Pipeline Lateral Buckling: Realistic Modelling of Geotechnical Variability and Uncertainty

Abstract

Abstract This paper describes and explores the effects of geotechnical variability and uncertainty on the design of pipelines for thermal buckling and axial walking. Geotechnical input parameters – ‘friction factors’ – are often assigned extremely wide ranges between low estimates (LE) and high estimates (HE), and these are assumed to apply uniformly along the pipeline in some design steps, or handled probabilistically in others. This paper clarifies the causes of geotechnical variability and uncertainty, identifying effects that are aleatory and epistemic. It proves useful to distinguish these effects on the basis of local and global influences that relate to short parts of the pipeline and the entire pipeline respectively. Field data from a case study pipeline is used to demonstrate these sources of PSI parameter variability and uncertainty. The aleatory (local) variability generally relates to the lay process and the resulting influence on the local soil strength and pipe embedment, and can also arise from in situ geotechnical heterogeneity. Epistemic (global) uncertainty relates to potential systematic errors in the assessment of soil strength and the average influence of the lay process. The epistemic (global) uncertainty budget also includes uncertainty related to the calculation models used to determine PSI parameters from soil properties, including their applicability for the site in question. Pipeline buckling design generally involves two types of analysis:probabilistic analysis of buckle initiation (and tolerability), usually using a Monte Carlo approach, e.g. using the BUCKFAST software;deterministic analysis of buckling, usually using finite element analysis. In the probabilistic analysis the full PSI LE-HE range is considered in each Monte Carlo realization, meaning the range is implicitly treated as aleatory (local). In the deterministic (FE) analysis a single set of PSI parameters are applied uniformly along the pipe meaning that the range is implicitly treated as epistemic (global). This study compares the calculated and observed as-laid embedment of the case study pipeline and shows the resulting range of lateral breakout and residual resistance. These parameters are calculated by a Monte Carlo method that yields the covariation of these parameters and forms a useful tool for design. Buckles tend to form at locations of low embedment and therefore low lateral breakout resistance which also have low residual lateral resistance, and can therefore tolerate higher feed-in. Conventional design practice treats initiation and tolerability as independent calculations, which is unnecessarily onerous. The influence of aleatory (local) variability is then explored using the lateral PSI parameters derived for the case study pipeline. FE analyses are firstly performed using LE, BE and HE lateral resistance applied uniformly along the pipe (mirroring conventional design practice). An additional analysis then uses the ‘real’ spatial variation in lateral resistance calculated directly from the embedment survey data. It is shown that the ‘real’ pipeline buckles at short (~20–40 m long) regions of local low lateral resistance. As these buckles lengthen, they experience a higher average lateral resistance. However, the resulting strains are significantly lower than calculated from the uniform HE friction case that is conventionally considered in design. These observations provide impetus for a more sophisticated treatment of geotechnical uncertainty and variability in pipeline buckling design. Current methods appear too onerous in some (but not all) respects. This paper provides techniques to reduce unnecessary conservatism, and to allow better definition and handling of geotechnical uncertainty and variability.