Sensor Fusion using Signal Processing and AI Tools Applied to CERN-CMS Detector
| Author | Aljamea, Hasan | |
|---|---|---|
| metadata.dc.contributor.department | University of Bahrain, College of Engineering | |
| metadata.dc.date.accessioned | 2026-09-20T15:44:51Z | |
| Publication date | 2024 | |
| Degree level | Master | |
| Abstract | Abstract : This thesis explores the application of Artificial Intelligence (AI) and Signal Processing (SP) techniques in the context of Sensor Fusion applied to the high-energy physics experiment at the Large Hadron Collider (LHC) of the European Centre for Nuclear Research (CERN). The central aim is to help in maintaining the integrity of the massive sensor array used at the Minimum Ionizing Particle Timing Detector (MTD) located at the inner layers of the Compact Muon Solenoid (CMS) and develop a user friendly platform to collect and compactly present the large sensor data to the monitoring operator. Hence, a major part of this work aims at detecting and raising warnings of potential operational issues before they escalate by applying different AI techniques for anomaly detection, namely various autoencoder architectures including Convolutional Neural Network (CNN), 2D Convolutional with Long Short-Term Memory (Conv2DLSTM), CNN with U-net architecture, and Conv2DLSTM with U-net architecture. The latter method turned out to provide the best performance with an accuracy of 96.2% on simulated data. Computer simulations indicate that this algorithm is quite robust against additive noise as its performance was found to degrade by only about 1% with a Signal-to-Noise Ratio (SNR) decreasing from 17dB to 5dB. The wavelet denoising algorithm was investigated to possibly mitigate this noise effect in light of the harsh operating environment at CMS with a marginal success since for low SNR sensor signals it caused the anomalies to be hidden hence missed by the algorithm. Obtaining practical data from the sensor array was a rather challenging task, and as of the writing of this work only partial data was obtained. Initial results of applying the algorithm to the partial real data in hand showed a comparable accuracy of 94.5 %. The MTD services developed in this work using FastAPI Python library and xDAQ C++ Framework as well as the proposed anomaly detection methods should contribute to safely running the MTD close to its optimum point. Keywords— CERN, CMS, MTD, DAQ, Anomaly Detection, Incremental Learning | |
| metadata.dc.identifier.other | https://digitalrepository.uob.edu.bh/id/e7b3e04c-c498-4ed2-85ea-4f0be8f3f946 | |
| metadata.dc.identifier.other | Drupal node 2265 | |
| metadata.dc.identifier.other | https://digitalrepository.uob.edu.bh/en/dar/sensor-fusion-using-signal-processing-and-ai-tools-applied-cern-cms-detector | |
| metadata.dc.identifier.uri | http://165.227.56.122/handle/123456789/6218 | |
| Language | en | |
| Link | https://digitalrepository.uob.edu.bh/en/_flysystem/fedora/2024-11/sensor-fusion-using-signal-processing-and-al-tools-applied-to-cern-cms-detector.pdf | |
| Title | Sensor Fusion using Signal Processing and AI Tools Applied to CERN-CMS Detector | |
| Type | Master's Thesis |
Files
Original bundle
1 - 1 of 1
Loading...
- Name:
- node-2265-1-sensor-fusion-using-signal-processing-and-al-tools-applied-to-cern-cms-detector.pdf
- Size:
- 9.26 MB
- Format:
- Adobe Portable Document Format