ARTIFICIAL NEURAL-NETWORK PREDICTIONS OF DEGRADATION OF NONMETALLIC LINING MATERIALS FROM LABORATORY TESTS

Artificial neural networks are computer simulations that have the pote ntial to find the same patterns that corrosion practitioners recognize to relate experimental test results to lifetime predictions. This pot ential was used to construct an artificial neural network to recognize the pattern between results from a sequential immersion test for orga nic nonmetallic lining materials and their ability to function as lini ngs in actual applications. The network was shown to predict field per formance. The network was incorporated within an expert system to simp lify data input and output, to allow for simple consistency checks bet ween sample appearance and network output, and to make the final predi ction.