A. Kakouei; M. Masihi; M. Shirani
Abstract
Determination of different facies is one of the important and basic tasks of geological engineering characterization of reservoir rocks from well logs and core data. Our objective is to identify and determine different facies of the South Pars Field using RBF and PNN neural networks in order to perform ...
Read More
Determination of different facies is one of the important and basic tasks of geological engineering characterization of reservoir rocks from well logs and core data. Our objective is to identify and determine different facies of the South Pars Field using RBF and PNN neural networks in order to perform static and dynamic simulation. These networks are utilized to identify facies of the South Pars Field for the first time in Iran. In this study, we use different parameters of mentioned networks such as ‘spread’ and ‘goal numbers’ to improve networks operation. In this regards, the optimum values of these two parameters were 0.01-10 and 0.02-0.04 respectively. The results show that the RBF and PNN neural networks are robust means to determine and model the facies of the South Pars Field in Iran.
A. Sayadi; M. Manjezi; H. Shahr Abadi
Abstract
Reserve evaluation is one of the most important parameters for mine designing and several methods have been developed in this regards. Among these methods, in addition to geostatistical methods, artificial methods such as Artificial Neural Networks (ANN) are suitable for reserve evaluation. In this research, ...
Read More
Reserve evaluation is one of the most important parameters for mine designing and several methods have been developed in this regards. Among these methods, in addition to geostatistical methods, artificial methods such as Artificial Neural Networks (ANN) are suitable for reserve evaluation. In this research, geometrical and block model of ٍEsfordi phosphate mine are prepared and the reserve is estimated. The block model contains 100 thousand blocks with dimensions of 25×25×5 m. To estimate the grade of each block,both methods of geostatistical and ANN methods are used. For geostatistical estimation, normal kriging is applied. In ANN a perceptron multilayer network is used and for training of network LM method is considered. Based on geostatistical and ANN methods, the amount of estimated reserve is 16.5 Mt with an average grade of 11.44% and 17.5 Mt with an average grade of 11.83%, respectively, considering a cut-off grade of 6%. The results obtained from these two methods are identical to each other and difference is less than 6%. This estimation is a requisite for improving present design of the mine with an objective of selective mining up to sea level of +1720.
