Research On Data Fusion Of Toxic And Harmful Gas Detection Based On Sensor Array

The detection of the concentrations of various components in toxic and harmful mixed gases has been widely used in coal mine gas,petrochemical,environmental protection,urban underground corridors and other fields.Due to the cross-sensitivity problem of common gas sensors,the output data of the sensor often includes the crossinterference of other non-target gases;the elimination of the cross-interference effect of a single gas sensor can only rely on the selectivity of the gas sensor itself,which is difficult to achieve Qualitative identification and quantitative measurement of multicomponent gases.Therefore,it is of great practical value to use data fusion to process the output signals of a sensor array composed of multiple sensors to achieve accurate detection of the concentration of each component in a toxic and harmful mixed gas.This article first compares common gas detection methods,and selects the gas sensor method for the detection of toxic and harmful gases.By comparing and analyzing common gas sensors,three electrochemical gas sensors are finally selected to measure the concentration of O2,H2 S,and CO,and a catalytic combustion.The gas sensor measures the concentration of CH4.Then,the working principle and characteristics of the selected electrochemical gas sensor and catalytic combustion gas sensor are introduced in detail;it provides support for the hardware design and the solution of cross-interference problems in the following chapters.The hardware and software design of the detection system was completed according to the requirements.Hardware part: This design is divided into data acquisition board(sensor array)and main control board.Software part: complete the lower computer software programming(data acquisition board and main control board);use Lab VIEW and Matlab mixed programming to complete the design of the upper computer.The cross-sensitivity of gas sensors affects the accuracy of measured values,and the input and output show a complex non-linear relationship.In this paper,data fusion is used as an important way to solve the problem.It can be used to approximate any non-linear function relationship in a global sense.Support vector machine as a data processing method to solve the problem of cross-interference between gas sensors;this method not only takes advantage of the excellent performance of support vector machine when processing small samples,but also avoids the difficulty in determining the structure of the neural network and easy to fall into the local area Minima issues.In practical engineering applications of support vector machines,a high-performance kernel method is the key;after the kernel function is determined,there is the problem of the best matching among the parameters of the support vector machine.The parameters are set reasonably.The determination of the parameters of the support vector machine is a non-linear global optimization problem.To solve this problem,this design introduces particle swarm optimization(PSO)to optimize the parameters in the support vector machine.The optimal problem introduces the Adaptive Particle Swarm Optimization(SCPSO)to optimize the parameters of the support vector machine.This method overcomes the problem of slow optimization of the genetic algorithm and also makes up for the basic particle swarm algorithm(PSO)due to Following the local optimization problem caused by the convergence of the optimal particle process,it provides an effective way to solve the cross interference problem of gas sensors.Based on the repeatability,high and low temperature characteristics,and crosssensitivity characteristics of the sensor,the design completed repeatability experiments,high and low temperature experiments,and cross interference experiments.At the end of the article,by introducing the average absolute percentage error as an evaluation criterion,the sensor array acquisition concentration in a toxic and harmful mixed gas environment,the prediction result based on support vector machine,the prediction result based on PSO optimization support vector machine,and the support vector machine based on SCPSO optimization are calculated The average absolute percentage error of the prediction result relative to the theoretical value;the comparison results show that the support vector machine model based on SCPSO optimization shows better performance and has practical engineering application value.