Inductive thermodynamics from time series data analysis
pp. 384-394
in: Setsuo Arikawa, Ayumi Shinohara (eds), Progress in discovery science, Berlin, Springer, 2002Abstract
We propose an inductive thermodynamics from time series data analysis. Using some modern techniques developed in Statistical Science and Artificial Intelligence, we construct a mathematical model from time series data. We introduce an effective potential for a steady state distribution and construct thermodynamics following the recent work by Sekimoto-Sasa. We apply our idea to neutron noise data from a test nuclear reactor. We interpret the slow transformation of the control bar as work. The response to the transformation appears as excess heat production in accordance with the second law.