Electrical & Computer Engineering, Department of
First Advisor
Michael Hoffman
Second Advisor
Sina Balkir
Date of this Version
Summer 7-2023
Citation
A. Inamura, Asset Cueing Nuclear Radiation Anomaly Detection Using An Embedded Neural Network Resource, M.S. Thesis, Department of Electrical and Computer Engineering, University of Nebraska-Lincoln, 2023.
Abstract
Nuclear radiation detection is inherently a challenging task, coupled with a high background variation or increase in anomalies, the accuracy for detection can plummet. A key factor in the success of nuclear detection hinges on the sensor’s ability to generalize its model and directly leads to the model’s robustness. The goal of this project is to develop algorithms suitable for use on the University of Nebraska-Lincoln’s Pingora chip, a low-power, system-on-chip device with an active neural processing unit (NPU) made for nuclear radiation detection. The thesis aims to improve Pingora’s overall generalization ability in nuclear radiation source detection. A multiphase multi-layer perceptron neural network (MLPNN) design was used to train the network offline until a low error rate was achieved. The development dataset includes over 100,000 samples with varying source presence. The difficulty of working with this dataset was the high variation in the data characteristics for both background and source samples. Regardless, the model achieved on average a 12% error across all test sets, including the worst-case dataset, which was defined as the dataset that includes the least identifiable characteristics.
Advisors: Michael Hoffman and Sina Balkır
Comments
A THESIS Presented to the Faculty of The Graduate College at the University of Nebraska In Partial Fulfillment of Requirements For the Degree of Master of Science, Major: Electrical Engineering, Under the Supervision of Professors Michael Hoffman and Sina Balkir. Lincoln, Nebraska: July, 2023.
Copyright © 2023 April Inamura