Application of RBF Neural Network to Temperature Compensation of Gas Sensor

Gas sensor is vulnerable to the impact of environmental temperature, thereby limiting its accuracy. In order to overcome this shortcoming, the paper proposes a new temperature compensation method based on RBF neural network, which is realized with Visual C++ 6.0 program software. The result of experiment indicates that the biggest error of the sensor outputs may be up to 20.0 percent before temperature compensation. After we adopted the temperature compensation method based on BP neural network, the biggest error reduced to 1.44 percent, even down to 0.12 percent through the method based on RBF neural network. Therefore this way has better effect on the temperature compensation so that the gas sensor may have higher accuracy and temperature stability after compensation.