Issue 42

G. Testa et alii, Frattura ed Integrità Strutturale, 42 (2017) 315-327; DOI: 10.3221/IGF-ESIS.42.33 316 Under these varying operating conditions, a safe pipeline requires a combination of design and operational measures. It is recognized that design and operational measures are not mutually exclusive, and the interacting combinations must be appropriately considered. Conventional pipeline design uses the Allowable Stress Design (ASD) approach, which limits pipeline stress to a prescribed fraction of its specified minimum yield strength, such as 72% of the yield strength in hoop direction and 90% of the yield strength for combined hoop and longitudinal stresses. This criterion, while appropriate for buried pipelines, is difficult to satisfy for pipelines that must withstand ground movements or buckling due to operating conditions. Moreover, the ASD approach makes no distinction between load-controlled and displacement-controlled conditions, between stable and unstable failure modes, or between the loss of serviceability and loss of pressure containment. Thus, the safe-design process requires developing an understanding of the strains imposed on the pipe (strain demand) and the safe strain limits that the pipe can withstand without failure (strain capacity). Recognizing these limitations, an increasing number of industry standards allow application of strain-based design for loading conditions outside those typically considered for more conventional pipelines. Strain-based design is a specific application of a limit-state design approach: here, the capacity of the pipeline to withstand longitudinal strain without failure is quantified and compared to the strain expected in service under displacement-controlled conditions [2]. Quantification of the strain demand side of the design condition requires a comprehensive understanding of the pipeline route, which includes soil characteristics and geothermal analysis, wall-thickness differences, and mechanical properties of adjacent pipe joints [3]. Tensile strain capacity is generally governed by the strain capacity of the girth weld region and it is determined through a combination of tests and finite element analysis or semi-empirical models. These models are typically based on attempts to modify fracture assessment criteria in the form of failure assessment diagrams [4]. They often lead to highly conservative criteria with large uncertainties. These diagrams imply the presence of a plastic collapse load and, if the pipeline stress-strain response is relatively flat in the plastic region, are not representative of strain capacities. In strain-based design, fracture resistance is usually assessed by CTOD fracture toughness. As for other fracture mechanics concepts, the CTOD fracture criterion is valid only when some conditions are satisfied. CTOD controlled crack growth under plain strain condition is ensured when:   0 , cr b B a W a       (1) where  = 50 and  = 0.1 for bend and compact tension specimens, b is the specimen ligament, B is the thickness and W the specimen width [5]. Gordon et al. [6] showed that the limits for the CTOD controlled crack growth are material dependent. Based on experimental measures on different material grades, it was postulated that the crack growth limit for CTOD controlled crack growth in R-curves is 15% of the initial uncracked ligament, although this condition alone would not be sufficient to ensure size/geometry independent results. Today, in ASTM 1820 the  limit is reduced to 35 [7]. In high toughness material grades operating over the mid-to-upper end of the ductile-to-brittle (DTB) transition region, these limits may not be fulfilled and, therefore, material resistance can be even strongly affected by the loss of constraint occurring at the crack tip. In these cases, fracture resistance and R-curve shows a significant geometry dependence that may hinder the transferability from laboratory samples to full-scale components [8]. When dealing with large plastic deformation, damage modelling is a valid alternative to investigate and predict material failure. Recently, the use of finite element analysis simulating crack growth by means of damage modelling has been introduced in DNV design recommendations for submarine pipeline systems [9]. In the literature, several examples using porosity models (Gurson-Tvergaard-Needleman (GTN), and Rousselier) are available. Fehringer et al. [10] investigated the stress triaxiality effect on the strain capacity of 20MnMoNi5-5 grade steel calibrating Rousselier model parameters on round notched bar tensile test results and validating predicting crack growth in CT-0.5T samples. Acharyya and Dhar [11] calibrated the GTN model parameters based on compact tension (CT) specimen fracture data and then used the model to predict crack growth in circumferentially cracked pipes. Xu et al. [12] performed a systematic numerical investigation predicting circumferential crack growth in pipes with different sizes and properties using the complete GTN model finding a good transferability to single edge notch in tension (SENT) specimen. Geometry transferability is a major requirement for any micromechanical model. The GTN suffers the transferability of model parameters to different stress triaxiality [13]. For this reason, crack data of laboratory samples with constraint similar to that expected in full-scale components are necessary for model parameters calibration. Furthermore, numerical solutions obtained with porosity models show mesh dependency because of softening in the flow curve caused by damage. Because of this, a reference element length is often introduced as an additional material parameter.