Prediction method of pipeline corrosion depth based on the correlation and Bayesian inference

Abstract

The number of samples for detecting corrosion characteristic value is difficult to reach a large enough size in practical engineering, which leads to the pipeline corrosion evaluation results tend to be aggressive. For this reason, the influence of sample size on the inference results was analyzed, and based on the Bayesian theory and the uncertainty of measurement, the Bayesian inference method for the pipeline corrosion depth under the condition of small sample size was proposed. Then, the correlation between the corrosion depth and the length was considered, and the prediction method of corrosion depth based on the correlation and Bayesian inference was developed. Thereby, the corrosion depths under different defect lengths was inferred with the pipeline corrosion detection data, and further the effectiveness of the method was verified. The results indicate that: the new method could better reflect the influence of sample size on the inference results, the prediction results are more conservative and consistent with the engineering experience, and so it is safer and more favorable to the engineering application. The research results could provide more accurate information to the prediction of pipeline corrosion depth, as well as theoretical reference to the prediction of the characteristic value of other corroded pipelines with consideration given to the correlation of random variables.

Keywords

• oil and gas pipelines
• corrosion depth
• interval inference method
• small sample size
• bayesian inference
• correlation