Corrosion Predication in Refinery Based on Not Equal Time Interval Grey System Model

Abstract

The equipment exposed to wash erosion seriously when the refinery process high-sulfur crude oil. Used FLUENT to emulate fluid field can help us to find the place to monitor. Measured the wall thickness of the elbow on the entrance of the atmospheric distillation tower top exchangers by ultrasonic thickness meter every month, established standard GM(1,1) model to predict the corrosion is effective, but when the corrosion data are not equal time interval ranged, the standard GM(1,1) model is out of function. Rebuilt data sequence, Established not equal time interval model with grey system theory, used the data of 1~4 and 4~10 month to predict the wall thickness of 18 month to test the model, the average relative error is 1.3%. Compared mean relative error, absolute degree of incident, ratio of mean square deviation, lower probability of error with standard GM(1,1) model and improved GM(1,1) model, though the precision of not equal time interval GM model is a little low, but it still passed error check, is eligible.