Grainger College of Engineering, All Events

View Full Calendar

Final Doctoral Defense: Jifu Zhao, PhD Candidate

Event Type
Seminar/Symposium
Sponsor
Department of Nuclear, Plasma, and Radiological Engineering
Location
111K Talbot Laboratory
Date
Apr 8, 2019   12:00 pm  
Speaker
Jifu Zhao, PhD Candidate
Cost
Free and Open to the Public
E-Mail
gwitmer2@illinois.edu
Phone
217-333-2295
Views
1
Originating Calendar
NPRE Events

Implementation and Simulation of Mobile Sensor Networks for Nuclear Radiation Detection

From preventing the threat of nuclear weapons proliferation to monitoring the transportation of special nuclear materials, nuclear radiation detection plays an important role in national security applications. However, changing background radiation, shielding effects, and short collection time make radiation detection a challenging problem. Anomaly detection, source localization, and isotope identification are three major parts of radiation detection. The concept of mobile radiation sensor networks, which utilize multiple mobile radiation detectors, has been proposed to solve these problems. This work mainly focuses on developing and testing methodologies for anomaly detection and radioactive source localization using mobile sensor networks. A collection of techniques and analyses for radiation detection are presented and evaluated. More specifically, in this work, a mobile sensor network simulation system is first developed to simulate the scenario where multiple radiation detectors move around a city. Based on the simulated data, the performance characteristics of mobile sensor networks for radiation detection are studied and quantified. Next, focusing on geospatial modeling of radiation count data, Poisson kriging is proposed to estimate background radiation level and perform anomalous source detection. The proposed method is validated using simulated source data injected in measured background radiation data and results indicate that the proposed algorithm can identify the anomalous radiation source with 90% accuracy under certain conditions. Additionally, source localization techniques based on maximum likelihood estimation are explored in detail. Simulation and experimental results show that source localization error can be reduced to be within 3 meters. Lastly, an exploratory study of spectrum-based anomaly detection techniques is presented. The performance of different machine learning techniques is evaluated and compared using simulated radiation data.

link for robots only