Abstract: To obtain the overall corrosion degree of tank bottom plates and reduce the cost of magnetic flux leakage detection, a calculation model for the corrosion rate of tank bottom plates based on the Weibull extreme-value distribution is constructed. Characteristic parameters are estimated using the Best Linear Unbiased Estimator (BLUE) method. Return period theory is introduced to realize extrapolation from sample extremes to population extremes, thereby predicting the minimum thickness and calculating the corrosion rate. The results show that the estimated corrosion rates of the middle plates and annular plates are 0.314 mm/a and 0.307 mm/a respectively; the predicted minimum thicknesses are all smaller than the minimum values from ultrasonic thickness measurement samples, demonstrating the conservatism of extreme-value extrapolation; the relative error between model predictions and magnetic flux leakage detection results ranges from -3.39% to 9.14%; the risk level of the tank bottom plates is medium to high; the traditional uniform corrosion evaluation method underestimates the actual corrosion risk. The findings can provide technical support for optimizing inspection strategies and rationally allocating inspection resources.

Key words: tank, Weibull distribution, corrosion rate, best linear unbiased estimate, regression period