Research on Prediction and Prevention Technology of Heading Face Outburst Based on BP Neural Network
|Course||Detection Technology and Automation|
|Keywords||coal and gas outburst BP neural network forecasting systemdevelopment|
Coal is the main source of energy in China. With the increase of coaldemand and excavating depth of mining, magnitude of coal and gas outburst isgreatly enhanced. Coal and gas outburst has restricted the development of coalindustry and the advance of economic efficiency seriously. Therefore, the researchfor prediction method of coal and gas outburst would contribute to improve thesafety of mine greatly. Considering the status of plentiful outburst factors and morefrequent outburst accidents, modern control theory and outburst mechanism would becombined to search new ideas and methods for predicting outburst.Firstly, the status of predicting outburst was summarized at home and abroad.The growing process and general rule of outburst were also expatiated. Main factorswere analyzed in outburst, such as stress in coal seam, gas and structure of coal. Atthe same time, outburst mechanism was investigated.Next, Standard BP network was improved by Levenberg–Marquardt algorithmon the basis of neural network theory. Using MATLAB, the model to forecastoutburst was built. Structural parameters, fittest number of nerve cell in hidden layerand training function were confirmed by programming. Through theoreticaldeduction and simulation results, the convergence rate and generalizationperformance of L–M algorithm was indicated better than other improved arithmeticsin outburst prediction.Wuyang Coal Mine of Lu’an Company was regarded as the actual object ofinvestigation, and its main outburst factors were analyzed roundly. Parameters werechose as the input of neural network model, such as gas pressure, gas absorptionindex of drill cuttings, drilling cuttings weight, Protodyakonov coefficient and depth.Coal samples from different place were mensurated in locale and laboratory, andtwenty sets of representative data were elected as the training samples for predictionmodel. Another six groups of samples were used to prove generalizationperformance and reliability of the improved model. Actual outburst status of six testsamples was in accordance with the result obtained by the model. Thus, it`s verified that outburst danger of mine could be predicted exactly by built model.Finally, graphical user interfaces and system programming were accomplishedby MATLAB. Coal and gas outburst prediction system was developed, which couldcombine several indexes forecast with neural network intelligent computing. At thesame time, its practical application was carried, and the visualization and theinformation technology have been realized to predict outburst.