| In recent years,with increasing awareness of people's safety,a frequency of gases explosion accidents has being begun to decline,but a large of those kinds of accidents still occur.So it is very important to improve a better performance of an existing gas detector.As the main component of gases is methane,a real-time and accurate detection of its concentration is a primary condition to avoid accidents.Compared to a traditional catalytic methane sensor,an infrared methane sensor has the advantages of hard poisoning,high precision,long service life and so on,so a use of an infrared spectrum technology is a future development trend of methane sensors.The infrared methane sensor is integrated by many electronic components,but compared with the first-class products in the world,the currently domestic products are handicapped by their designing and manufacturing levels of core components,and there is stilla long way to go,so it is very critical and necessary to use an embed technology with a software design about error compensations.In this paper,starting from the second chapter,a detection principle of an infrared methane sensor is studied more deeply and carefully,and it is found that a temperature has a non negligible effect on a wavelength of light sources and sensor components,they are mainly shown in the following aspects:(1)With increasing of temperatures,a center wavelength of the infrared light source will be reduced gradually.(2)A absorption coefficient in Lambert-Bill law will change accordingly.(3)Changes of temperatures will affect a performance of semiconductors,resistors and capacitors in a sensor.An influence on semiconductors is the largest,because those semiconductors are composed of P-N junction units,and a reverse leakage current will increase or decrease once per 10?changing of an ambient temperature.A widely appllication of semiconductors in an integrated operational amplifier circuit,a voltage regulator circuit and a logic chip will,therefore,produce complex nonlinear drift errors,an effect of temperatures on resistance changes with the resistance materials,a resistivity of metal resistances are positively correlated with changing of temperatures.But resistivities of insulators and semiconductors are negatively related to the changes of temperatures.Furthermore,the temperature changes have the same effect on a capacitance and a loss tangent.Based on the above mentioned problems,considering all these factors,especially focusing on the third point,they will have a great impact on the accuracy,precision and reliability of the sensor.And a mechanism of these effects is still not very clear,so it is impossible to use a unified equilibrium formula to sum up.Therefore,a temperature-impact verification experiment is first designed,the data obtained are used to study the electrical changes of our existing sensors when measuring the methane concentrations under different temperatures,and the nonlinear variations of experimental data provide a reliable basis for a further verification about a complexity of temperature-impact mechanisms.Furthermore,according to the complex and changeable characteristics of the experimental data,a necessity and feasibility of using Gaussian regression algorithm for temperature compensations are demonstrated,and the Gaussian regression process as a practical machine algorithm based on a probability theory is realized.According to the existing experimental sample,a mathematical model which induces a relationship between input values and output values is proposed and constructed.On the basis of the model,a statistical theory related to Gaussian regression processing is studied and the characteristics of each basic covariance function in the Gaussian process are analyzed.And performing an addition and multiplication of a kernel operation to construct a new type of composite covariance function which is more suitable for theexperimental sample,combined with original measurement data of the infrared methane sensor and its distribution features,the temperature compensation model based on Gaussian regression process is established.According to the original model constructed by the different covariance function,an optimization basis about the maximization of marginal distribution is proposed,and the maximum likelihood estimation method is selected to optimize the parameters of each model.A control effect of all the parameters on the data fitting and fitting errors is analyzed using MATLAB software.The average absolute error of predicted methane concentration after temperature compensation is0.0572,and the mean square error is 0.0057,which means that this method can achieve an effective temperature compensation for nonlinear errors in the infrared methane sensor.In addition to comparing the Gaussian regression model constructed by the different covariance function,the temperature compensations based on the BP neural network model and the least square Newton interpolation model are also compared through a simulation experiment on the LabVIEW platform,the validity and rationality of Gaussian regression model for temperature compensations are verified.In addition,based on the Gaussian regression model for temperature compensations,a system structure and software of the infrared methane sensor is studied.The system structure in it contains an opticalmeasurement part and a circuit design part,the optical measurements mainly include the researches about an infrared light source,a detector,a gas chamber and so on.Considering the characteristics of the infrared absorption with methane,IRL715 incandescent lamp and PYS3228 infrared pyroelectric detector are selected.In order to improve the reliability of the sensor,the dual channel optical system is applied;a power modulation circuit and an amplifier filter circuit are designed.Its software part adopts the top-down modular design method,including the designs and implementations of original signal acquisitions,the methane concentration predictions,serial communications and so on,in which the methane concentration prediction module contains the Gaussian regression model for temperature compensations,the serial communication module is used for the communications between the sensor and the computer.When those programd are called by the MSP430 microcontroller in the sensor,the corresponding functiond can be realized to predict the methane concentrationd in an actual environment accurately. |
PDF Full Text Request