Research And Application Of Data Mining In Production Direction System

Nowadays, with the competition of the iron and steel industry, it's the only way for the Pellet Factory's surviving and development to reducing the energy consuming, increasing the output and the qualities. Lots of production data basing on the time sequence remain in the control center database of the production automatic system. Using these data reasonably and efficiently can optimize the scheduling and increase the profits. This thesis is going to utilize these existing data, in order to forecast the output and distribute various raw materials according to the proposing production.First, the thesis analyzed the specialties and the existing problems of the controlment of the flow industry producing, and then reviewed the developments and the actual state of the research on the subject of production direction. Based on these above and in accordance with the ongoing systems' bugs and the production data's characteristics, through clustering statistics and relativity analysis, a statistics model was brought forward, then I present the model for production prediction. Secondly, during the construction of this scheme model, at first the evaluate of the products' integrated qualities in a long range resulted that the quality is stable. Following, with the usage of the fuzzy neural network the production prediction model was consummate. Finally, based on the production prediction model a production planning model was brought out .The thesis aimed at the specialties of the production data and so the techniques of statistics datamining was choosed. Adopting the Least Square Estimation Method, the statistics model of the relationship between the inputs and outputs was given. As the validation of this model was not good, the neural network techniques was introduced and the BP neural network was used to achieve the nonlinear model. In the final expression of the rules the fuzzy theory was introduced and it successfully avoids the over fitting and the instability of the neural network, so it consolidates the model's flexibility and reliability.