode optimization with multiple predictive tools for real-time operational management. (vi) Implemented artificial neural networks in the conventional process graph (P-graph) framework. (vii) Highlight the future of AI and process engineering in biosystems via a commercial-based multi-omics paradigm, Industrial Process Improvement, Data-driven Modelling, Process Optimization, Machine Learning, Industrial Systems, Energy-Intensive Industries, Artificial Intelligence
File Description: text/html
Relation: http://hdl.handle.net/11012/195697
Availability: http://hdl.handle.net/11012/195697
