Assessment of insulated piping system inspection using logistic regression

Abstract

Corrosion under insulation (CUI) is a common problem not only in chemical process plants but also in utility and power plants. According to empirical study, CUI is mainly driven by the operating temperature where CUI is more susceptible when the equipment or piping system is operating between −12 and 121 °C. Other factors such as insulation type and equipment or pipe location are also seen to be the contributing factors to CUI. However, to date, it is not clear which factors are more important in contributing to CUI occurrence. This paper presents a methodology for predicting the likelihood of CUI occurrence for insulated piping system using a logistic regression model. Logistic regression, a special case of linear regression, requires binary data and assumes a Bernoulli distribution. Using historical data, the variables of operating time in year, pipe operating temperature, type of insulation and pipe size are modelled as factors contributing to CUI. The outcome of this model does not produce the probability of failure to be used in quantitative risk-based inspection (RBI) analysis. However, the result rather uses the historical inspection data to provide the decision makers with a means of evaluating which pipe to be inspected for future planning of scheduled inspection, based on the likelihood of CUI occurrence.