SVM-Based Soft-Monitoring Network System for Industrial Dust Emissions

Abstract

The industrial dust emission is characterized by scattered emission points with accurate emission parameters difficult to be obtained. The soft-sensing system for industrial dust emissions based on the B/S and C/S mixed mode was studied in terms of the topological network structure, logical structure and a series of key technologies. The network system was constructed with a data flow model, which is based on signal acquisition layer, data acquisition layer, and the area management and central management layers. The user/client cluster is constituted with area client, center client and mobile client. Interanet/Internet was employed to build up a visualization platform for environmental monitoring. Algorithm for the identification of industrial dust concentrations and emissions is proposed on the basis of image processing and Support Vector Machine (SVM). With the help of the characteristic parameters of dust image, vectors to be detected was constructed and normalized to get the real-time concentration and emissions from the SVM. The actual running situation showed that the System not only provided real-time visual monitoring and the accurate concentration of emissions data, but also simple and reliable for end-users and excellent in system scalability.