A generalised neural-fuzzy well log interpretation model with a reduced rule base

A generalised neural-fuzzy interpretation model combining the advantages of neural networks and fuzzy logic as an interpretation tool for well-log data analysis is reported in this paper. The model makes use of an artificial neural network to learn the underlying function from a set of training data and then it is used to generate a fuzzy rules-based model. The fuzzy rules-based model enables a log analyst to gain a better understanding of the model. Furthermore, the rule set may be manipulated to modify the performance of the model by incorporating the experience or knowledge of the analyst. However, the number of fuzzy rules generated can be very large. A method is proposed to substantially reduce this number to suit the conditions of the well under investigation. This allows easier interaction between the operator and the model while maintaining prediction accuracy