Search

Os processos, sistemas e pessoas envolvidos na cadeia de produção mineral são responsáveis pelo aumento considerável na geração de dados operacionais. Neste cenário, torna-se necessária a utilização de ambientes com capacidades de armazenamento e processamento de dados cada vez mais robustos para que métodos de análises inteligentes possam ser aplicados, a fim de realizar previsões do desempenho produtivo por meio de algoritmos de aprendizado de máquina. O objetivo deste estudo é aumentar a previsibilidade do desempenho operacional de equipamentos de mina, diretamente relacionado com a tomada de decisão nas atividades de planejamento de lavra de curto prazo. Para alcançar este objetivo foi proposta uma metodologia de aprendizado de máquina para o planejamento de lavra que abrange tarefas desde a preparação dos dados envolvidos no processo produtivo, até a tomada de decisão e replicação da rotina de análise para os próximos períodos que serão planejados. A metodologia foi aplicada no planejamento de lavra dos turnos operacionais em uma mina de cobre localizada na região norte do Brasil com o objetivo de prever o desempenho de produção de escavadeiras considerando variáveis de planejamento de lavra, geológicas, geográficas, de manutenção, da programação de horas, das turmas de trabalho escaladas para o turno e dos turnos de trabalho a serem planejados. A partir de diversos algoritmos de aprendizado de máquina, foram criados 175 modelos preditivos, treinados e testados por meio de validação cruzada, para que fosse definido uma modelagem que melhor se ajustasse ao padrão dos dados operacionais. Estes resultados comprovaram que sistemas preditivos podem ser inseridos na rotina operacional do planejamento de lavra de curto prazo, auxiliando na tomada de decisão para a alocação de equipamentos de maneira realística, eficiente e ágil. The complex combination of controls, systems and human aspects presented in the mine value chain are today responsible for an increasing amount of digital data in the mining industry. In this scenario, it is imperative the use of reliable and intelligent systems that can store and process the data to predict mine equipment performance. The objective of this study is to improve the prediction of operational performance in mine equipment for the short-term planning. For this purpose, it is proposed a machine learning (ML) methodology to map the production process from data collection until planning, with replication of the generated routines for subsequent short-term period analysis. The methodology was applied to predict the operational performance of excavators during working shifts in an open pit copper mine located in Northern Brazil, considering a series of variables such as operational, geological, geographic, maintenance. 175 predictive models were generated during the study, which were tested through cross-validation to improve the model adjustment to the collected data. The results obtained using this methodology confirmed that the use of ML predictive models provides a better understanding of the operation and allocation of mine loading fleet through the use of fast and realistic predictive routines.

Small-scale mining usually operates under high geological uncertainty conditions. This turns mine planning into a complex and sometimes inaccurate task, resulting in low productivity and substantial variability in the quantity and quality of the mineral products. This research demonstrates how the application of a novel methodology that relies on traditional and low-cost geophysical methods can contribute to mine planning in small-scale mining. A combination of resistivity and induced polarization methods is applied to enhance mine planning decision-making in three small-scale mining operations. This approach allows for the acquisition of new data regarding local geological settings, supporting geological modelling and enhancing decision-making processes for mine planning in a timely and low-cost fashion. The results indicate time savings of up to 77% and cost reductions of up to 94% as compared with conventional methods, contributing to more effective mine planning and, ultimately, improving sustainability in small-scale mining.

The strategic mine plan is a crucial step for the success of mining companies, and for its development, it is necessary to use a number of inter-related variables that are usually estimated independently. These variables include operational data that is traditionally isolated in information islands between the different departments in the mine or they are consolidated into individual models. This reduces the holistic view of the deposit, thereby causing a negative impact on the results of the strategic planning itself. In order to improve the process and to maximize the production and/or value of a mining project, there needs to be an integration of the geology, the mine plan, the processing and the geometallurgy data. In order to accomplish this, a new methodology is proposed for the creation of a technological model. This model can be interpreted as the consolidation of the different models required for a better understanding of the geological and technical information of the deposit. This concept was developed and applied at a copper and gold mine site located in Brazil. Based on the evaluation of different blasting and mill productivity scenarios through a pit-to-plant approach, it was possible to obtain operational short-term gains such as a 10.7% increase in the plant production rate and a 2.2% increase in the crusher's feed rate with little or no capital investment.