READ ME File For 'Dataset for Accurate and Efficient Design and Optimization of Thermoe-lectric Cooler via Machine Learning Technologies'

Dataset DOI: 10.5258/SOTON/D3741

ReadMe Author: Ruomeng Huang, University of Southampton 

This dataset supports the publication: Accurate and Efficient Design and Optimization of Thermoe-lectric Cooler via Machine Learning Technologies
AUTHORS: Haoji Yang, Yuxiao Zhu, Harold Chong and Ruomeng Huang
TITLE: Accurate and Efficient Design and Optimization of Thermoe-lectric Cooler via Machine Learning Technologies
JOURNAL: Applied Thermal Engineering
PAPER DOI IF KNOWN: 10.1016/j.applthermaleng.2025.129224


This dataset contains:
The raw data of figure 1 to 8.

The figures are as follows:

Figure 1. (a) The schematic of the TEC model applied in this study. (b) Architecture of the ANN developed to predict the cold-side and hot-side temperatures. The input contains the geometrical conditions (W_TE,H_TE) and operating conditions (I, T_amb, h, ρ_c, ρ_t, and Q_in/A), the output layer contains the cooling performance values (T_C and T_H). (c) Flowchart of the GA-based TEC optimization, using either the ANN or COMSOL simulation for the fitness function calculation. (d) Flowchart of the GA-based TEC current optimization with a two-stage objective. 
Figure 2 (a) Training loss curve of the final ANN model. Parity plot comparing the predicted (b) T_C and (c) T_H from the ANN model with values obtained from COMSOL simulations.
Figure 3 Comparison of cold-side temperature T_C values predicted by the ANN model (lines) and COMSOL simulations (dots) under varying (a) heat input density Q_in⁄A, (b) convective coefficient h, and (c) ambient temperature T_amb. For each parameter sweep, other parameters are fixed as Q_in⁄A=1 W⁄〖cm〗^2 , h=1750 W⁄(m^2∙K), T_amb=300.15K, ρ_c=10^(-9) Ω∙m^2, ρ_t=10^(-5)  (m^2∙K)⁄W, W_TE=3 mm and H_TE=2 mm.
Figure 4. Comparison of cold-side temperature T_C values predicted by the ANN model (lines) and COMSOL simulations (dots) under varying (a) leg height H_TE, (b) leg width W_TE, (c) electrical contact resistivity ρ_c, and (d) thermal contact resistivity ρ_t. For each parameter sweep, other parameters are fixed as Q_in⁄A=1 W⁄〖cm〗^2 , h=1750 W⁄(m^2∙K), T_amb=300.15K, ρ_c=10^(-9) Ω∙m^2, ρ_t=10^(-5)  (m^2∙K)⁄W, W_TE=3 mm and H_TE=2 mm. 
Figure 5. Comparison of the T_(C,min) values predicted by the ANN model (lines) and COMSOL simulations (dots) under varying (a) Q_in⁄A with different h, (b) Q_in⁄A with different T_amb, (c) h with different T_amb, (d) ρ_c with different ρ_t, and (e) W_TE with different H_TE conditions.
Figure 6. Mapping result of the (a) T_C value against varying Q_in/A and I conditions, (b) T_(C,min) value against various T_amb and h conditions, and (c)T_(C,min) value against various H_TE and W_TE conditions.
Figure 7. Convergence curve of genetic algorithm for minimum cold-side temperature under an operating condition of (a) Q_in⁄A = 0.6 〖W/cm〗^2, (b) h = 1250 〖W/(m〗^2·K). Other parameters are fixed as T_amb=300.15K, ρ_c=10^(-9) Ω∙m^2, ρ_t=10^(-5)  (m^2∙K)⁄W. T_(C,min) optimized by GA coupled with ANN (red dots) and COMSOL simulation (blue dots) as a function of (c) heat input density Q_in⁄A, (d) convective coefficient h.
Figure 8. Dynamic TEC current optimization results under three different scenarios: (a) Scenario 1 – unconstrained optimization aiming to minimize cold-side temperature (T_C); (b) Scenario 2 – optimization with a target temperature constraint of 303.15 K; (c) Scenario 3 – dual-objective optimization minimizing power consumption while maintaining T_C below 303.15 K. For each scenario, the corresponding heat input density Q_in/A, optimized cold-side temperature T_C and current I are shown. The fixed parameters include h=2500 W⁄(m^2∙K), T_amb=300.15K, ρ_c=10^(-9) Ω∙m^2, ρ_t=10^(-5)  (m^2∙K)⁄W, W_TE=3 mm and H_TE=3.5 mm.


Date of data collection: Jan 2024 to April 2025

Information about geographic location of data collection: United Kingdom

Licence: No

Related projects: EPSRC and AWE Ltd. for the ICASE studentship No. 16000087


Date that the file was created: June, 2022