Construção de um modelo preditivo para determinação do TML em cargas de finos de minério de ferro

Abstract

Iron ore is one of the most relevant products of the national economy and a significant contributor in the Brazilian trade balance, being the main mineral resource exported by Brazil. Having almost all its volume transported by sea, the mining production chain has a relevant concern with the humidity of its cargoes, whose uncontrolled elevation constitutes a risk to the safety of bulk vessels involved in transoceanic transport. Controlling the moisture content in the loads and their maximum limit, known as TML – transportable moisture limit – is an obligation of all ore shippers, according to IMO (International Maritime Organization), the regulatory body of this type of transport. The mandatory determination of the TML for the loads of ore fines inserts in the logistics chain between mine, railroad and port, an additional complexity during the formation of their cargoes. When declared unfeasible, a new cargo proposal requires the repetition of the TML test, which may incur delays in vessel shipments, causing financial losses, both due to the decrease in productivity, as well as taxes and fines in chartering service. Since the TML is an empirical factor, which must be previously determined for almost all proposed loads, having the possibility of knowing in advance a predicted value for this parameter is very relevant to the production chain, improving the decision-making capacity and predictability of operations. In this context, this work proposes a predictive statistical model to estimate the TML of iron ore fines loads, starting from physicochemical parameters previously analyzed by production laboratories. The application of the multiple linear regression statistical method, with four independent variables, produced a model that is capable to predict values for TML with reasonable accuracy in a test dataset. This approach, when applied to the operational routine, should configure a planning alternative that starts from existing data, without increasing costs or requiring additional computational resources, contributing to a better progress in the sequencing of loads along the production chain of fines iron ore.