Radiological clearance at CERN: classical machine learning methods for waste classification

Area of Science:

• Nuclear physics
• Waste management
• Machine learning applications

Background:

• High-energy particle accelerators at CERN produce potentially radioactive waste.
• Radioactive waste requires clearance procedures for release from regulatory control.
• Characterizing waste is complex due to varied activation scenarios, radionuclides, and materials.

Purpose of the Study:

• To develop a data-driven probabilistic method for radiological waste characterization.
• To simplify clearance procedures for radioactive waste from particle accelerators.
• To define threshold values for clearance eligibility of irradiated electric cables.

Main Methods:

• Combining physics-based activation simulations with classical machine learning algorithms.
• Applying the developed method to irradiated electric cables from CERN.
• Focusing on measurable quantities: activity, dose rate, and mass.

Main Results:

• A probabilistic approach was successfully applied to irradiated electric cables.
• Threshold values for clearance eligibility were defined based on measurable quantities.
• The method simplifies measurement and decision-making for waste clearance.

Conclusions:

• The data-driven probabilistic method enhances the efficiency of radioactive waste management at CERN.
• The approach aligns with international guidance for waste clearance procedures.
• This method offers operational advantages for regulatory control release.