Research On Dust Concentration Detection Method Based On Multi-sensor Information Fusion

As a companion of industrial production,dust has great harm to industry,human health and living environment,reducing the efficiency of industrial production and greatly increasing the risk for pneumoconiosis.Therefore,accurate and real-time detection of dust concentration is particularly important.In this paper,a method of dust concentration detection based on multi-sensor information fusion is studied by combining theory and experiment.According to the insufficiency that the optical window of photoelectric dust concentration sensor is prone to be polluted and needs constant calibration and large amount of maintenance,in this paper,electrostatic sensor is applied to realize detection of dust concentration Since a single sensor has its own limitations and cannot get the dust concentration in a very timely and accurate way,and if the sensor fails,the detection will be greatly affected.In order to solve this problem,multi-sensor information fusion is introduced.On the basis of previous studies,BP(Back-propagation)neural network fusion algorithm is applied.In view of the insufficiency of the traditional BP neural network,in this paper,it is improved by using the BP neural network with momentum term adaptive learning rate.In comparison with MATLAB simulation,it is concluded that the improved BP neural network has faster convergence speed.Moreover,since the initial weights of the improved BP neural network are randomized assignment,and are mostly obtained by experts' subjective experience,therefore,in this paper,based on the trust function and likelihood function of the DS evidence theory,the exact initial weights are obtained,and DS-BP neural network fusion model is established,which can accelerate the convergence speed,reduce the error and improve the fusion precision.DS-BP neural network is applied to dust detection.The experiment is carried out by using the experimental device of simulating the dust environment in mines.The electrostatic sensor is placed in the different position of the device,and the NI data collection equipment and LabVIEW software are used to collect the signal.After the wavelet threshold de-noising and feature extraction,the signal is processed by the DS-BP neural network fusion algorithm,and the precise dust concentration is obtained.Through the experiment,it is proved that in comparison with the improved BP neural network,the application of DS-BP neural network fusion algorithm in the dust detection can decrease the times of convergence,reduce the error and improve the accuracy of dust concentration.