Abstract
In this paper, an expert system has been outlined to classify the defects in metallic petroleum pipelines using acoustic techniques with non-destructive evaluation (NDE) protocols. the proposed system maps the quantitative defect data through a novel perception-based kernel that has its roots in multidimensional fuzzy set theory to map the relative weights given to various features; mathematical or statistical, to the decision surface to deduce the type of the defect. The system has a centralized database which holds the defect information in the form of known and calculated features. The known features and their quantitative representations are used to initialize the database. Then experiments are conducted on known defects and the collected experimental data is then modeled into autoregressive process models using state of the art h. deconvolution algorithm. With each feature set, a classifier tag is associated that assigns a class number to that defect. The classifier tag is then used to classify any new data using the fuzzy classifier.