A Fuzzy Expert System on Qualitative Mineralogy

This project involves the evolution of an existing computer software system designed to teach mineralogy and assist in the identification of a mineral specimen. The current system operates in a proprietary hypertext-based software environment that needs to be upgraded to run in a Windows-based browser such as Firefox or MS-Explorer.

The system uses a fuzzy expert system to present information on over 250 minerals and to assist a student in learning how to identify an unknown specimen. It uses observations made by the user to increase or decrease the degree of belief in a particular mineral name that is initially chosen by the user. In this way the user acquires the ability to make observations about the properties of rocks and minerals such as colour, streak, luster, S.G., hardness, crystal structure and habit, cleavage, twinning, and numerous unique characteristics such as twinning and fluorescence; and apply these facts to identify minerals.

The research involves evolving the existing rules and fuzzy logic structure into a form compatible with web-based programming. Some of the existing rules can be made to operate in a more efficient manner using an Agent-based approach to reuse rules and apply them across the different properties rather than being applied as separate entities. The system uses a number of Fuzzy Associative Memory maps to take input information and determine the degree of belief in an output. Input and output is typically of a linguistic nature except where measurements have been made such as S.G. and hardness. The method of Defuzzification uses a Weighted Average approach although some rules use Weighted Inferencing similar to an Artificial Neural Network. Examination of ways to streamline these two alternatives will occur during the research. The ability to easily add new minerals into the system will also be examined.

The research plan for this work is as follows:

Week 01 – Familiarization with the existing system
Week 02 – Learning the AI elements in the system
Week 03 – Transferring the hypertext documents into HTML/XML
Week 04 – Continuing with hypertext documents and database
Week 05 – Creating rules for on-line application/manipulation of data
Week 06 – Continuing to create rules
Week 07 – Continuing to create rules
Week 08 – Testing the system
Week 09 – Testing the system
Week 10 – Developing the User Interface
Week 11 – Final clean-up and write-up of a brief report
Week 12 – Presentation of the new system to our department

The ideal candidate for this project would possess knowledge in web-based programming, data management systems, and would wish to learn more about AI techniques such as fuzzy logic, expert systems, and artificial neural networks. An interest in minerals and earth sciences would be an asset, but expertise in this area is not required to do this research.

Faculty Supervisor:

John Meech






University of British Columbia


Globalink Research Internship

Current openings

Find the perfect opportunity to put your academic skills and knowledge into practice!

Find Projects