* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Download The Economic Optimization of Mining Support Scheme Based on
Pattern recognition wikipedia , lookup
Multi-armed bandit wikipedia , lookup
Fuzzy logic wikipedia , lookup
Mixture model wikipedia , lookup
Gene expression programming wikipedia , lookup
Catastrophic interference wikipedia , lookup
Agent-based model in biology wikipedia , lookup
Neural modeling fields wikipedia , lookup
Convolutional neural network wikipedia , lookup
The Economic Optimization of Mining Support Scheme Based on Optimal Fuzzy Model WANG Min, RU Zhong liang College of Civil Engineering,Henan Polytechnical University, [email protected] P.R.China, 454000 Abstract The support scheme is one of the main factors that affect the safety of coal mines. It is not only affected the mine’s safety but also caused a big cost for the enterprise. This paper bring out a optimal fuzzy model combined genetic algorithms (GAs) with artificial neural network(ANN). ANN is used to establish the complexity relationship between the geologic condition and the support parameters, and GA is adopted to in the model to search the optimal cost for the support scheme. This model then is applied to a coal mine; a safe and economic support scheme is brought out by it. Key words fuzzy model, support scheme, genetic algorithms, natural network. 1 Introduction The support system is a major engineering of the coal mine since the improper design of support systems can lead to under-design and costly failures or over-design and high tunnel costs[1]. In these cases, the goal of any support system is to achieve safety production and economical cost. The support needed to accomplish this objective depends on the mechanical computing and economical optimization. From the mechanical view, many researchers have done many works range from the rock properties, bolt mechanics and numerical simulation et al. A review of support mechanisms provided by liners has been given by Stacey [2]. Finite difference method is the most usually method used in rock engineering, such as FLAC software, Caia [3], Medhurst and Reed[4], Wang and Tannant [5] used discrete element models to simulate numerically a liner round a tunnel. In the early 1990s, with the development of computer hardware and software, the concept of Soft Computing was introduced to the engineering[6]. Soft Computing is an evolving collection of artificial intelligence methodologies, such as Fuzzy Logic (FL), Artificial Neural Networks (ANN), and Genetic Algorithms (GAs). Research has been deployed in the direction of applying SC to engineering design. Yu[7] proposed an intelligent method combining artificial neural networks and evolutionary calculation for the effective displacement back-analysis of earth-rockfill dams, and applied it to two dams projects. A genetic algorithm-based searching technique is adopted in the system to provide the optimal or near-optimal combination of production sequences[8]. SC techniques yield rich knowledge representation (symbol and pattern), flexible knowledge acquisition (machine learning), and flexible knowledge processing. Also it can either be deployed as separate tools or be integrated in unified and hybrid architectures. In this paper a GA optimal model was developed by combined neural network and genetic algorithm. The neural network model sets up the complexity relationship between the rock mechanics and support parameters. The genetic model can search the optimal cost of the support design. Combined those two model, a GA optimal fuzzy model is brought out. In this paper, section 2 introduces the fuzzy model base on neural network, section 3 expatiate the genetic algorithm and propose the optimal fuzzy model, section 4 is a case which apply the model to a coal mine and deigned a support scheme. 2 . Fuzzy model for the support scheme Artificial neural network (ANN) is an artificial intelligence technique designed to stimulate human brain activity. An ANN is a massively parallel-distributed processor that has a natural propensity for storing experiential knowledge and making it available for use. It allows for self-learning, self-organization, and parallel processing, and is well suited to problems involving matching an input pattern to a set of output patterns where deep reasoning is not required. The training processes of ANN 654 are usually complex and high dimensional problems. A fatal drawback of the commonly used gradient-based BP algorithm, which is a local search method, is its easy entrapment into local optimum point. The artificial neural network is developed by using a number of such artificial neurons as described above by trial and error method. Many different configurations of the artificial neuron can be made to develop different network configurations. The configuration so developed is trained using a predetermined learning algorithm. In this paper, back-propagation neural network(BPN) were used to analyze the relations between the geological parameters and the support scheme. The network is consisted by three layers, as shown in Fig 1. The input layer consists of neurons, corresponding to the elements of input vector, in this study make geological parameters surrounding rock elastic module, coal elastic module, coal thickness, pillar width, tunnel area etc as the input data. These neurons send the values of independent variables to the neurons of succeeding hidden layer. Each neuron of the hidden layer is assigned a fixed center, that is of the same dimension as the input vector. The centers are usually chosen from the training dataset. A neuron of the hidden layer receives input vector and emits the output vector, which is the support scheme parameters support type, bolt length, bolt space, bolt section area etc. In the course of training, a collection of about 100 successful support schemes in our country is repeatedly presented as the training data. The weights in the network are adjusted until the errors between the target and the predicted outputs are small enough, or a pre-determined number of epochs are passed. The perceptions are then validated by presenting with an input vector not belonging to the training pairs. L Erokc Ecoal D Dcoal . . . . . . · · · · · · · · · Acoal S Hcoal T Input Layer Hidden Layer Output Layer Fig. 1 The sketch of neural network 3. Optimal model based on GA-NN 3.1 Genetic algorithms Genetic algorithm (GA) was dated to the 1960s by Holland [9]. It has been popularly used in many areas: constrained or unconstrained optimization, scheduling and sequencing, transportation, reliability optimization, artificial intelligence, and many others. Genetic algorithms are stochastic search techniques based upon the mechanism of natural selection and natural genetics. GA continually exploits new and better solutions without any pre-assumptions, such as continuity and unimodality. GA has been successfully applied to many complex optimization problems and shows its merits to traditional optimization methods, especially when the system under study has multiple optimum solutions[10]. In GA, potential solutions to a problem are represented as a population of chromosomes and each chromosome stands for a possible solution at hand. The number of chromosomes in a population is referred to as population size. The chromosomes evolve through successive generations (same as 655 do-loops in a computer program). offspring chromosomes are created by merging two parent chromosomes using a crossover operator, or modifying a chromosome using a mutation operator. During each generation, the chromosomes are evaluated on their performances with respect to the fitness functions (i.e.,objective functions). Fitter chromosomes have higher survival probabilities. After several generations, chromosomes in the new generation may be identical, or certain termination conditions are met. The final chromosomes hopefully represent the optimal or near-optimal solutions to a problem. 3.2 Optimal fuzzy model From section 2 we can get the support pattern and the parameters, but it would not be the best one, because that those schemes haven’t consider their economic. So we must set up an optimal model can consider both safety and cost of support system. In this study the support system optimization problem can be formulated in the following form: min Fcos t ( x1 , x 2 , x3 , L) st. Fsafety ( x1 , x 2 , x3 , L) > [ Fs ] (1) Where, Fcos t is the cost of the support system, x1 , x 2 , x3 , L means the factors of the support. So our objective is to get the lowest cost while satisfied the safety conditions. Hybridize the GA and NN fuzzy model, the following optimal model was designed(Fig. 2). Start Generation of initial population y1 New generation y2 yn … … … Mutation y1 y2 yn Crossover No Selection Fsafety>[F] Yes No min Fcost Yes Output Fig. 2 The sketch of support scheme optimal fuzzy model 4 Case study Apply the optimal fuzzy model to a mine in Shanxi province, this mine’s major geological parameters surrounding rock elastic module, coal elastic module, coal thickness, pillar width, tunnel area are listed in table 1. The subentry costs of the support engineering are list in table 2. The Bp network we used has been trained by many successful support schemes in many other coal mines. A tunnel support scheme was designed using the optimal model, and GA was used to search for minimum project direct cost. Table 3 lists the proper support schemes and the optimal scheme cost. From the table we can see that the most economical scheme, but its safety factor is only 1.13, considering the safety production, so we abandon it, and choose 656 the scheme No.6 as the best and safety support scheme. The optimal support scheme can save about 7.6% cost for the mine. Rock Elastic Module(GPa) 22.0 Cablel(yuan/m) Coal Elastic Modul(GPa) 4.3 1 2 3 4 5 6* Pillar width (m) 6.0 Table 2 Subentry costs of the support Bolt(yuan/m) Anchor(yuan) 24.52 Scheme No. Table 1. Geological parameters Tunnel Depth Coal thickness (m) (m) 70 3.5 26.74 30.0 0.80 0.85 0.82 0.81 0.92 0.90 Labour(yuan) 40.0 Table 3 Support schemes of the tunnel Number of Bolts Bolt Space(m) Safety factor 472 433 459 467 410 414 Tunnel area (m2) 12.5 1.86 1.34 1.54 1.58 1.13 1.27 Total cost(yuan) 46,086 42,278 44,816 45,597 40,032 40,422 * the optimal scheme be adopted. 4 Conclusion This paper bring up a GA-based optimal fuzzy support model that can help the mines to get the best optimal and safety tunnel support scheme. Compared with the traditional methods, the GA-based optimal model proposed in this paper has several advantages. First, this paper adopts fuzzy neural network theory to construct the relationship between geological parameters and support parameters; the network has been trained by many successful support schemes, so it can obtain the proper support parameters quickly and reliably. Second, the evolutionary strategy we used in the model is genetic algorithm. GA is a stochastic global search, so it exploits new and better solutions without any pre-assumptions, so the solution is the best point in the global space. Third, the optimal model not only consider the economy of the scheme but also take account into the safety factor, so the support scheme obtained from the optimal model is a robust result. Acknowledgments. Financial supports from Henan Science Foundation(2008A440005), Henan Polytechnical Doctor Foundation(648508) and Key Teacher Foundation(649043) are highly appreciated. References [1] Evert Hoek. Tunnel support in weak rock[J]. Keynote address, Symposium of Sedimentary Rock Engineering, Taipei, Taiwan, 1998,12 [2] Stacey TR. Review of membrane support mechanisms, loading mechanisms, desired membrane performance, and appropriate test methods[J]. Journal of Souther African Institute Mining Metallurgy 2001(101):343~51. [3] M. Caia, P.K. Kaisera, H. Morioka. FLAC/PFC coupled numerical simulation of AE in large-scale underground excavations[J]. International Journal of Rock Mechanics & Mining Sciences, 2007(44): 550~564 [4] Medhurst, T.P., Reed, K.. Ground response curves for long wall support assessment. Mining 657 Technology: IMM Transactions section A , 2005(114):81~88. [5]Wang C, Tannant DD. Rock fracture around a highly stressed tunnel and the impact of a thin tunnel liner for ground control[J]. International Journal of Rock Mechanics & Mining Sciences, 2004(41):490~497. [6]K.M. Saridakis, A.J. Dentsoras. Soft computing in engineering design-A review[J]. Advanced Engineering Informatics, 2008(22): 202~221 [7]Yuzhen Yu. An intelligent displacement back-analysis method for earth-rockfill dams[J]. Computers and Geotechnics, 2007(34): 423~434 [8]Sou-Sen Leu, Shao-Ting Hwang. GA-based resource-constrained flow-shop scheduling model for mixed precast production[J]. Automation in Construction 2002(11): 439~ 452 [9]Holland J. Adaptation in natural and artificial systems[M]. Ann Arbor, MIT University of Michigan Press.1975 [10]Back T., Hammel U, Michalewicz Z.. Handbook of Evolutionary Computation[M]. Oxford University Press, New York, and Institute of Physics Publishing, Bristol,1997. 658