Study on the Integration of the Creep Law from Artificial Neural Network Methodology in the Delayed Hydride Cracking Assessment of the CANDU Pressure Tubes

The paper objective is to implement the creep law in the assessment of Delayed Hydride Cracking (DHC) for CANDU pressure tubes, following the process-zone methodology stated in the CAN/CSA N285.8 standard. A complex flaw has been chosen, represented by a Bearing Pad Fretting Flaw (BPFF) that contains also a Debris Fretting Flaw (DFF), for the analyses. For the stress relaxation, at the flaw tip, it is used for the first time a creep law, which has been developed at RATEN ICN, based on the artificial neural network (NN) methodology. The NN model is called Multilayer Feedforward Neural Network, and it gives an explicit relationship for the radial strain rate of the Zr-2.5%Nb pressure tube. A finite element model was implemented in specialized software, VisualFEA, to obtain the stress-strain field and the process zone methodology was run in a MATLAB environment employing specific scripts. As results, the convergent solutions are obtained after some trials for the process-zone displacements. These allow evaluating the prevention against DHC from the specified complex flaw of the CANDU pressure tubes in operation after certain effective full power hours (EFPH). (author).