Evaluation of sensing response of H2S sensing materials toward H2S molecules using artificial neural network and response surface methodology
Year: 2026
Type: article
Abstract: In this work, temperatures of 25 to 275 °C, times of 6.5 to 4091 s, concentrations of 0.0001 to 400 ppm, and surface areas of 17.11 to 223.6 m 2 /g were simulated through artificial neural network (ANN) and response surface methodology (RSM) for predicting the H 2 S sensor response. The number of collected data points was 548. The optimum values of each parameter were calculated using RSM. The optimized variables were 97.11 ℃ of temperature, 2785 s of response time, 259.77 ppm of H 2 S concentration, and a surface area of 197.78 m 2 /g. Regarding RSM, the dependency of sensor response on the parameters is quadratic. The multi-layer perceptron (MLP) and radial basis function (RBF) models were used in the ANN. Levenberg-Marquardt backpropagation algorithm (trainlm) was used. The numbers of neurons of 49 and 29 were considered for the first and second hidden layers for the MLP network, respectively. The number of neurons was calculated as 198 for the RBF model. The R 2 values of MLP, RBF, and RSM models were attained at 0.9897, 0.9894, and 0.9821, respectively. Also, the % AARD values of the MLP and RBF models are 17.33 and 66.20, respectively. It is concluded that the MLP model could predict the sensor response better than the RBF model. The transfer function for the first and second hidden layers was tansig in the MLP model. The best MSE validation performance of the MLP and RBF networks was 0.001061 and 0.001486 at 69 and 198 epochs, respectively. The MLP 3D diagram also proved that H₂S sensing materials such as Fe 2 O 3 -Ba 0.99 Ce 0.01 TiO 71 , SnCl 2 , and Fe-600 are good candidates for sensing the H 2 S molecules.
Source: Scientific Reports
Language: English
Subfield: Electrical and Electronic Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: gold
Estimated APC paid: $2,690
OpenAlex ID: W7236839178
