Multi-MFL Measurement Assessment using Gaussian Mixture Model

Magnetic flux leakage (MFL) is the most popular in-line inspection (ILI) technique to inspect pipeline corrosion. However, the random measurement error of MFL cannot be negligible due to measurement variations. Since the ILI is usually conducted with two different types of MFL devices, data from two MFL systems are compared to assess the measurement result in this paper. Because the difference between two measurements consists of both random and systemic errors associated with individual MFL system, measurement conversion is conducted first to eliminate the influence of the systemic error. In this paper, a Gaussian probability density function is employed to represent the measurement data because of the randomness of measurement error. Then, the process to eliminate the systemic error between two systems becomes the alignment of two Gaussian mixtures. Besides, the L2 distance is employed as the distance measure to align Gaussian mixtures and assess the measurement confidence. The validity of proposed method is proven in the numerical example.