Core axial power shape reconstruction based on radial basis function neural network

Abstract

The core axial power shape reconstruction method based on radial basis function neural network (RBFNN) is proposed, and 18-node axial core power shape can be reconstructed from 6-segment in-core detector signals or 6-segment ex-core detector signals. Alternating conditional expectation algorithm is also used to validate the effectiveness of RBFNN algorithm. The results show that no matter what kind of detector is used, the RBFNN algorithm performs better than the ACE algorithm. The RBFNN algorithm has a good anti-noise ability when the in-core detector signals with noise are used, while the correct axial power shape can’t be reconstructed from the ex-core detector signals with noise. By analyzing the ex-core axial spatial weighting function’s condition number, the determination of the axial power shape from ex-core detector signals is found to be a typical ill-posed inverse problem. A regularized RBFNN algorithm is used to eliminate this ill-posedness and get the physically meaningful axial power shape.