عیار

In the present research, a probabilistic neural network based on the Bayesian probabilistic ‎algorithm was employed to classify the grade of Ali-Abad copper deposit in Yazd. For this ‎purpose, induced polarization (IP) and resistivity (Rs) geophysical data and rock type of ‎exploration borehole cores as geological information corresponding to four geophysical ‎profiles, DD-1, PD-2, PD-3 and PD-4 were used as input parameters as well as the copper ‎grade of the boreholes as target parameter. To achieve the goal, 488, 528, 188, and 456 data ‎were randomly collected from the sections related to DD-1, PD-2, PD-3 and PD-4 geophysical ‎profiles so that 75% of total data were selected for training and 25% to test the probabilistic ‎neural network. The performance of the proposed approach was evaluated by confusion ‎matrix through the ratio of summation of data on the main diameter to the total test data, as ‎well as determination of Commission and Omission errors. The results of the research show ‎that the probabilistic neural network could estimate the test data for DD-1, PD-2, PD-3 and ‎PD-4 profiles with accuracy of 60, 74, 60 and 83.3%, respectively which are reasonable ‎considering the type of available data. In addition, the results were qualitatively evaluated ‎through plotting isograde maps of four exploratory cross-sections over the geophysical ‎profiles. This process was carried out using the assay data of exploration boreholes, gridding ‎and the grid interpolation with the high accurate kriging estimation method, which was leaded ‎to favorite results.‎

Grade estimation is one of the important stages for mining assessment. Grade values have asignificant effect on scheduling, designing and management of the mine. Therefore, it is importantto apply a method, which is able to estimate the necessary parameters with a high accuracy. One ofthe direct methods for figuring out the grades is to use exploration wells, which, because of theirhigh costs, usually it is impossible to use them extensively. In this study, a novel method based onfuzzy logic, neural networks and genetic algorithm is presented. This algorithm, by applying geneticalgorithms for optimization of the architecture and the neuro-fuzzy parameters, is able to achievebetter results as compare with other traditional methods for grade estimation. This is because ofusing different artificial intelligent methods. Foe this aim, Darrehzar copper porphyry was selectedas the case study. According to the results obtained our proposed algorithm able to capture theexisting pattern between the inputs and outputs finely and to estimate the grade with a highprecision.