Corrosion Prediction from Polarization Scans Using an Artificial Neural Network Integrated with an Expert System

Artificial neural networks are forms of artificial intelligence that learn correlative patterns between input and output information without a specific model. Then, they use the learned relationships to make predictions. An artificial neural network was constructed to recognize certain relationships in potentiodynamic polarization scans to predict if crevice corrosion, pitting, and general corrosion are possible concerns. The network so constructed was shown to make appropriate predictions using scans not included in the original training. The resulting network was incorporated within an expert system to provide an easy way to input data, to provide simple consistency checks, and to interpret the numerical output of the neural network to make the final prediction.