Anomaly Detection in CMS HGCAL Using Radial Distribution Function and Autoencoder

The High Granularity Calorimeter (HGCAL) is a key component in the Phase-2 upgrade of the CMS detector at the LHC. Due to its unique hexagonal geometry, conventional Convolutional Neural Networks (CNNs) are not directly applicable. In this work, we explore the Radial Distribution Function (RDF) as an alternative feature representation to detect anomalies in HGCAL occupancy maps. We construct an Autoencoder model using both CNN and RDF features, compare their performance on Monte Carlo (MC) simulated datasets with injected anomalies, and demonstrate the superiority of RDF-based features in this context.