Please use this identifier to cite or link to this item:
https://www.um.edu.mt/library/oar/handle/123456789/92358
Title: | Application of artificial intelligence for energy efficiency analysis of compressed air systems |
Authors: | Farrugia, Daniel (2021) |
Keywords: | Artificial intelligence Pneumatic machinery Machine learning Machinery -- Maintenance and repair Support vector machines Leak detectors Neural networks (Computer science) |
Issue Date: | 2021 |
Citation: | Farrugia, D. (2021). Application of artificial intelligence for energy efficiency analysis of compressed air systems (Bachelor’s dissertation). |
Abstract: | Compressed air systems are highly inefficient sources of energy, however, compressed air use is not foreseen to subside any time soon. The fundamental problem with compressed air is the wide range of inefficiencies present, both on the generation side and on the distribution and utility side. Leakages and fault constitute a large part of the losses associated with compressed air systems and, hence, leak and fault detection methods are continually being researched, in hopes of discovering improved techniques. Conventional leak and fault detection methods require planning, maintenance personnel as well as dedicated equipment to be carried out. Moreover, the subsequent analysis of the gathered information is typically done manually and on site. The dawn of the fourth industrial revolution has brought about a new perspective towards maintenance, namely Condition-based Predictive Maintenance. This combines the advantages of condition-based maintenance and predictive maintenance into one. Real-time remote condition monitoring is an extensively researched are of study, and the benefits of such approaches are the automatic nature of deriving results or making decisions and remote diagnostics makes maintenance-related activities easier to plan and carry out. In this project, the use of supervised machine learning techniques is proposed for automatic leak and fault detection within the demand side of compressed air systems. The relevant literature indicated that the benefits of artificial intelligence applications are not being exploited with respect to compressed air systems. Six prototype machine learning classification algorithms were developed in order to analyse pre-collected data to determine the feasibility of the proposition. The results obtained were promising. One feature obtained 100 per cent accuracy across all of its test runs, and another instance resulted in a 99 per cent accuracy for fault detection purposes. More research is necessary to develop certain implementations which for the time being may seem untimely. The overall project objective was deemed to be accomplished, owing to the fact that leak and fault detection was possible through the proposed methodology. |
Description: | B.Eng. (Hons)(Melit.) |
URI: | https://www.um.edu.mt/library/oar/handle/123456789/92358 |
Appears in Collections: | Dissertations - FacEng - 2021 Dissertations - FacEngIME - 2021 |
Files in This Item:
File | Description | Size | Format | |
---|---|---|---|---|
21BENGME008.pdf Restricted Access | 4.54 MB | Adobe PDF | View/Open Request a copy |
Items in OAR@UM are protected by copyright, with all rights reserved, unless otherwise indicated.