An optimization algorithm to assess different reserves models for open pit short term planning

Abstract

This paper summarizes a research to develop an optimization algorithm used to support short term planning decisions in an open pit operation. In particular, it is sought to use this algorithm to evaluate block models that have been computed using different geostatistical techniques. This leads to several production schedules, each representing the best production option for a given block model. Then, these schedules are compared against a most likely optimal schedule that uses blast hole grade information as a main input. This comparison provides sufficient information to decide upon the geostatistical method that better forecasts a given production schedule. The optimization algorithm is oriented to minimize production deviations with respect to the medium term plan (yearly budget) subject to uniform grade fed to the plant over a period of time (e.g. a week), minimum tonnage fed, shovels digability, among others. The algorithm will report the path that needs to be followed at different mining faces in order to reach optimality. The method proposed is of direct optimization, in which all the possible combinations in short time steps are evaluated explicitly hence allowing decision making regarding the path to be followed. Finally, the geoestatistical method that better predicts the actual mining outcome is determined. This path minimizes deviations regarding the medium term production targets. The method is implemented and validated with actual data: a reserves model is built from drillhole data using several geostatistical models, namely, ordinary kriging and Gaussian simulation. The models are compared with blasthole data that is effectively used to make the decision of sending the blocks to the processing plant or to the waste dump. The responses from different geostatistical models are compared with the optimized schedule determined with the optimization methodology described.