Purdue Reactor Integrated Machine Learning (PRIMaL) Dataset

Purdue University Reactor Number One (PUR-1) was recently upgraded with a fully digital instrumentation and control system and is now capable of providing real-time multivariate research data with sampling up to the millisecond for over 2000 different parameters. These parameters cover a wide spectrum, ranging from process measurements to network traffic, as well as derived values such as the system change rate, throughout many reactor states such as startup, steady state, rod movement, and shutdown conditions. The proposed Purdue Reactor Integrated Machine Learning dataset (PRIMaL) encompasses 22 parameters with more than 1 million datapoints corresponding to various reactor operation cycles and curated to serve as unique classification and time-series forecasting benchmarks of various degrees of complexity. This dataset represents an entry point for researchers to bring artificial intelligence and machine learning (AI/ML) techniques into the world of nuclear engineering and beyond.