Reconstruction problem by maximum entropy method applied on a mixture of experts

Vincent Vigneron (1, 2) , Jean-Marc Martinez (1) , Marie-Christine Lepy (3) , Jean Morel (3)

1 SERMA - Service d’Études des Réacteurs et de Mathématiques Appliquées
2 CREL - Centre de Recherches et d'Etudes Logistiques
3 LNHB (CEA, LIST) - Laboratoire National Henri Becquerel (CEA, LIST)

Layered Neural Networks, which are a class of models based on neural computation, are applied to the measurement of uranium enrichment. The usual methods consider a limited number of γ-ray and X-ray peaks, and require previously calibrated instrumentation for each sample. But since, in practice, the source-detector ensemble geometry conditions are critically different, a mean of improving the above conventional methods is to reduce the region of interest ; this is possible by focusing on the KαX region where the three elementary components are present. The measurement of these components in mixtures leads to the desired ratio. Real data are used to study the performance of neural networks and training is done with a Maximum Likelihood Method. We show that the encoding of data by Neural Networks is a promising method to measure uranium 235U and 238U quantities in infinitely thick samples.

Mots clés: Layered Neural Network; neural network; signal processing; neural computation; uranium enrichment; detector geometry; Maximum Likelihood Method

Domaines: Traitement du signal et de l'image [eess.SP]; Physique Nucléaire Expérimentale [nucl-ex]; Instrumentations et Détecteurs [physics.ins-det]

Vincent Vigneron, Jean-Marc Martinez, Marie-Christine Lepy, Jean Morel. Reconstruction problem by maximum entropy method applied on a mixture of experts. International Conference on Engineering Applications of Neural Networks (EANN '95), Aug 1995, Helsinki, Finland. ⟨hal-00221538⟩