T-S Fuzzy Rule Extract Based On Data Mining

System model is all long regarded as a critical step in automation and optimization. "Model First" is also pursued all the times in industrial applications. Along with technology development at very fast speed and production scale-up, a lot of systems get more and more complicated. Especially increasing hot industrial competition all over the world and people's deeply understanding, changing and harmonizing of nature, which make system model encounter an unprecedented challenge, task is arduous and urgent to deal with, and at the same time, data mining is appearing. So how to use it to solve the problem is becoming a research hotspot.Data mining is a new technique of data processing and analyzing, which develops rapidly with the development of the technique to store data, the enlargement of the scale of database, the accretion of people's need to take information from data set. It is considered as the offspring of computer technology, database technology, artificial intelligence, statistics and so on. In this paper the evolutive history, basic principles, some important research fruits about data mining, and the research status about data mining based on neural network and fuzzy neural network were introduced, and it also introduced the application status and future development on data mining in the field of industrial process optimization.In this paper, it makes a research on data mining based on fuzzy neural network, and implements fuzzy neural network interference system based on second-order T-S fuzzy, which is based on first-order adaptive fuzzy neural network. In spite of the system based on second-order T-S fuzzy model is complex compared with first-order system, but to some problem, if only select correct model structure and parameter, its function is better than the first-order system.The adaptive fuzzy neural network system based on second-order T-S fuzzy model is introduced in detail, including system math descriptions, system structure and system learning algorithm. A typical example is given to prove that the adaptive fuzzy neural network system based on second-order T-S fuzzy model has its superiority.At the end of this article, we use the true data from matte converting in Pierce-Smith converters to test the approach mentioned above is valid. Because the data from industrial process exist a err, and the differences between the variables'dimensional, so in order to avoid the influence from noise signal, and the phenomenon "big number eats little number", corresponding de-noising and normalization should be done before training system with this data .At last, we test and verify the method mentioned above by predicting the end-point in matte converting in Pierce-Smith converters, the simulative experiment based on real industrial data show that the method and technique to build model are feasible and the model build in this paper has a good performance in predicting the output of the matte's grade.