USE OF ARTIFICIAL NEURAL NETWORK MODELS FOR PREDICTING COATED COMPONENT LIFE FROM SHORT TERM EIS MEASUREMENTS

摘要

The objective of this study was to relate the results from electrochemical impedance spectroscopy (EIS) collected in 24 hours to salt spray exposure data collected over 1500 hours for conversion coated metal surfaces. To develop such a relationship, an approach based on artificial neural networks (ANNs) was used. The output of this study was a matrix of weights and threshold values that predicted the salt spray performance of coated components based on EIS results. A model based on phase angle data input from EIS measurements collected after 24 hours exposure to 0.5M NaCl was able to account for 85 percent of the variation in the salt spray time to failure from a randomly selected subset of the sample population. This exercise illustrates the utility of ANNs in corrosion prediction and suggests that they may play a key role in making lifetime predictions for components in service based on laboratory measurements. Conversion coated substrates and salt spray exposure data used in this study were furnished by the National Center for Manufacturing Sciences Ann Arbor MI, as part of the Alternatives to Chromium in Metal Finishing Study [1].