| Coal quality detection is a necessary technology to guarantee the economic and safe operation of coal-fired power plants.Current detection methods have some disadvantages,including complex sample preparation processes and low detection efficiency,which cannot meet the growing detecting needs of coal-fired power plants.Laser-induced breakdown spectroscopy(LIBS)is a non-digesting elemental analysis method with the advantages of simple sample pretreatment,no special requirements for the detection environment,fast analysis speed,simultaneous multi-element analysis,remote detection etc.Thus,LIBS is very suitable for online detection of industrial application.In this thesis,coal quality and ash fusibility detection method in industry based on LIBS were investigated.The specific research contents are as follows:A LIBS platform for coal quality detection was set up in laboratory.With the investigations of sample preparation method,spectral line selection,experiment parameter optimization,and application of multi-layer perceptron regression model to predict four common analysis targets,the coal quality detection was achieved with good accuracy.The result shows great analysis capacity of LIBS in coal quality detection and lays foundation for subsequent industrial applications.Based on the laboratory optimization,an online coal quality detection system was developed.The system can perform real-time and online detection of industrial coal quality.Five common analysis targets(including ash,volatile matter,fixed carbon,hydrogen content and bomb calorific value)were detected.The detection results can be obtained in 30 minutes.The result of coal quality detection under industry condition was analyzed from three aspects of static precision,static accuracy,and dynamic precision.Furthermore,a long-term detection experiments of 153 coal samples was performed.The results demonstrated the good stability and reliability of the coal quality online detection.A method for predicting the ash fusibility temperature directly from the coal spectrum was also proposed in this thesis,which will simplify the flow temperature prediction process Partial least squares regression model,multi-layer perceptron regression model and support vector machine regression model were used in the prediction of flow temperature.The repeatability and accuracy of each model were compared.The result shows that the support vector machine regression model can achieve better repeatability and accuracy under industry condition. |
