Study on Early-Warning Assessment in ChineseCoalMine Safety Based on Genetic Neural Networks
Yong-wen Ju,Li-xia Qi, Qian-li Sun
School of Management and Economics, NorthChinaUniversity of Water Resources and Electric Power
Zhengzhou, P.R.China
(, , )
Abstract - The early-warning and pre-control process to recognize potential safety hazard of coal mine based on characteristics of production safety is put forwards in the paper. The warning evaluation index system of coal mine safety which influenced by human, machine and equipment,environment, management and information is established.Thenit conducted an empirical study by using an evaluation method of neural network based on genetic algorithm. Evidence shows that the methodhas better adaptability and high accuracy by combining with an example insupportingpersistent effectmechanism for the safety production of coal mine.
Keywords - Coal minesafety,geneticalgorithm, neuralnetwork, early-warning assessment
IINTRODUCTION
At present, an increase rate inChina coal mine demand annually is about 10%, which promotes the coal industry development andproduces kinds of coal accidents at the same time. According to the statistics, China is one of the countries that have the highest frequency of coal mine accidents.
The current supervision mechanism of coal mine safety in China is based on the emergency plan management, and the early warning mechanism had not beenreally set up. It is necessary to build long-term preventionmechanism combined withthe theory and evaluation method of coal mine safety and early warningto monitor,diagnose, control and correct production activities of coal mine. It could providetheoretical basis and technical support for preventing and reducing coal mine accidents.
IILITERATURE REVIEW
Domestic and foreign scholars have carried on theactive exploration and research in the assessment method of coal mine safety. The main assessment methods include:1) Fuzzy comprehensive evaluation: Ding Xia-jun (2004) [1],Sun Jia(2005)[2],Gao Wen-hua(2008)[3],Sun Jian-hua(2009)[4] built the index system and estimate the safety of coal mine by fuzzy comprehensive evaluation, they obtainthe conclusion consistent withactual situation______
National social science fund projects（12CGL101）;
Graduate students’ education innovation fund project plan of North China University of Water Resources and Electric Power（YK2011-03）;
Henanprovincesocial science planning project (2009BSH005)
by the quantitative evaluation. 2)Grey relation method:Xu Yi-Yong(2003)[5],Cao Shu-gang(2007)[6],Fu Yong-shuai（2009）[7]established the evaluation index systembased on the reality of coal mine production.Gray correlation analysis is used to evaluate the coal mine safety, and the different levels of security evaluation results are inducedaccordingly.3)Unascertained mathematicsevaluation method:Yan Le-lin(2004)[8] built the index system,confidence identification criteria and rating criteria based on unascertained mathematics theory, the safety indictor of coal mine is analyzed through building the unascertained measure model. 4) Neural network evaluation method:Huang Hui-yu(2007)[9],Zhou Zhong-ke(2011)[10],Gao Xiao-xu(2011) [11],Ding Bao-cheng(2011) [12] built the index system and estimate the safety of coal mine by neural network .Then the practicality and effectiveness of the model are verified.
From what we have analyzed above, the assessment methods that are used commonly include the AHP,fuzzy comprehensive evaluation, grey relation method, unascertained mathematicsevaluation method andneural network evaluation.However,these methods are short ofthe ability ofself-learning,it is difficult to get rid of thesubjective uncertainty and understandingof the ambiguityin the decision-making process. The neural network evaluation could avoid thedefect,but there are some inadequaciessuch as slow convergence velocity and potential trapping into local search. Genetic algorithmcan find the global optimum and have a good robustness.Therefore, it has fast convergence velocity and strong self-learningability through the combination of genetic algorithms and neural networks.In this paper, the coal mine production safety is assessed by using an evaluation method of neural network based on genetic algorithm.
The“human-machine- environment”evaluation index system of coal mine safetyisestablished based on the accidentcausing theory. Zhang Yu-lin(2008)[13],Xu Yang(2009)[14]consider that the accident was caused due to unsafe state of human,machine and environment. Sun Jian-hua(2009)[4]think that the coal mine safety evaluation index system should include the factors of human, legislation,machine, engineering technology and disaster prevention.
Security information managementplays an important role in the process of coal mine production.Coal mine safety is affected because of backwardinformatizationconstruction in coal mine enterprise. The author thinks that the safety production of coal mineis influenced by the factors of human,machine,environment, management and information. The accident was the interaction of these factors’defects.So, “human-machine-environment- management- information”evaluation index systemof coal mine safety is established in this paper.
So,the coal mine safety is assessed by genetic neural network based on the evaluation index system. The genetic neural network has been applied in evaluation of the corporation’core competence and risk project. But there is no related research in the safety assessment of coal mine. The index system and assessment methods built in this paper have great realistic meanings and long-term meanings to enhanceearly-warning and evaluation theory of coal minesafety.
IIIEARLYWARNING ASSESSMENT INDEX SYSTEM OF COAL MINE
The index system of “people-machine-environment-
management-information” isshown in table 1.
IVEmpirical analysis based on early warning model of coal mine
The majorpartofgenetic neural networkis to optimize the weights of network.First it finds the optimal solutionbygenetic algorithm;it can narrow down the searching range. Then it will use the BP neural networkto find the optimalsolution[15].The specific stepsareas follows:
A.Determinethenetworkstructureofthemodel
The layersof neural networkinclude input,hidden and output layers.
1) Set the input of the network:To make the original datamore suitable for neural network through pretreating.The quantitative indicators should be normalized, and the qualitative indexes should be quantified. The number of network inputnodesare equal to the index number ofevaluation indexsystem. Therefore,the inputnodes in this paper are29.
2)Determine theoutput nodes and hidden layer nodes：The output nodes should be corresponded to the early warning assessment. The output nodesare 5 and the hidden layer nodesare15 in this paper based on the experience formula.Thecorresponding output results alertare shown in table 2
TABLE I Early warning assessmentindex system of coal mine
Rule layer / Index layerhuman factor
(X1) / Violation rateofemployees (X11)
Average level of education (X12)
training timeper month(X13)
machine factor
(Y1) / level of mining mechanization (Y11)
the rate of support equipment at good condition (Y12)
the rate of ventilation equipmentat good condition (Y13)
the rate of dust-proof equipment at good condition (Y14)
the rate of fire-fighting equipmentat good condition (Y15)
the rate of drainage equipmentat good condition (Y16)
the rate of lifting equipmentat good condition (Y17)
the rate of mechanical and electrical equipment good condition (Y18)
the rate of transport equipmentat good condition (Y19)
the rate of gas drainage equipmentat good condition (Y110)
Environment factor
(Z) / Thefactors of geological environment(Z1) / the averagefault throw (Z11)
number offaultbarsper unit area (Z12)
coal thickness coefficient of fault (Z13)
The degree of difficultyof controlling the roof (Z14)
factors of mine disaster (Z2) / spontaneous combustion period (Z21)
coal dust explosion index (Z22)
average Gas Emission (Z23)
mining surfacerichwatercoefficient (Z24)
factors of work environment (Z3) / pass rate of controlling dust pollution (Z31)
pass rate of controlling sound pollution (Z32)
Managementfactor(U1) / degree of perfectionon management system (U11)
capacityof emergency rescue (U12)
timeliness and effectiveness of management (U13)
information factor (V1) / degree of informatization (V11)
capacity of information recognition(V12)
capacity of information processing(V13)
TABLEⅡOutput result of neural network correspond to the alert
Output result / 10000 / 01000 / 00100 / 00010 / 00001Safe alert / highest / higher / medium / lower / lowest
B. Optimize weights of network by Genetic algorithm
The steps which optimize weights of network are as follows:
1)Population initialization:Acombination oftheinitializationfunction and therandom function is chosento selecttheinitial population. Cross-scale, crossover probability and mutation probability are included.
2)Determineencoding mode and evaluationfunction: Calculateselection possibility of each individual and selectindividual which have the biggest sufficiency value for the next-generation [16].
3)Operateselection, crossover and mutation:Population is operated by stochastic universal sampling,two-point crossover and uniform mutation.
4) Output the individualwith best fitness degree value:Select the neural network which has minimum errors and thresholds to train until the error reachesthe precision. Set the termination conditionand the error is less than 0.0001.
C. Empirical analysis of the model
In order to verify the feasibility and practicality of the model,network training is operated with monitoring data of five selected coal mining enterprises in four quarters in 2011. The error curve of network training isshown in figure 1. The training result shows that:the network training is completed becausethe network error is less than 0.0001 after 201 times of training.Select five samplesto test the network and it isshown in Table 3.The test resultsare consistent withthe practical situation. Therefore, it is proved that the model that build in this paper have right evaluation to the safety situation of coal mine, and it provides a scientific basis for policy-makersto judge the safety conditions and formulate the countermeasures.
VConclusions
The early-warning assessment systemin coal mine safety is a significant process that can preventand controlthe accident.This paper analyzes and identifies the potential accidents and risk factors which may affectsafety production, and build the genetic neural networks; it is proved that the model built in this paper has right evaluation to the safety situation of coal mine.Therefore, the model can assess the safety production in coal mine,and warn the weak in the production. Coal production could enter the safe orbit through timely adjustment inproduction by administrator.
Compared with previous studies,the contributions of this paper include：Firstly, the factorscontributedcoal mine safety production isanalyzedcomprehensively through the early warning index system including the information factors.Secondly,genetic neural network is able to achieve ideal empirical results when assessing the coal mine safety.
Fig.1. Training curve of BP algorithm
TABLE Ⅲ The test sample of model
X11 / X12 / X13 / Y11 / Y12 / Y13 / Y14 / Y15 / Y16 / Y17 / … / V11 / V12 / V13 / T / O1.00 / 0.80 / 0.60 / 0.40 / 0.60 / 1.00 / 0.80 / 0.87 / 0.53 / 1.00 / … / 0.67 / 0.12 / 0.37 / 0 0 0 1 0 / 0 0 0 1 0
1.00 / 0.75 / 0.87 / 0.43 / 0.80 / 1.00 / 0.87 / 0.33 / 0.60 / 0.00 / … / 0.80 / 0.75 / 0.60 / 0 0 1 0 0 / 0 0 1 0 0
0.88 / 0.43 / 0.27 / 0.28 / 0.60 / 0.60 / 0.80 / 0.13 / 0.13 / 0.27 / … / 0.60 / 0.43 / 0.13 / 0 0 0 1 0 / 0 0 0 1 0
1.00 / 0.52 / 0.40 / 0.51 / 0.50 / 0.87 / 0.87 / 0.60 / 0.60 / 0.60 / … / 0.50 / 0.52 / 0.60 / 0 0 0 1 0 / 0 0 0 1 0
0.58 / 0.25 / 0.13 / 0.38 / 0.77 / 0.47 / 0.33 / 0.07 / 0.00 / 0.00 / … / 0.77 / 0.25 / 0.00 / 0 0 1 0 0 / 0 0 1 0 0
REFERENCES
[1]Ding Xia-jun, Wang Bai-shun. “Application of fuzzy comprehensive evaluation in coal safety assessment”,Mining Safety & Environmental Protection,vol.31,pp.55-57,2004.
[2] Sun Jia,Sun Dian-ge,Li Lli-li,Jiang Zhong-an. “Application of fuzzy comprehensive evaluation in coal safety assessment of’one ventilation and theer-prevention’”, Mining Safety & Environmental Protection,vol.32,pp.74-78,2005.
[3] Gao wen-hua, Chen hong-zhang,Xie ke-ming.” Application of fuzzy comprehensive evaluation inLocal coal mine safety”, Mining Safety & Environmental Protection,vol.35,pp.83-85,2008.
[4]Sun jian-hua,Guo ying-xia,Zhang jin-peng,Qin xian-li,Zhang ke-chun. “The method based on multi-level fuzzy comprehensive in coal mine”, Safety in Coal Mines,vol.5,pp.125-128,2009.
[5] Xu yi-yong, Dai guang-long. “Minesafety evaluation based on grey system theory”, Mining Safety & Environmental Protection,vol.30,pp.10-11,2003.
[6] Cao shu-gang, Xu a-meng, Liu yan-bao, Zhang Li-qiang. “Comprehensive assessment of security in coal mines based on grey relevance analysis” ,Journal of Mining & Safety Engineering,vol.24,pp.141-145,2007.
[7]Fu Yong-shuai. “Study the applied research of gray theory in the coal mine safety evaluation” ,Ph.D. dissertation, Coal Science Research Institute,Beijin,China,2009.
[8] Yan Le-lin, Xu Jing-cai, Xu Man-gui, Wang Guo-qi. “Application of unascertained mathematics in preliminary evaluation of coal mine safety” .Mining Safety & Environmental Protection,vol.31,pp.24-25,2004.
[9] Huang Hui-yu, Li Cong-dong. “Research on the model for safety grade evaluation of coal mines based on artificial neural networks”, Industrial Engineering Journal, vol.10,pp.112-115,2007.
[10] Zhou Zhong-ke, Wang Li-jie. “Study on safety early-warning assessment in coal mine based on bp neural networks”, Journal of Safety Science and Technology, vol.7,pp.134-138,2011.
[11] Gao Xiao-xu, Dong Ding-wen, Yang Ri-li. “Application of BP neural networks to coal mine intrinsic safety assessment model”, Journal of Xi＇an University of Science and Technology, vol.31,pp.780-785,2011.
[12] Ding Bao-cheng, Shen Yu-zhi. “Application of compensatory fuzzy neural network in coal mine safety early warning”, Journal of Liaoning TechnicalUniversity, vol.13,pp.591-593,2011.
[13] Zhang Yu-lin. “Research on Coal Mine Safety Evaluation”,Ph.D. dissertation, Shenyang Institute of Technology,Shenyang,China,2008
[14] Xu Yang, Zhou Yan, Sun Xin, Yu Yue.“Synthesized assessment method for coal mine safety based on fuzzy analytic hierarchy process”,China Safety Science Journal, vol.19,pp.147-152,2009.
[15] Lin Guo-xi, Xuan Hui-yu. “Application research on tax forecasting based on genetic algorithms and BP artificial neural networks”,Systems Engineering-Theory Methodology Application, vol.14,pp.145-148,2005.
[16] Ni Lin, Liu Lin. “Comprehensive evaluation of vehicle recovery of reverse logistics based on genetic neural network”,Application Research of Computers, vol.28,pp.2866-2867,2011.