Physics-based Digital Twins for Autonomous Thermal Food Processing: Efficient, Non-intrusive Reduced-order Modeling
Authors:
Maximilian Kannapinn,
Minh Khang Pham,
Michael Schäfer
Abstract:
One possible way of making thermal processing controllable is to gather real-time information on the product's current state. Often, sensory equipment cannot capture all relevant information easily or at all. Digital Twins close this gap with virtual probes in real-time simulations, synchronized with the process. This paper proposes a physics-based, data-driven Digital Twin framework for autonomou…
▽ More
One possible way of making thermal processing controllable is to gather real-time information on the product's current state. Often, sensory equipment cannot capture all relevant information easily or at all. Digital Twins close this gap with virtual probes in real-time simulations, synchronized with the process. This paper proposes a physics-based, data-driven Digital Twin framework for autonomous food processing. We suggest a lean Digital Twin concept that is executable at the device level, entailing minimal computational load, data storage, and sensor data requirements. This study focuses on a parsimonious experimental design for training non-intrusive reduced-order models (ROMs) of a thermal process. A correlation ($R=-0.76$) between a high standard deviation of the surface temperatures in the training data and a low root mean square error in ROM testing enables efficient selection of training data. The mean test root mean square error of the best ROM is less than 1 Kelvin (0.2 % mean average percentage error) on representative test sets. Simulation speed-ups of Sp $\approx$ 1.8E4 allow on-device model predictive control.
The proposed Digital Twin framework is designed to be applicable within the industry. Typically, non-intrusive reduced-order modeling is required as soon as the modeling of the process is performed in software, where root-level access to the solver is not provided, such as commercial simulation software. The data-driven training of the reduced-order model is achieved with only one data set, as correlations are utilized to predict the training success a priori.
△ Less
Submitted 7 September, 2022;
originally announced September 2022.
Reference material for natural radionuclides in glass designed for underground experiments
Authors:
P. P. Povinec,
M. K. Pham,
J. Busto,
C. Cerna,
D. Degering,
Y. Hamajima,
K. Holy,
M. Hult,
M. Jeskovsky,
M. Koehler,
A. Kovacik,
M. Laubenstein,
P. Loaiza,
F. Mamedov,
C. Marquet,
J. Mott,
M. Mullerova,
F. Perrot,
F. Piquemal,
J. -L. Reyss,
R. Saakyan,
H. Simgen,
B. Soule,
J. Stanicek,
I. Sykora
, et al. (1 additional authors not shown)
Abstract:
A reference material designed for the determination of natural radionuclides in solid samples (glass pellets) is described and the results of certification are presented. The material has been certified for 7 natural radionuclides (40K, 226Ra, 228Ra, 228Th, 232Th, 235U and 238U). An information value is given for 210Pb. Radon (222Rn) emanation experiments showed results comparable within participa…
▽ More
A reference material designed for the determination of natural radionuclides in solid samples (glass pellets) is described and the results of certification are presented. The material has been certified for 7 natural radionuclides (40K, 226Ra, 228Ra, 228Th, 232Th, 235U and 238U). An information value is given for 210Pb. Radon (222Rn) emanation experiments showed results comparable within participating laboratories, however, the number of data and precision was too low to carry out a certification process. The reference material may be used for quality management of analytical laboratories engaged in the high-sensitive analysis of radionuclides in the construction materials of detectors placed in ultra low background underground laboratories.
△ Less
Submitted 12 July, 2015;
originally announced July 2015.
