Substantiation of model for prediction of ore body contours and useful components content in reserves based on artificial neural networks

The ores of the Talnakh deposit are complex, from them they extract: copper, nickel, cobalt, metals, platinum groups, gold, silver, as well as selenium, tellurium, ruthenium and sulfur. Ores have a complex mineral composition and are characterized by a variety of forms of finding useful components. The main ore minerals of all types of ores are: pyrrhotite, chalcopyrite and its varieties, pentlandite and cubanite. Nickel, copper, platinum, palladium, gold and silver are represented mainly by their own minerals; rhodium, ruthenium, osmium, iridium, selenium, and tellurium are in scattered form. As a result of the fact that the ore deposits of the Skalistaya mine lie at a considerable depth (more than 700 m), its ores are classified as dangerous for mountain impacts. Purpose of the study. The commissioning of disseminated ore mine field sections, taking into account the risk of mining impacts, determines the need for accurate information on the content of useful components in a particular section of the deposit, which in turn will increase the efficiency and safety of mining reserves, as well as increase the economic efficiency of the mine. Research methods and materials. Constant geological control, taking samples, rock samples is not always possible or difficult as a result of the high complexity of both field and office work. In addition, the cost of additional exploration of deposits (field sites, production blocks) is significantly high, and the period for obtaining results can reach up to several months. The model developed by the authors for predicting the contours of the ore body and the content of useful components in reserves based on the use of an artificial neural network will make it possible to accurately predict the shape of the ore body, the conditions for its occurrence, as well as the composition and value of the content of useful components in the contours of the mining unit. Research results. The forecast model is based on an artificial neural network of direct distribution, which allows predicting the desired parameters with an error value not higher than 5–10 %. The model developed by the authors allows us to obtain high accuracy of forecasting the conditions of occurrence, the shape of the ore body and the content of components from the forecast data of each forecast well. The period for obtaining geodata on the condition of occurrence, shape and content of useful components is much shorter than the period for detailed exploration of reserves (several hours versus several months, respectively). Discussion of research results. The obtained forecast data are presented in the form of a table on the content of useful components in the disseminated ore reserves on each meter of each forecast exploration well of the grid of forecast wells of the forecast production block. The ore body contours are determined by the predicted content of useful components in the reserves of disseminated ores of the production block, taking into account the data of the grid of forecast exploration wells. Conclusions. To obtain timely and highly accurate information on reserves for both rich ores and disseminated ores, taking into account the depth of development and the presence of danger from mountain impacts, it is necessary to use the forecast model developed by the authors based on an artificial neural network of direct distribution. The model developed by the authors allows us to obtain high accuracy of forecasting the conditions of occurrence, the shape of the ore body and the content of components from the forecast data of each forecast well

Artificial neural network · Computer Science · Mineral Processing and Grinding · Mining and Gasification Technologies · Mining Techniques and Economics · Artificial Intelligence