基于高斯回归过程的红外甲烷传感器温度补偿算法研究

发布时间：2018-04-03 10:52

本文选题：甲烷检测　切入点：温度补偿　出处：《太原理工大学》2017年硕士论文

【摘要】：近年来,随着人们安全意识的不断增强,瓦斯爆炸事故发生频率开始有所下降,但特大型事故仍时有发生。因此提高现有瓦斯检测仪的性能就显得至关重要。由于瓦斯的主要成分是甲烷,实时、精确地检测出甲烷气体浓度是避免事故发生的首要条件。相对于传统催化型甲烷传感器,红外甲烷传感器具有不易中毒、精度高、寿命长等诸多优点,利用红外光谱技术检测甲烷气体是未来甲烷传感器的发展趋势。由于红外甲烷传感器是由许多电子元器件集成后的产品,目前国内产品受制于核心元器件的设计与制作水平,与世界一流产品仍有不小的差距,所以设计误差补偿型软件嵌入且集成于其中就显得十分关键和必要。本文从第二章开始,对红外甲烷传感器的检测原理进行了深入研究,阐述并论证了温度对红外甲烷传感器中的光源波长、元器件有着不可忽略的影响,表现在:1)随着温度的升高,红外光源的中心波长会逐渐减少;2)朗伯比尔定律中的吸收系数也会发生相应的变化;3)温度的变化会影响传感器中的半导体、电阻、电容等元器件的性能,其中半导体受到的影响最大。因为半导体是由P-N结单元构成的,环境温度每变化10℃,其反向漏电流就会增大或减少一倍,在传感器中大量使用的半导体的集成运放电路、电压稳压电路、逻辑芯片都会由此而产生复杂的非线性漂移误差,温度对电阻的影响则由电阻材料的差异而有所不同,金属电阻的电阻率与温度呈正相关变化,绝缘体和半导体的电阻率变化则为负相关,温度对电容的容量和损耗角正切值同样有一定的影响。基于上述问题,这些因素综合在一起,尤其重点考虑第三点,就会对传感器的可靠性和检测精确性产生极大的影响,并且这些影响机理至今还不十分清楚,无法使用统一的平衡公式来归纳表征。因此,本课题首先设计了温度影响验证实验,所获数据用于研究本课题已研发的传感器在不同温度条件下测量甲烷浓度时的电性变化,而实验数据复杂的非线性变化也为进一步验证和论述温度影响机理的复杂性提供了可靠的依据。再者,根据实验数据复杂多变的特质,论证了利用高斯回归算法来进行温度补偿的必要性和可行性,实现了高斯回归过程这一种基于概率论的实用型机器算法,且根据已有的实验样本,提出和构建了一个归纳输入值与输出值关系的数学模型。以此模型为基础,本文研究了高斯回归过程中相关的统计学理论,分析了高斯过程中各个基本协方差函数的特点并进行了相加、相乘的核运算,从而构造了更适合于实验样本的新型复合式协方差函数,结合红外甲烷传感器中原始的测量数据及其分布特点,建立了基于高斯回归过程的温度补偿模型。针对不同协方差函数所构建的原始模型,进一步提出了边缘分布最大化的模型优化依据,并选择最大似然估计法来实现各个模型参数的优化,利用matlab软件将模型中各个参数在优化前和优化后对数据拟合度、拟合误差的控制效果进行了对比与分析;经过最优化后的温度补偿模型预测值平均绝对误差为0.0572,均方差为0.0057,表明此方法能够为不同温度下传感器的非线性误差实现有效的补偿。除了比较由不同协方差函数构建的高斯回归温度补偿模型外,还与BP神经网络模型、基于最小二乘的牛顿插值混合模型进行了比较,在LabVIEW平台上经过仿真模拟实验,验证了本文所建立的高斯回归温度补偿模型的有效性及合理性。此外,本文基于高斯回归温度补偿模型对红外甲烷传感器中的系统结构和软件部分进行了整合性设计研究,其中的系统结构设计包含传感器中的光学测量部分与电路设计部分,其中光学测量部分主要是对红外光源、探测器、气室等进行了研究;考虑到甲烷对红外光的吸收特性,选择IRL715白炽灯为红外光源和PYS3228型热释电探测器;为了提高传感器的可靠性,使用双通道光路系统并设计了电源调制电路和放大滤波电路。软件部分采用了自顶而下(Top-Down)的模块化设计方法,包括原始信号的采集与处理、甲烷浓度预测、串口通信、超限报警等程序的设计与实现,其中甲烷浓度预测模块包含了本文所建立的高斯回归温度补偿模型,串口通信模块则用于传感器与上位机之间的通信。当这些程序受到传感器中MSP430单片机的调用后,就能实现相应的功能,可对实际环境中的甲烷浓度进行精确预测。
[Abstract]:In recent years, with the growing awareness of safety, gas explosion accident frequency began to decline, but large accidents still occur. Therefore to improve the performance of the existing gas detector is very important. Because the main component is methane gas, real-time, accurately detect the concentration of methane gas is the primary condition to avoid accidents happen. Compared with the traditional catalytic methane sensor, infrared methane sensor is not easy to poisoning, high precision, long life and other advantages, the use of methane gas detection technology is the future development trend of infrared methane sensor. Because the infrared methane sensor is integrated by many electronic components of the product, the current domestic products subject to the core the components of the design and manufacture level, with first-class products in the world is still not a small gap, so the design error compensation software embedded in and integrated in the It is very crucial and necessary. This paper starts from the second chapter, the detection principle of infrared methane sensor is studied, expounds and demonstrates temperature on the wavelength of the light source infrared methane sensor, components have an impact can not be ignored in: 1) with the increase of temperature, the center wavelength of infrared light source will be reduced gradually; 2) absorption coefficient of Longbow Bill's law will change accordingly; 3) the change of temperature will affect the semiconductor sensor resistance performance of capacitors, wherein the semiconductor has the greatest influence. Because the node is composed of P-N semiconductor elements, the ambient temperature changes every 10 DEG C the reverse leakage current will increase or decrease as the integrated operational amplifier circuit of semiconductor used in sensors, voltage regulator circuit, logic chip will be generated from this complex nonlinear drift error The influence of temperature on the resistance, the difference of resistance material is different, the resistivity is positively correlated with temperature change of metal resistance and resistivity change of insulator and semiconductor is negatively related to temperature on the capacitance and loss tangent also has certain influence. Based on the above problems, these factors together, especially focus on third points, will have a great influence on sensor reliability and precision of detection, and the influence mechanism is still not very clear, unified balance formula cannot be used to summarize the characterization. Therefore, this paper designed the temperature influence experiments, the data obtained for the study of the changes of electric measurement of methane concentration in different when the temperature sensor has been developed in this paper, the nonlinear change of complicated experiment data for further verification and discussion of temperature influence mechanism To provide a reliable basis for complexity. Furthermore, according to the experimental data of complex traits, and demonstrates the necessity and feasibility of temperature compensation using Gauss regression algorithm, the realization of Gauss the regression process of a practical machine algorithm based on probability theory, and according to the existing experimental samples, proposed and constructed an inductive input the math model of the relation between value and output value. Based on this model, this paper studies the theory of statistics related to the Gauss regression process, analyzes the characteristics of each process in the basic Gauss covariance function and the sum multiplied by the nuclear operation, so as to construct a new type of composite covariance function is more suitable for the experimental samples, combined with the measurement the original data and distribution features of infrared methane sensor, a temperature compensation model based on the Gauss regression process for different covariance function The original model, further puts forward the optimization model based on the marginal distribution of the maximum and optimal selection to achieve the various model parameters of maximum likelihood estimation, all the parameters in the model before and after optimization of data fitting using MATLAB software, the control effect of the fitting error are analyzed and compared through temperature compensation; the model optimized the predictive value of the average absolute error is 0.0572, the standard deviation is 0.0057, it is shown that this method can for nonlinear error under different temperature sensor to achieve effective compensation. In addition compared by different covariance function to construct the Gauss regression model of temperature compensation, and the BP neural network model, Newton interpolation hybrid model based on least squares compared in the LabVIEW platform through the simulation experiments, verified the Gauss regression model of temperature compensation effectively And reasonable. In addition, the Gauss regression model of temperature compensation of the system structure and software of infrared methane sensor is studied in the integrated design based on the design of system structure which contains part of the optical measuring part and circuit design of sensor in optical measurement, which is a major part of the detector to the infrared light source, gas chamber etc. The study; considering the characteristics of infrared absorption of methane, IRL715 incandescent lamp as light source and PYS3228 infrared pyroelectric detector; in order to improve the reliability of the sensor, the use of dual channel optical system and the design of the power modulation circuit and the amplifying and filtering circuit. The software part adopts a top-down modular (Top-Down) the design method, including the original signal acquisition and processing, serial communication, methane concentration, design and implementation of alarm procedures, the methane concentration in the pre Measuring module includes the Gauss regression model of temperature compensation, the serial communication module is used for communication between sensor and computer. When these procedures by sensor MSP430 microcontroller after the call, you can achieve the corresponding functions, can accurately predict the methane concentration in the actual environment.

【学位授予单位】：太原理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP212

【参考文献】