Knowledge Representation and Expert Systems for Mineral Processing Using Infobright

Open source tools for Knowledge Representation in databases and the implementation of a real time expert system for mineral processing operations (size reduction and enrichment) are discussed. The use of a column-oriented database system (Infobright IEE) to store quantitative data from sensors that measure feed size distribution, feed rate, aeration rate, pulp density, pH and temperature allows low latency database query responses and real time process control and analysis. Qualitative metadata can be generated with the use of mathematical process models (simulation outputs, reduction equations, transforms), and from the natural language analysis of process data (reagents and ore mineralogy). The toolkits Wordnet and the Natural Language Toolkit (NLTK) are proposed for metadata generation, processing qualitative text information present in process databases, and for generating data for subsequent inference engine rule checking. We took advantage of the power and ease of the programming language Python to implement a framework for fuzzy and rough set rules generation, and to create an on-line-analytical-processing (OLAP) system for reporting production process parameters.