Monitoring the waste to energy plant using the latest AI methods and tools

Loading...
Thumbnail Image
Journal Title
Journal ISSN
Volume Title
Kemian tekniikan korkeakoulu | Master's thesis
Date
2018-02-27
Department
Major/Subject
Chemical Engineering
Mcode
CHEM3027
Degree programme
Master's Programme in Chemical, Biochemical and Materials Engineering
Language
en
Pages
84+2
Series
Abstract
Solid wastes for instance, municipal and industrial wastes present great environmental concerns and challenges all over the world. This has led to development of innovative waste-to-energy process technologies capable of handling different waste materials in a more sustainable and energy efficient manner. However, like in many other complex industrial process operations, waste-to-energy plants would require sophisticated process monitoring systems in order to realize very high overall plant efficiencies. Conventional data-driven statistical methods which include principal component analysis, partial least squares, multivariable linear regression and so forth, are normally applied in process monitoring. But recently, latest artificial intelligence (AI) methods in particular deep learning algorithms have demostrated remarkable performances in several important areas such as machine vision, natural language processing and pattern recognition. The new AI algorithms have gained increasing attention from the process industrial applications for instance in areas such as predictive product quality control and machine health monitoring. Moreover, the availability of big-data processing tools and cloud computing technologies further support the use of deep learning based algorithms for process monitoring. In this work, a process monitoring scheme based on the state-of-the-art artificial intelligence methods and cloud computing platforms is proposed for a waste-to-energy industrial use case. The monitoring scheme supports use of latest AI methods, laveraging big-data processing tools and taking advantage of available cloud computing platforms. Deep learning algorithms are able to describe non-linear, dynamic and high demensionality systems better than most conventional data-based process monitoring methods. Moreover, deep learning based methods are best suited for big-data analytics unlike traditional statistical machine learning methods which are less efficient. Furthermore, the proposed monitoring scheme emphasizes real-time process monitoring in addition to offline data analysis. To achieve this the monitoring scheme proposes use of big-data analytics software frameworks and tools such as Microsoft Azure stream analytics, Apache storm, Apache Spark, Hadoop and many others. The availability of open source in addition to proprietary cloud computing platforms, AI and big-data software tools, all support the realization of the proposed monitoring scheme.
Description
Supervisor
Jämsä-Jounela, Sirkka-Liisa
Thesis advisor
Sun, Bei
Keywords
monitoring, waste-to-energy, artificial intelligence, deep learning, cloud computing
Other note
Citation