An Advanced Machine Learning and Artificial Intelligence System for Demonstrating Radiation Regulatory Compliance in DOE Accelerator Facilities

Department of Energy (DOE) facilities are required to demonstrate regulatory compliance for the radiation exposure of workers, general public, and environment. Compliance is generally demonstrated by monitoring radiation fields both onsite and offsite with both passive (dosimetry) and active, real-time monitors. DOE facilities utilize several traditional methods to demonstrate compliance with Federal limits. Calibrated, fixed radiation monitors may measure radiation doses in real-time or the monitors may integrate the dose with passive devices (e.g. area monitoring dosimetry). As part of experiment and facility designs, Monte-Carlo simulations (FLUKA, GEANT4, etc.) are often employed to determine the expected radiation dose at various key areas of the site. Routine surveys with hand-held radiation detectors supplement the fixed monitoring and the design simulation to ensure localized radiation fields are not above design limits. All of the approaches for demonstrating compliance rely on the expertise of individuals for interpreting data, configuring proper estimations, performing calculations, etc. Additionally, changes in facility configurations requires significant personnel effort in configuring equipment, reanalyzing data, and rerunning calculations. In this proposal, Applied Research LLC (ARLLC), Thomas Jefferson National Accelerator Facility (TJNAF or JLab), and Old Dominion University (ODU) propose the combination of domain knowledge (beam characteristics, fixed structural shielding, earthen burden (the soil and foliage added to the dome of the experimental halls as additional shielding), etc.), machine learning (ML) and/or artificial intelligence (AI) to correlate a variety of multi-modal onsite signals and the radiation fields seen in accessible areas of the accelerator site and the site boundary. The ML/AI will consider the complex influence of environmental parameters that affect the radon contribution of the measurements. We will focus on actual data obtained from JLab. In Phase 1, the coded beam and location data are fed into a deep learning model to predict doses at serval designated important locations in JLab’s facility. According to a market analysis and projection report, it forecasts the radiation detection, monitoring, and safety market to reach 2.3 billion by 2022 from an estimated 1.7 billion in 2017, at a CAGR of 5.7% during the forecast period. Our team consists of a small business focusing on pattern recognition, a university with strong expertise in AI and ML, and a national laboratory (JLab) with strong experience in radiation monitoring. Given the strong expertise of our team, the likelihood for success is extremely high. We expect that our annual revenue will be 5 million three years after the end of Phase 2.