EXTRACTING KNOWLEDGE FROM CARBON DIOXIDE CORROSION INHIBITION WITH ARTIFICIAL NEURAL NETWORKS

The artificial neural network (ANN) has a proven reputation of accurately modeling the interacting relationships in a complex non-linear system. However, an ANN model is often considered a "black-box" in the sense that its estimates appear incomprehensible. This limitation is alleviated by using knowledge extraction techniques and algorithms. Better understanding of these relationships is significantly important to the oil industry, where the factors that affect corrosion are not well understood. To provide insight, this paper presents a number of different techniques to extract knowledge from an ANN trained with a CO(2) corrosion dataset. These techniques include Network Interpretation Diagrams, Garson's Algorithm, Sensitivity Analysis, Family of Curves and Surfaces, and TREPAN-Plus. From a knowledge-based perspective, these methods can provide the oil industry with the ability to determine the role of input variables in predicting corrosion inhibition. The limitations and advantages of each of these techniques are also discussed.