Study On Setting Pressure Extraction And Working Resistance Prediction Of Hydraulic Support Based On WSN

Increasing coal mining height has put forward higher requirements to hydraulic support supporting ability. The setting pressure and the weighting value are the two most important indicators of the hydraulic support, which have realistic guiding significance to the mine safety production. Currently, the setting pressure and the weighting value are mostly determined by the subjective experiences of the hydraulic support operator, which makes a lot of hydraulic supports in the actual work setting pressure below its design value. In order to solve this problem, in this paper, the hydraulic support working resistance monitoring system based on Wireless Sensor Network(WSN, Wireless Sensor Networks) was proposed. By analyzing the hydraulic support working resistance, a method of setting pressure extraction and the working resistance prediction based on ELM for hydraulic Support were proposed. The research content mainly includes:Firstly, the software of hydraulic support working resistance monitoring system was designed. We established the integrated development environment with Microsoft visual studio 2008, Microsoft SQL Server 2005 and realize the data receiving, storage, display, query, curve drawing etc using c # and SQL language.Secondly, the method of setting pressure extraction was proposed. First, learning from non-uniform quantization thinking, the hydraulic support working resistance was non-uniform quantified by using of A-law compression algorithms. Because early working resistance changes drastically in setting pressure area, the data is bound to produce the most quantitative segments. Then the energy ratio method was used to extract the value in setting pressure area. Finally, to verify effectiveness of the algorithm, we did simulation by comparing the setting pressure extracted by our method and the experience estimates.Finally, according to the theory of extreme learning machine, the working resistance history data was used as training samples. By analyzing the impact on the performance of ELM produced by hidden layer neurons number, the number of hidden layer nodes was set as 20 to determine the ELM network model. The simulation experiments show that the forecast effect is good and the error rate is around 2% except the situation when the hydraulic support is moving and the error rate is more than 10% during this time.