Análise multivariada para estudo de um processo industrial de concentração de minério de ferro

Abstract

In industrial processes, product standards and qualities can be assessed based on various process responses. From predetermined process conditions, established standards for raw materials should be ensured, avoiding significant variations, therefore performance and quality should be monitored throughout the production chain. Within the process, hundreds of explanatory variables can be found, with respective adopted levels, or naturally established. In the light of these many possibilities, it is more difficult to assess and obtain responses from processes. In the case of the mining industry, which is endowed with complex systems for valuing mineral reserves, originating products that will constitute raw materials for the most diverse basic industries, we are faced with processes and systems as complicated as any other industry. Thus, the processes for valuing mineral reserves are difficult to deal with in terms of the variability of the ores that feed the mills and scale of processed mass. In these processes, the explanatory variables can easily reach hundreds, to which they are linked to the same process response. The following study, is related to the industrial process of an iron ore mining company in the municipality of Itabirito, in Minas Gerais, in which the industrial process has several unitary operations, or stages of processing. In this work, it was evaluated the productive system of mineral goods, dedicated to the production of inputs for the steel industry. This productive system performs the concentration of the useful mineral (iron minerals, hematite or goethite, for example) for obtaining metallic iron. One of the statistical tools used was the multivariate analysis (k-means), which was a guide of laboratory tests to validate hypotheses formulated along process analysis and improvements in the industrial standard. The problem focused was the production of mineral product, which raw material for obtaining metallic iron, with a level of silica contaminant above that specified by the customer, consisting of a production failure (defect). This failure leads to a loss of production, because the product with a high level of contaminant is segregated, directly affecting the company’s final result, revenue and customer service. The objective of this study, therefore, was to identify the probable causes of contamination of the product, in order to prevent the failure from occurring and, consequently, promote greater recovery of the useful mineral mass as a product. For the basis of the evaluation, the mapping of the industrial process was developed to extract the relevant information for the organization of the database and its contents. The information available in the database came from hundreds of sensors distributed in the plant’s various equipment, as well as from the results of periodic sampling at the plant, which are inputs for chemical and physical determination of the ore in flows of input and output of the power plant processes. This phase of the study allowed the visualization of the responses of each process with the corresponding explanatory variables. The industrial process responsible for the response associated with the content of silica contaminant in the product is associated with at least 50 variables. To control these variables, several actions are required, ranging from activities automatically performed by the equipment to the application of statistical tools to obtain the best process parameters. The analyses carried out allowed us to conclude that the control of such a process is complicated to be quickly evaluated through the senses of the process technicians, even the most experienced ones. It is then justified the need to adopt a complete methodological direction to undertake actions that change the parameters of the variables and allow the improvement of the process.