Innovative methodology for predicting pipe thickness loss based on electromagnetic flaw detection results

Abstract

The paper proposes an innovative approach to forecasting based on the results of electromagnetic flaw detection to determine maximum pipe thickness losses over 10-m intervals of the immersion depth of the lower part of a pipe into a well. The main model uses rank transformation applied to a fuzzy regression model with fuzzy input variables and fuzzy output. The input variables include the main parameters of the reservoir in the well (temperature, density, and dynamic viscosity of the hydrocarbon mixture), while the output variable is the maximum loss of pipe thickness in the above–mentioned immersion intervals of the lower part of the pipe. The error of the output forecasts is determined using a numerical procedure to estimate the difference between fuzzy numbers. The forecasting is performed using a sliding method that combines rank fuzzy regression with clear nonlinear regression until the prediction error reaches the specified threshold value. The use of fuzzy regression with fuzzy input and output makes it possible to assess the impact of reservoir parameters on the condition of pipes and the potential for emissions and critical situations. Keywords: Fuzzy LR-type numbers; rank transformation; fuzzy regression; electromagnetic flaw detection; fuzzy outliers; sliding prediction; membership function; pipe thickness loss.