Parametric optimization and prediction of enhanced thermoelectric performance in co-doped CaMnO3 using response surface methodology and neural network

In this work, two different combinations of materials are prepared, and the effects on dual doping in Ca 1− x − y Gd x Sr x MnO 3 and Ca 1− x − y Ce x Sr x MnO 3 ( x = 0, 0.025, 0.05, y = 0, 0.025, 0.05) materials are evaluated, and its parameters are optimized and predicted by the Box-Benhken design in the RSM method. The activation energy was measured with respect to different thermoelectrical material concentrations. RSM design is validated using hybrid DBN-RSO. The results show that increasing temperature, increases the amount of doping, decreases the thermal conductivity (k) and increases the electrical conductivity (σ) and power factor (PF). A bigger number of merits was also reached by increasing the amount of doping and the temperature. At 1000 K, the Ca 0.95 Gd 0.05 Sr 0.05 MnO 3 material has a low thermal conductivity and the highest figure of merit (ZT) value of 0.24, which is more than Ca 0.95 Ce 0.05 Sr 0.05 MnO 3 . The predicted values from the DBN-RSO method provide results that are closer to the experimental observations. The highest score (ZT) that the DBN-RSO prediction received was 0.26. Besides, the regression value of 99% is obtained from the experimented and predicted values. It shows the confidence and fitness of values. Also, the DBN-RSO achieves closer results to the experimental design with the lowest error value.