Advanced DQM for a Drift Tube Detector

LCP mod.A - Final Project

Introduction

A novel approach for muons identification and track parameter estimation [1] consists in the implementation of an algorithm mixing artificial neural networks and analytical methods on a FPGA. This algorithm is being tested on a cosmic muon telescope at the Legnaro INFN National Laboratory (LNL), a detector composed by a set of drift-tubes (DT) reproducing a small-scale replica of those in use at CMS.

Recent developments in the search for new physics at LHC led to the implementation of a model-independent search strategy which exploits deep artificial neural networks [2, 3, 4]. The astonishing predictive power and flexibility of the New Physics Learning Machine (NPLM) algorithm can be conveyed to perform advanced Data Quality Monitoring (DQM) tasks [8].

Resources

• Description of the experimental setup → [5] chapter 1 (1.2.1 explains the actual configuration)
• How we track muons using scintillator signals → [5] section 2.2, 2.4 + chapter 3
• How the ML trigger algorithm reconstructs tracks (mean timer technique) → [6] (minimal summary) + [7] section 3.1 (complete documentation)
• Description of the NPLM algorithm → [8] chapter 2 (summary) + [2] (extended conceptual foundations)
• Application of NPLM to DQM → [8] chapter 3

Outline

1. Build a 2D dataset with t_drift and the crossing angle θ as features
   • using scintillator signals
   • using the ML algorithm reconstruction hidden within data
2. Study the correlation between the two features
3. Test the performance of NPLM using the 2D dataset built in 1.
   • NN architecture?
   • What is the average training time for the algorithm?
   • If we put a constraint on the crossing angle, does the algorithm detect the anomaly in the drift time distribution or it correctly sees the correlation between the two features?
   • What if we cut the angular feature but keep all the time information? Does it see it as a discrepancy?

---

Git Setup

1. Fork this repository clicking on the top-right button Fork.

2. Clone your forked repository → create a local repository in your machine.

   git clone https://<YourToken>@github.com/<YourUsername>/LCP_modA_finalProject.git

   where YourUsername it your GitHub username and YourToken is the token as copied from the GitHub webpage.

3. Get into the new folder:

   cd LCP_modA_finalProject/

4. Configure your username and email:

   git config --global user.name "<YourUsername>"

   git config --global user.email "<YourEmail>"

5. Define this repo as the upstream repository:

   git remote add upstream https://<YourToken>@github.com/niklai99/LCP_modA_finalProject.git

   Remember that in order to be able to push to the upstream you must be a contributor to this repo

6. Check

   git remote -v

7. Fetch for updates

   git fetch upstream

8. Check branches

   git branch -vv

Git Development Cycle

1. Sync the main branch that will have the latest completed code:

   git checkout main

   git fetch upstream

   git merge upstream/main

2. Before starting to code in your machine make sure everything is up to date:

   git pull

3. Now you can start developing code

4. Add files you want to commit (DO NOT add data files, weird folders and junk files):

   git add <NewFile>

5. Commit the tracked changes:

   git commit -m "<MeaningfulMessage>"

6. Push local changes into your remote repository on github (origin):

   git push origin main

7. Push local changes into the upstream repository:

   git push upstream main