RECOGNIZING ELETROCHEMICAL NOISE PATTERNS FROM MILD STEEL CORROSION IN OIL-WATER MIXTURES USING NEURAL NETWORKS

摘要

A two-phase system is experimentally simulated to produce bubble and slug flows, similar to thosernfound in pipelines. Different hydrocarbon/electrolyte test fluid mixtures in the range from 99.9/0.02 torn85/15 ratio by volume were prepared from diesel and 3% NaC1 solution. Electrochemical current noisern(ECN) is recorded by using a rotating system with 3 mild steel electrodes embedded in "activated" resinrnunder various flow conditions and test fluid mixtures. Depending on flow intensity and electrolyterncontent, ECN presents different types of signals, which are related to the wetting of the metal surfacernand corrosion intensity. Three types of neural networks have been tested to identify 3 signal patternsrnobtained from the experimental system as part of a corrosion monitoring system for water in oil systemsrnwhere early and systematic corrosion detection is desirable.