Font Size: a A A

Research Of Non-Linear Prediction And Identification Omen Of Coal And Gas Outburst By Support Vector Machines

Posted on:2010-04-05Degree:DoctorType:Dissertation
Country:ChinaCandidate:Z Y ChenFull Text:PDF
GTID:1101360278461437Subject:Safety Technology and Engineering
Abstract/Summary:PDF Full Text Request
The coal mine monitoring gas system has been established in lots of coal mines, but it is lack the early warning function to forecast coal and gas outburst. In this paper it is to study the development, dynamics, and other non-linear characteristics of the coal and gas outburst occurred by the coal mine monitoring gas system to monitor the gas concentration data, as well as to predict coal and gas outburst based on support vector machines. The main innovation results are as follows:1) Not only the cusp catastrophe models of coal and gas outburst are established respectively from the point of view of energy and momentum conservation, but also the chop mechanisms and the conditions of coal and gas outburst are concluded. So the new theory of disaster prediction and prevention from coal and gas outburst is put forward.2) The effect factors of gas emission from the working face have been analysed. The theory and monitoring data are indicated as follows: The gas emission trends, which are abnormity when the coal and gas outbust happens, while are uniformity and smallness when the coal and gas outbust does not take place from the working face, are consistency with the danger of coal and gas outburst in the working face. The dynamic characteristics of the gas emission from the working face are the better predictors of non-contact, which not only do not engross busywork time, but also do not broach forecast, then which have the advantage of continued and dynamic forecast,as well as make up the disfigurement which the coal mine monitoring gas system does not forecast coal and gas outburst.3) The contrast features of chaotic motion between coal and gas outburst happening and coal and gas outburst not taking place, are studied by using the Chaos Kinetic theory. The Hurst exponent of gas emission from the working face is calculated. Its results are less than 0.5, which are indicated that it has strong chop characteristic of coal and gas outburst. The retardation time of phase space is calculated by using the complex correlative function method. The retardation time of phase space of gas emission from the working face in the coal and gas outburst happened is biger than that of phase space of gas emission from the working face in the coal and gas no outburst. The traditional method defects of calculating the correlation dimension are pointed out, while the new method of calculating the correlation dimension, which is applicable for high-dimensional chaotic system, is put forward. The correlation dimension of gas emission from the working face in the coal and gas outburst happened is biger than that of phase space of gas emission from the working face in the coal and gas no outburst, while the result of the least inbuilt saturation dimension is opposition. The Wolf arithmetic of Lyapunov exponent is ameliorated in view of the less evolutive vector length and angle, and it is found the law that the Lyapunov exponent of gas emission from the working face in the coal and gas outburst happened is greater than that of phase space of gas emission from the working face in no the coal and gas outburst.4) It is pointed that the shortage of part forecast is rested with the correlative degree between the nearest point and the center point, The calculation of the nearest point of new method which can effectively prevent the neighboring pseudo-point, is put forward. The prediction model of Lyapunov , which the predictive value is positive and negative, is impacted on the overall accuracy of forecasting gas concentration. So, not only the Lyapunov of forecasting model to determine the predictive value of ideas, but also of determining the appropriate algorithm are put forward. Wavelet based on the integration of chaos and chaotic time series forecasting methods is also put forward. The very high precisions of gas concentration are forecasted dividably by improving weight of first-order local-region prediction method, by improving prediction model of Lyapunov, and by improving prediction models based on wavelet and chaotic integrated prediction of chaos.5) It is pointed out that the information to monitor the heading face of the danger of coal and gas outburst is identified by support vector machines in the coal mine monitoring gas system. The kernel pattern recognition function and the structure principles of algorithm applying the heading face of the danger of coal and gas outburst are studied. The time-domain characteristics of the vector, the frequency characteristics of the vector, the wavelet domain feature vector, and fractal and chaos characteristics of the vector are educed in the gas concentration of the heading face of the coal and gas outburst and no-outburst by the coal mine monitoring gas system. And support vector machines are in the application of forecasting conventionally coal and gas outburst, and of processing data from disaster KBD7 coal-power electromagnetic radiation monitor. The results show that the support vector machines identifier model, which solves the real problems of the small sample, non-linear, high-dimension, the local minimum value, and so on, has a good classification results. In this paper, by using Visual C# 2005 and Matlab7.0 language, the support vectormachines recognition system, which distinguishes continually non-contactly coal and gas outburst from no outburst by analysing precursor information of coal mine, with the original coal mine gas monitoring system communicating the internet for data, is empoldered. The forecasting results are believable, are consistent with the mine practice, so the support vector machine recognition system is provided with of great practical significance.
Keywords/Search Tags:coal and gas outburst, support vector machines, chaos, lyapunov exponent, pattern recognition
PDF Full Text Request
Related items