Research on temperature compensation method of non-dispersive infrared CO2 sensor
Aiming at the problem of non-dispersive infrared CO2 sensor are susceptible to temperature,a new method of double semi ellipsoidal and double channel gas absorption chamber and generalized gaussian radial basis function neu-ral network for temperature compensation is proposed in this paper.Firstly,the right focus of the semi ellipsoidal long axis coincides with the left focus of the right semi ellipsoidal long axis in the double semi ellipsoidal and double chan-nel gas absorption chamber structure,the infrared light source is located at the overlap of the left and right semi ellip-soidal long axes,and the detector is located at another focus of the semi ellipsoidal long axis,which enables the light to be received by the gas sufficiently,and reduces the loss of light propagation.Secondly,the differential calculation of the output voltage of a single wavelength double channel pyroelectric detector is used to obtain the concentration of CO2 gas.Finally,the influence of shape parameters and auxiliary shape parameters of the generalized gaussian radial basis function neural network on improving approximation performance is analyzed.The experiment shows that CO2 gas absorption rate reaches its maximum value at a long axis length of 4 cm and a short axis length of 3 cm in the double semi ellipsoidal and double channel gas absorption chamber structure.At room temperature 250 C,the measurement er-ror of different concentrations of CO2 is not much changed,with an absolute error fluctuation range of 0.004％～0.006％and the relative error fluctuation range is between 0.067％～0.100％.When the CO2 concentration is 2％,the measurement error varies greatly at different temperatures,with an absolute error fluctuation range of 0.001％～0.024％and the relative error fluctuation range is between 0.017％～0.400％.The average response time of a CO2 gas sensor with a measurement concentration of 2％for six times is 27.46 seconds,indicating fast response time.
NETL
NSTL
万方数据
