Identification of Core Anomalies via Neutron Noise under Multiplicative Detector Signal Distortion

Other conference contribution, 2026

This work addresses the identification of anomalies in nuclear reactor cores through the solution of an inverse problem based on neutron noise. Two methodologies are investigated: Exhaustive Search (ES) and Artificial Neural Networks (ANNs). The performance of both approaches is evaluated in the case of detector signals affected by a distortion. This distortion is modeled as a multiplicative component dependent on the signal itself. To reduce sensitivity to measurement quality, we introduce within the ES framework a Bayesian approach to produce solutions in terms of probability density function in the solution space. In addition, we develop a denoising technique that reconstructs physically consistent clean signals. Results show that ES achieves the highest accuracy and robustness when combined with signal reconstruction, and the ANN model trained on distorted data maintain stable performance across all distortion levels.