PSI - Issue 41

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

382

11

Fig. 9. Architecture of the optimized ANN.

Fig. 10. Importance of input parameters using the second optimized ANN.

Figure 11 presents the predicted and actual leak probability versus

IG SCC D  for 3 pipe sizes using the second

optimized ANN and confirm that IG SCC D  can be used to generalize the PFM model. The ANN approach has shown effective and flexible performance for estimating the pipe reliability because of its powerful capacity to deal with complex and nonlinear relationships between multiple parameters.

IG SCC D  for 3 pipe sizes using the second optimized ANN.

Fig. 11. Predicted and actual leak probability versus

5. Conclusion This paper proposed a reliability method integrating the MCS, sensitivity analysis, and ANN approach to assess the reliability of the pipeline with IG-SCC. The MCS generates the reliability data and input parameters for the ANN modeling and training. The trained ANN can be used to effectively and accurately estimate the reliability. The proposed reliability method is illustrated by three examples of AISI 304 SS pipes subject to IG-SCC. The trained ANN can be used to effectively and accurately estimate the initiation and leak reliability of damaged pipes. The computing time needed for the ANN method and for MCS to obtain the reliability curve shown in Figure 4, which consists of 40 probability values one printing the results for each year, is 1.3s and 44s respectively. Thus,