In this paper we have showed a statistically based approach to find patterns in the calcination process in ferronickel production. Since this is a very complex and hard to control environment, statistics arise as a powerful tool to discover relations between events and variables towards a satisfactory end. In our case the task was to find the fuel levels that would provide the desirable range in calcine temperature with the maximum likelihood. By applying data-based algorithms, like association rules and decision trees, we found the searched patterns in the form of rules, whereby the optimum control ranges are determined by probability.
Although this control is not applied to production yet, we could get a glimpse of some results by static simulation with real time process data. The results found in this article will support the adoption of this methodology to search for new patterns in this process.
1. Semel, F. J., D. J. Kasputis, and P. U. Gummeson. “Statistical Process Control in Iron Ponder Production and new product development.” SAE Congress. Detroit, Michigan, 1988.
2. Namikka, E., and G. J. Gibbon. “Identification of data mining techniques for industrial process analysis and control.” 1st African Control Conference AFCON. Cape Town, 2003.
3. Platon, R., and M. Amazouz. Applicaton of Data Mining Techniques for Industrial Process Optimization. Natural Resources of Canada, 2007.