MSOE Library Home	Login My List - 0 Help
	Home   Search   My Account

Basic Advanced Power History

Search:  Author KeywordsBarcodeBib No.Call Number Keyword Search (Enter part of a call number -- use wildcards)General Keyword SearchKeyword Search of Contents NotesKeyword Search of Credits Notes (Enter word or words)Keyword Search of FormatKeyword Search of Item DescriptionKeyword Search of Publisher's NameName Keyword SearchPublication Date Keyword SearchSearch Part of an ISBN NumberSearch Part of an ISSN NumberSeries Title KeywordsSubject KeywordsTitle Keywords       Refine Search     > You're searching: Walter Schroeder Library, Milwaukee School of Engineering

Item Information   Holdings       More by this author     Dellemann, Robert.     Subjects     Neural networks (Computer science)     Electric machinery -- Monitoring.     Leak detectors     Water leakage.     MSE Project.     Browse Catalog     by author:      Dellemann, Robert.       by title:      Design of a leak det...         MARC Display

Design of a leak detection system for automated animal watering using neural networks / Robert Dellemann. by Dellemann, Robert. Subjects Neural networks (Computer science)   Electric machinery -- Monitoring.   Leak detectors   Water leakage.   MSE Project. Description:  74 leaves : ill. ; 29 cm. Contents:  Thesis advisor(s): Dr. Larry Fennigkoh. Committee members: Dr. Ronald Gerrits, Hue Tran. Introduction and background -- Requirements -- Software design -- Conclusion -- Appendix A- Alicat flow meter data sheet B- Alicat flow meter theory of operation C- Matlab progam listing. Objective: The objective of this project is to research and design a system for detecting leaks in an automated animal drinking system using neural networks. Detecting a leak using a rule-based algorithm when a mouse drinks 3-6 ml per day is a daunting challenge. Neural networks have the ability to train on patterns, detect those patterns and thus discern anomalies that occur. It is the intention of this project to implement Artificial Neural Networks to detect leaks in an automated animal watering system. Methodology: Research was conducted with 58 mice on an Edstrom Automatic Watering System using an Alicat L-series flowmeter. The flowmeter was monitored, collecting flow rates every minute for over six months. Collected data were used to train two neural networks in series. First, an unsupervised, self-organizing map clustered the input data. Second, a supervised, back propagation network was trained to detect leaks greater than one milliliter per minute. Neural network algorithms were designed and written with the application program Matlab. Major findings: Testing proved that the system could detect a leak over one milliliter per minute. A leak was artificially created to test the system. Output from the network displayed, with 100% accuracy, that this artificially created flow was a leak. The system also detected, with 100% certainty, an actual leak that had occurred during the course of this research.   Copy/Holding information Location Collection Call No. Status   Walter Schroeder Library Master's Theses AC805 .D4565 2006 Available Add Copy to MyList Format: HTMLPlain textDelimited Subject:   Email to:

Horizon Information Portal 3.25_9885

© 2001-2013 SirsiDynix All rights reserved.