Research On Coal Ash Fusion Temperature Prediction System Based On Neural Network

In the article, the analysis method of gray correlation degree was used to analyze the correlation degree between coal ash chemical composition parameters or combination parameters (consisted of ash chemical composition) and flow temperature. Parameters with larger correlation degree were selected as input variables to predict coal ash flow temperature, so did parameters which were widely used to predict coal ash flow temperature in the literature. Network models were established using MATLAB generalized neural network, then according to the minimum prediction error, the most optimal models were selected to predict ash flow temperature. Then under Visual Studio2010development environment, combining MATLAB mixed programming, and using SQL Server2008database, a set of coal ash fusion temperature prediction system was developed based on neural network models.The research conclusions were as follows:(1)According to the size of acidic components (SiO2+Al2O3+TiO2) content,354kinds coal ash samples were divided into five categories, such as high silicon aluminium coal ash(≥85%), middle high silicon aluminium coal ash(80%-84%), middle silicon aluminium coal ash(65%-79%), low and medium silicon aluminium coal ash(50%-64%) and low silicon aluminium coal ash(<49%). Using grey correlation method, it was obtained that grey relational degree of combination parameters are almost more than its individual components, including the acidic components, the sum of the three components (calcium, iron and magnesium), silicon aluminum ratios, the sum of potassium and sodium. So the above parameters have larger influence on coal ash flow temperature.(2) Aimed at high silicon aluminium ash, which divided into two categories again, silicon aluminum ratios greater than or equal to2and (Fe2O3+CaO+MgO) content greater than or equal to8%, another silicon aluminum ratios less than2and (Fe2O3+CaO+MgO) content less than8%. After the analysis, it is found that the latter’s coal ash flow temperature were greater than1500℃, so only for the former and the rest four categories, network models were established respectively according to two prediction methods (a) and (b) using the MATLAB generalized regression network. Then network models above established were trained and tested respectively. For the better model prediction accuracy, the five models established by the method (a) were selected, and the testing maximum relative error was respectively1.34%,7.55%,10.50%,4.75%,8.16%,1.34%, the testing minimum relative error was respectively0.52%,0.00%,0.00%,0.00%,0.00%, the average relative error was0.93%,0.07%,0.67%,0.08%,0.70%. Finally, m files corresponding to five network models were constructed by MATLAB language and compiled into.net components using the MATLAB compiler.(3) Using C#and MATLAB mixed programming to transfer the formed.net component, flux flow temperature prediction system of Coal blending and adding limestone flux was established. Then through debugging the system works properly, and can be used to predict coal blending flow temperature quickly.(4) Coal quality database was established based on SQL Server2008database software, and using C#programming language the coal quality management was realized, namely query, add, edit or remove coal quality information as required.