Research On Dynamic Measurement Method Of Temperature Modulation For Semiconductor Gas Sensor

Although the rapid development of social economy and the enhancement of industrial production capacity has facilitated our lives,at the same time,the waste gas produced by industrial production,automobile exhaust,food processing,etc.also brought serious environmental pollution problems.Among them,volatile organic compounds(Volatile Organic Compounds,VOCs)are not only the main causes of acid rain and smog but also bring great harm to human physical health.Therefore the detection and control of VOCs gas has become an urgent problem.Metal oxide gas sensors have been widely used in the detection of volatile organic compounds due to their advantages of low cost,simple fabrication,quick response and high sensitivity.However,the problem of poor selectivity in actual complex gas environments has become a bottleneck restricting the development of semiconductor gas sensors.For this reason,this paper presents a dynamic test method,and achieves ideal results.This thesis is mainly composed of three groups of experiments,the contents of which are as follows:(1)The dynamic test method of temperature modulation is a common method to solve the poor selectivity of metal oxide sensors,but there has not been a clear method to control the waveform of the dynamic response signal to achieve the expected expectations.This experiment first describes the correspondence between static performance index and dynamic response signals from the perspective of static testing,and proposes a selection method of semiconductor sensors suitable for dynamic testing.Then taking rectangular wave as an example,the response time and power consumption in practical application can be shortened by adjusting its period,duty ratio and operating temperature range without reducing the quality of dynamic response signal.(2)In order to solve the problem that the characteristic peak is not obvious in the dynamic test,the SnO2 gas sensor was developed and the temperature modulation test was carried out.This experiment first proposed the sawtooth rectangular wave temperature modulation method,which significantly improved the selectivity of the sensor and the characteristic peak of the current curve.The use of support vector machines for pattern recognition proves the superiority of support vector machines in the case of small samples.Ethanol,methanol,propanol,2-butanone and butyl acetate were successfully identified by using the pattern recognition method of support vector machine under the sawtooth rectangular wave temperature modulation.Compared with BP neural network and decision tree,SVM has the advantage of small sample size.The results show that the combination of sawtooth rectangular wave temperature modulation and support vector machine pattern recognition can effectively improve the selectivity of the gas sensors.(3)In order to solve the problem of low accuracy of dynamic testing,the original experimental scheme are improved to perform dynamic testing.The method of continuously obtaining experimental data is adopted,and the concentration gradient between the samples is reduced from 100 ppm to about 1.5 ppm,which is close to the continuous concentration data.In this experiment,principal component analysis(Principal Component Analysis,PCA),K-nearest neighbor algorithm,polynomial regression and normal distribution are applied to the process of pattern recognition.When the confidence interval is located at 95%,the prediction accuracy reaches about 5ppm.The results show that the continuous data acquisition method combined with PCA and other algorithms can effectively improve the accuracy of dynamic testing.