Discrete wavelet analysis of external corrosion clusters on pipelines for burst capacity prediction

Abstract

The present study applies the two-dimensional discrete wavelet transform to analyze naturally occurring corrosion clusters on external surfaces of in-service steel transmission pipelines. The corrosion depths in a cluster are measured by high-resolution laser scans. The Daubechies wavelet with two vanishing moments (DB2) is adopted in the analysis, and the widely used RSTRENG model is employed to compute the burst capacity of pipelines with corrosion clusters. A methodology is proposed to determine level- and sub-band-dependent thresholds such that those wavelet coefficients below the thresholds have a negligible impact on the burst capacity and can be ignored for the reconstruction of the cluster. Statistical analyses of the preserved wavelet coefficients indicate that most of the features in the natural corrosion clusters are circumferential features, but their influence on the burst capacity is less significant than longitudinal features. Features that have large impacts on the burst capacity prediction are approximately 4–16 mm in width and 4–32 mm in length. It is unnecessary to decompose the natural corrosion cluster to a scale greater than 32 mm because the wavelet coefficients at higher levels have a negligible impact on the burst capacity prediction.