Multivariable modeling to predict mud entrance in Block Caving operation

Abstract

Caving mining currently represents a low-cost and massive exploitation option. However, its productive and economic attractiveness is affected by some challenges that operations must overcome. Such is the case of the mudflows, defined as violent inflows of a mixture of fine material and water into productive work. This phenomenon has caused damage to equipment and infrastructure, and even loss of life. To address this problem, various measures have been taken: drainage programs, planning strategies and conservative operation, and redefinition of reserves that do not consider the extraction of columns of saturated ore. Mainly this last measure involves leaving ore unexploited. Addressing the previously described problem, the present work describes the multivariate modeling process to predict the mud entrance to drawpoints in cave mining. In the first part of this work, the identification of relevant variables in the problem of mud entrance into the Diablo Regimiento sector was carried out, as well as a quantification of remaining reserves not extracted due to early closure of drawpoints. This work is titled Statistical analyzes of wet muck at Diablo Regimiento sector at El Teniente's Mine . The main variables identified as relevant in the wet muck declaration were: mud proximity, drawing uniformity, distance to topography, percentage of fines, accumulated drawn height, moisture content, horizontal distance to points with mud in the upper old sector, under points with mud in the upper old sector, presence of mud at point of the same drawbell, and belonging to the cave initiation area. In the second part of this work, the creation of predictive models of mud entrance to the drawpoints of the Diablo Regimiento sector was carried out. This was done based on observations at drawpoints: moisture and fines, as well as environmental variables. This work is entitled "Predictive models to estimate mud entry in Cave Mining - Diablo Regimiento Sector, El Teniente Mine". With regard to the creation of predictive models to predict the mud entrance, the Logistic Regression technique was used as a predictive technique for dichotomous events. In this regard, 12 models were developed for the Diablo Regimiento sector, 3 of which presented a high predictive quality. These models were evaluated using model prediction quality indicators, making them the best with a sensitivity of 89%, specificity of 71%, and area under the ROC curve of 0,88. The best of the 3 models considered the following variables: moisture content, mud proximity, drawing uniformity, accumulated drawn height, and horizontal distance to drawpoints with mud in the upper old sector. In the third part of this work, the application of the technique developed for prediction was carried out to a case of mining design in a sector in the Feseability Engineering stage. This work considers Garcés, who introduces the water entry variable into the system considering stream gauging. This work is entitled "Predictive models of wet muck entry for caving mining: application for the determination of dry reserves in the Andesita Project". Regarding the application case, it was shown that the developed prediction technique is useful in solving engineering problems, both in planning and in mining design. Its potential is such that it was used for the design of the Andesita Project, a sector that is currently under construction at the El Teniente Mine and which will begin production within the next few years. The inclusion of a comminution model that allows predicting the quantity and behavior of fines that are generated as a result of the movement of the mineral column, due to the extraction process of the columns, is seen as an opportunity to improve the modeling carried out. Based on the research carried out, it is concluded that the use of the Logistic Regression technique is reliable in predicting the entry of water-mud to drawpoints in block caving. In addition, predictive models allow us to identify which variables are the ones that most affect the mud entrance. Thus, this technique is useful in solving engineering problems, both in mine planning and design. Based on the foregoing, the following potentialities are identified in the developed technique: Risk analysis: it allows creating risk maps of water-mud entrance to a sector. This is useful both in mining projects in their engineering stages, since it allows making mining design decisions. As there are areas with a higher probability of water-mud entrance, more robust drainage systems are designed, or the use of semi-autonomous or remote-controlled LHD shovels is considered Mine planning: by being able to predict where the water-mud entrance will occur, different mining plans can be analyzed, thus being able to determine the volume of reserves to be exploited in a sector. Also, the technique allows to compare different cave initiation points, being able to identify which option is better from the point of view of the recovery of reserves. In terms of medium- and short-term planning, the technique allows defining controlled extraction zones in order to minimize the loss of reserves. Operation: in terms of operation, the technique allows decisions to be made on which points to privilege extraction and on which to carry out controlled extraction. The fact of having a probability of water-mud entrance generates an alert and an objective value for decision-making.