PSI - Issue 41

Mohamed Amine Belyamna et al. / Procedia Structural Integrity 41 (2022) 372–383 Mohamed Amine Belyamna et al. /Structural Integrity Procedia 00 (2022) 000–000

376

5

2.3. Residual Stresses Residual stresses influence both crack initiation and propagation. The calculations reported here concern the SCC behavior of the tree pipe sizes, presented in Table 1. The local residual stress at the pipe (size1and size 2) inside surface is treated as being normally distributed. The through—thickness distributions of stress are assumed to vary linearly between local values sampled at the inner and outer surfaces. For pipe size 3, the inner surface had a mean tensile stress of 262 MPa. The stress through the pipe wall changes from compressive in the inner quarter-wall thickness to tensile in greater depths (Guedri et al., 2009). The damage parameter IG SCC D  is a function of the stress, which consists of both the applied service-induced pressure, thermal, and residual stresses. The crack-tip stress intensity factor is given by Where ap K and res K are the stress intensity factor’s attributable to the applied stress and residual stresses, respectively. 2.4. Failure Criteria The net-section stress criterion is applicable to very tough material, and the failure is due to insufficient remaining area to support the applied loads given by equation (7), i.e., net-section stress due to applied loads becomes greater than the flow stress of the material f  . Where p A is the cross-section area of the pipe, cr A is the area of crack, LC  and f  are the load-controlled components of stress and the flow stress, respectively. The flow stress of the material f  used in equation (7) was taken to be normally distributed, with an expected value of 296 MPa and a standard deviation of 29 MPa. For leakage failure criterion was that a crack depth equal to the pipe wall thickness. 3. Proposed Probabilistic IG-SCC Model In this study, we discuss only the basic ideas of an ANNs. A more detailed introduction to ANN can be found in (Lippmann, 1987; Meireles, 2003). An ANN is a powerful method for problems that can't be expressed easily with a formula. It's widely known that the architecture and abilities of an ANN are usually considered analogous to the human brain. Gurney (1997) defined an ANN as an interconnected assembly of simple processing elements called nodes, and the processing ability of the network is stored in the interunit connection strengths called weights which are obtained by learning a set of training patterns. Various learning algorithms can be utilized depending on the nature of the training data and the expected output results. The main advantage of the ANN is a capability to learn information from samples. Additionally, ANNs possess the ability to implicitly detect complex nonlinear relationships between independent and dependent variables (Xu, 2017). Following training based on samples, ANNs can predict accurate solutions under any undefined inputs in many research fields. The calculation time and the complexity of the ANN used to predict the reliability of the corroded pipe is significantly influenced by the parameters of the selected model. Generally, these models are developed according to different parameters, such as geometric parameters, material parameters and parameters of operating conditions. The number of input neurons ANN is generally determined by the parameters of the limit state. i R is the internal radius of the pipe, h is the pipe wall thickness, ap res K K K   (6) LC h (2R h ) a A ab 2 ( ) R            i p i net   f   (7)