On hybrid-fuzzy classifier design: An empirical modeling scenario for corrosion detection in gas pipelines

In this paper, a customized Fuzzy Inference System is presented to classify the corrosion and distinguish it from the geometric defects or normal state of the steel pipes used in gas/petroleum industry. The presented strategy is hybrid in the sense that it utilizes both the soft computing as well as conventional parametric modeling through H
_infin
optimization methods. An experimental strategy is first outlined through which the necessary data is collected as A-scan which are the ultrasonic echoes pulses in ID. Then, using empirical modeling approach a parametric transfer function is obtained for each pulse. In this respect, each A-scan is treated as an output from a defining function when a pure metal's A-scan is used as its input. Three defining states are considered in the paper; healthy, corroded, and defective, corresponding to the healthy or very much less corroded metal, corroded metal, and metal with any artificial or other defects, respectively. Impulse responses for each of these parametric models are plotted and human heuristics is then utilized in coming up with a set of quantitative features that can be used in distinguishing these classes. This feature set is then supplied to the Fuzzy Inference system as input to be used in distinguishing various classes under study. The main contribution of this work is to elaborate the fact that corrosion modeling provides easier approach in classifying the A-scans better rather than the raw A-scan data which is more prone to noise errors and more dependent on the measuring device's parameters.