Implementation of Artificial Neural Network for Detection of LPG through Pd-Doped SnO2 Based Thick Film Gas Sensor

Pd-doped SnO2 thick film sensor was produced on alumina substrate and their LPG sensing behavior was investigated. In the present paper studies have been made to analysis the sensitivity, response of SnO2 based with different Pd-doped thick gas sensor for LPG detection using different nonlinear ANN technique. To achieve this thick film Pd doped SnO2 sensor was fabricated on a 1" x 1" alumina substrate. LPG gas consists of sensitive layer SnO2 based with different Pd- doped, a pair of electrodes underneath the gas sensing layer serving as a contact pad for sensor and heater element is fabricated on the backside of the substrate. On both pure and Pd-doped samples measurements of conductance and capacitance have been made by fixed temperature at fixed gas concentration. The sensitivity of sensor has been studied at different Pd doped at a constant temperature of 350°C upon exposure to LPG. Sensing properties such as, sensitivity and selectivity measurement are carried out in the range 0- 5000 ppm at 350°C. The maximum response of 72.08% at 350°C was obtained under the explore of 0.5% of LPG. Practical Sensitivity of LPG is used to measure the sensitivity of different Pd- doped SnO2 based thick film gas sensors by using ANN algorithm. Two training algorithm of feed-forward algorithm namely Levenberg-Marquardt feed forward propagation and Random Weight /Bias Rule with adaptive learning rate were used. The performance of ANN models with different nonlinear algorithm is evaluated on practical sensitivity of sensor with different network transfer functions. Experimentally, we found that ANN model with training algorithm is more suitable for sensitivity of the sensor. Results present in the paper shows ANN as an effective tool in the area of SnO2 based thick film sensor design.