rafaelhuntstokes/lartpc

A simulation of DUNE's far detector.

LArTPC

A simulation of DUNE's far detector. This software package was created as part of an MSci research project, under the supervision of Professor Ryan Nichol @ UCL. The simulation takes event data, generated in GEANT4, of CC- supernova neutrino/argon interactions, and uses it to output TPC images. These images are sensative to a variety of the physics parameters of the detector, such as electron lifetime, diffusion coefficients, electron velocity, radioactive activity and electronic noise.

The simulation outputs a set of supernova neutrino event images and images comprised only of noise. A convolutional neural network is used to investigate classification accuracy under a range of different conditions.

The Files

1. simulation_tpc.py - the simulation code that creates the TPC images
2. model_utils.py - code that creates CNN model and includes test/train/validation methods
3. run_model.py - code uses CNN defined in model_utils.py and processes the output to create confusion matrices and graphs of results. Allows specification of inputs to CNN.

NOTE: the required GEANT4 data for the simulation, electron_data.npy, is too large to upload here. A smaller subfile containing a few events will be uploaded shortly.