Study On Pre-warning Of Spontaneous Fire In Coal Mine Based On Computational Intelligence

Pre-warning of spontaneous fire(including pre-assessment of coal seam spontaneous fire danger) has received increasing attention from coal mine safety researchers and engineers. Taking advantage of diverse disciplines, this dissertation introduces several new theories or methods in computational intelligence, such as support vector machine(SVM), particle swarm optimization(PSO) and rough set(RS), to meet the practical needs of developing spontaneous fire pre-warning technologies. This research mainly consists of several tasks as presented below:1) Improvement of PSOTo avoid premature convergence to a local optimum, which might occur in standard PSO, an improved PSO(IPSO) algorithm is proposed by integrating a chaotic map and a new strategy of inertia weight adjustment based on particle spacing. IPSO is tested against three standard functions, and other four kinds of PSO algorithm are used to compare. The test results show that, IPSO has a stronger global optimization capability and gives a higher accuracy than the other four algorithms.2) IPSO based optimization of SVM parametersIn order to improve SVM prediction accuracy which is highly dependent on SVM parameters, IPSO is applied to optimize SVM parameters, and two algorithms, namely, IPSO-v-SVR and IPSO-SVC are proposed. These two algorithms are tested on UCI data sets, and several other algorithms for regression or classification are used to compare. The test results show that, IPSO-v-SVR has the best accuracy among the regression algorithms, so does IPSO-SVC among the classification algorithms.3) Coal seam gas content prediction based on IPSO-v-SVR and coal seam spontaneous fire danger pre-assessment based on IPSO-SVCIPSO-v-SVR is test on a data set of coal seam gas content. The results show that, IPSO-v-SVR can predict coal seam gas content at a relatively high accuracy for small sample case.IPSO-SVC is test on a data set of coal seam spontaneous fire danger. The results show that, IPSO-SVC can classify coal seam spontaneous fire danger at a certainly high accuracy. Moreover, IPSO-SVC is used to assess the spontaneous fire danger along two working seams that are not included in the above data set. The assessment results fit well with the actual situations.4) Fusion of RS and IPSO-SVC, and its application in pre-warning of spontaneous fire in caving zoneAn index system for pre-warning spontaneous fire in caving zone is established, which makes use of the analyses of gas sample taken from gas drainage borehole. By fusion of RS and IPSO-SVC, an RS-IPSO-SVC algorithm is put forward. RS-IPSO-SVC is tested on a data set of spontaneous fire in caving zone. The results show that RS-IPSO-SVC’s accuracy is higher than the algorithms used to compare, and it is certainly applicable in pre-warning of spontaneous fire in caving zone.5) Development of coal mine spontaneous fire information management & pre-warning systemTo meet the practical requirement of coal mine enterprises on spontaneous fire information management, and the need for data refreshment during application of intelligence algorithm, a coal mine spontaneous fire information management & pre-warning system of B/S framework is developed based on.NET technology. This system has been applied in Yangdong coal mine, Fengfeng Group Company, and has achieved a good effectiveness in indentifying the abnormal phenomena related to spontaneous fire hazard, and in collecting new data to help enrich data set.Through above researches, several main conclusions are drawn as follows:1) IPSO has a stronger global optimization capability.2) IPSO can significantly help find the optimal SVM parameters.3) IPSO-v-SVR, IPSO-SVC and RS-IPSO-SVC can be separately applicable with relatively high accuracy in prediction of coal seam gas content, pre-assessment of coal seam spontaneous fire danger, and pre-warning of spontaneous fire in caving zone.