J. Cent. South Univ. (2012) 19: 1953-1960
DOI: 10.1007/s11771-012-1231-y
Integrating unascertained measurement and information entropy theory to assess blastability of rock mass
ZHOU Jian(周健), Li Xi-bing(李夕兵)
School of Resources and Safety Engineering, Central South University, Changsha 410083, China
? Central South University Press and Springer-Verlag Berlin Heidelberg 2012
Abstract: Due to the complex features of rock mass blastability assessment systems, an evaluation model of rock mass blastability was established on the basis of the unascertained measurement (UM) theory and the actual characteristics of the project. Considering a comprehensive range of intact rock properties and discontinuous structures of rock mass, twelve main factors influencing the evaluation blastability of rock mass were taken into account in the UM model, and the blastability evaluation index system of rock mass was constructed. The unascertained evaluation indices corresponding to the selected factors for the actual situation were solved both qualitatively and quantitatively. Then, the UM function of each evaluation index was obtained based on the initial data for the analysis of the blastability of six rock mass at a highway improvement cutting site in North Wales. The index weights of the factors were calculated by entropy theory, and credible degree identification (CDI) criteria were established according to the UM theory. The results of rock mass blastability evaluation were obtained by the CDI criteria. The results show that the UM model assessment results agree well with the actual records, and are consistent with those of the fuzzy sets evaluation method. Meanwhile, the unascertained superiority degree of rock mass blastability of samples S1-S6 which can be calculated by scoring criteria are 3.428 5, 3.453 3, 4.058 7, 3.675 9, 3.516 7 and 3.289 7, respectively. Furthermore, the proposed method can take into account large amount of uncertain information in blastability evaluation, which can provide an effective, credible and feasible way for estimating the blastability of rock mass. Engineering practices show that it can complete the blastability assessment systematically and scientifically without any assumption by the proposed model, which can be applied to practical engineering.
Key words: rock mass; blastability; unascertained measurement (UM) model; information entropy; prediction
1 Introduction
Rock mass blastability indicates the ease with which a rock mass can be fragmented by blasting during the excavation process. It’s also the comprehensive reflection of the integrated physical and mechanical properties of rock under dynamic loading. Moreover, the classification of rock blastability not only was applied to estimating explosive consumption for blasting engineering, but also provided the basic parameters for the blasting design (BD). Therefore, it is very important to give a practical and reasonable way to classify the blastability of rock mass [1-3].
Several approaches have been used for estimating blastability [1, 3-6]. Some researchers have tried to correlate it with the data available from laboratory and field testing of rock parameters; some others have related it with rock and blast design parameters; yet some have tried to estimate blastability through approaches based on the drilling rates and/or blast performances in the field. The traditional rock mass classification methods, only focused on rock itself, used some isolated indexes which do not consider characteristics of blasting engineering but those of rock mass [1]. RAKISHEV [4] devised ?ve blastability classes according to the value of an index named critical fracture velocity. HEINEN and DIMOCK [5] proposed a graphical method for the assessment of blastability index according to seismic propagation velocity in rock mass. LILLY [6] developed an empirical method to assess rock mass blastability. In fact, many factors may affect the rock-blasting behavior besides the characteristic of rock mass. The latest improvements in computer methods have also opened up new vistas for the researchers to use various artificial intelligence algorithms for determination of blastability [1, 6-9]. FENG [7] established an arti?cial neural network (ANN) approach for comprehensive classification of rock stability, blastability and drillability. XUE et al [2] proposed the attribute recognition model (ARM) of assessment of classification for rock mass blastability in engineering blasting based on the theory of attribute mathematical theory. LATHAM and LU [3] proposed a blastability designation (BD), as a part of their developed energy-block transition (EBT) model, for prediction of blasted block size distribution (BBSD). AZIMI et al [8] investigated a fuzzy logic based blastability designation predictor model. A feed-forward back-propagation neural network was developed by HAN et al [9] to classify rock mass blastability. However, the developed ANN model has some limitations, such as black box approach, arriving at local minima, overfitting problem and absence of probabilistic output. Furthermore, there is not proper method to determine the number of hidden layers in the ANN model. The developed fuzzy logic model determines the fuzzy rules with difficulty.
Comprehensive classification of the blastability of rock is one of the most complex, knotty problems in rock mechanics and blast design. For this case, an effective method to assess the blastability of rock mass should be presented based on the unascertained mathematics theory in the current work. Unascertained mathematics has been proposed by WANG [10], which is a tool to describe subjective uncertainty quantitatively. It deals mainly with unascertained information which differs from stochastic information, fuzzy information, and grey information. Unascertained information refers to information demanded by decision-making over which the message itself has no uncertainty. Since the 1990s, LIU et al [11] and other scholars [12-15] have done much work and unascertained mathematics has been successfully used in many fields. It is, therefore, motivating to investigate the capability of UMM in rock mass blastability evaluation.
2 Details of unascertained measurement (UM) model
2.1 Determining evaluation space
Suppose that Γ has n objects to be assessed, and the assessment object space is Γ=(Γ1, Γ2, …, Γn). Each object of Γi (i=1, 2, …, n) has m evaluating indices, so the evaluating index space is Ψ=(Ψ1, Ψ2, … , Ψm). Then Γi can be denoted to be m-dimensional where is the measured value of evaluation object, Ri, with respect to evaluating index For each (i=1, 2, …, n; j=1, 2, …, m), we assume that there are p evaluation grades of (Δ1, Δ2, …, Δp).
The evaluation space is Θ, denoted as Θ=(Δ1, Δ2, …, Δp). Suppose Δk (k=1, 2, …, p) is the k-th evaluation grade, and the k-th grade is higher than that of the (k+1)-th one, denoted as Δk>Δk+1. If the grading rank (Δ1, Δ2, …, Δp) satisfies Δ1>Δ2>…>Δk+1 or Δ1<Δ2<…<Δk+1, (Δ1, Δ2, …, Δp) is called the ordered partition class of evaluation space Θ [11-15].
2.2 UM of single index
Denote the UM as = , where is the degree of belonging to the k-th evaluation grade of Ck , which satisfies the following principles [11-15]:
1) Nonnegative and limited principle
(1)
2) Convergent principle
(2)
3) Additive principle
(3)
As an UM, these three principles must be met simultaneously; otherwise, it should be called as an estimate. The UM matrix of single index is constructed as
(4)
In practice, decision-maker should decide the specific structure based on the background, knowledge and prior knowledge [11-15]. At present, there are several commonly used methods to construct measure function, including straight-line method (SLM), quadratic curve method (QCM), sine curve method (SCM) and exponential curve method (ECM). Among them, the SLM function is the most widely used and the simplest UM function. The straight-line unascertained measure function is used in this work and the function graph is shown in Fig. 1.
Fig. 1 Straight-line UM function
The corresponding function expression of the above
UM function is
(4)
2.3 Identification weight of index
When the membership of every index has been known, the “importance” of the weight has only one meaning, that is, the importance for the index classification data to determine the classification of the index. So, the index weight used in determining the composed membership by the single index is and only is the identification weight of the index. In this case, the common methods used to get the weight, such as Delphi method, Brainstorming, and Analytic hierarchy process (AHP) are helpless. The method of determining the index identification weight by using the information entropy is introduced as follows [2, 12].
SHANNON proposed the conception of information entropy, which is employed to measure the uncertainty roughly [12, 16]. Suppose that wj is the relative important extent of measured index Ψj compared with
other indices. wj satisfies 0≤wj≤1, and , which
is called the weight of Ψj. Index weight vector w is characterized by w=(w1, w2, …, wm). Then, wj is given by
(5)
(6)
where 0≤χi≤1. The evaluation matrix of UM of single index is known, so wj can be obtained by Eqs.(5) and (6).
2.4 Composite UM of multiple indices
Parameter υik is denoted as the degree of the assessment object Γi belonging to the k-th evaluation grade of Δk. When υik is equal to υ (Γi∈Δk), υik is called the composite UM of multiple indices. After the single index matrix and the identification weight are derived, considering the meaning of the identification weight, the composite UM of multiple indices can be worked out. The comprehensive assessment vector is defined as
(7)
where υik satisfies 0≤υik≤1 and .
2.5 Principle of identification
Because the classification of the comment ranks (Δ1, Δ2, …, Δp,) is ordered, e.g., Δk is “better” than Δk+1, the maximum measure identification (MMI) principle is not available. The credible degree identification (CDI) principle is needed. Let the credible identification be λ, where 0.5≤λ≤1, and it is always 0.6 or 0.7 [11]. If the evaluation space meets Δ1>Δ2>…>Δp, Let
(8)
Then, φi is the k0 -section appraisal grade Δki. The implication is as follows. The confidence that the grade of φi is not higher than k0 is λ or the confidence that the grade of sample φi is higher than k0 is 1- λ.
Suppose that the score value of Ct is It, JΔi is given by
(9)
where is the unascertained superiority degree of evaluation object Δi, so J=(JΔ1, JΔ2, … JΔi) is called the vector of unascertained superiority degree. The superiority degree of Δi is ordered according to the magnitude of JΔi.
3 UM model for evaluating blastability of rock mass
The approach for the development of the UM-based assessment for the blastability of rock mass can be divided into seven stages: 1) Choice of the criterion indices for rock mass blastability classification; 2) Collection of data sets of rock mass; 3) Construction of UM function of single index; 4) Determination of the entropy weight of criterion indices; 5) Combined UM of multiple indices of rock mass blastability; 6) Establishment of the final UM model for evaluating the blastability of rock mass with the help of credible degree identification (CDI) principle; 7) Evaluation and validation of the UM model by evaluation with testing data and comparing it with literature correlations, as shown in Fig. 2.
3.1 System of assessment indices
Selecting the correct classification of rock blastability evaluation is to ensure that evaluation results are reasonable, economic and reliable in the pre-conditions. On the basis of the related studies [1-8], twelve factors affecting the blastability evaluation for rock mass are selected as the evaluation indices, which are strength, resistance to fracturing, sturdiness of rock, elasticity of rock, resistance to dynamic loading, hardness of rock, deformability, resistance to breaking, in-situ block sizes, fragility of rock mass, integrity of rock mass and discontinuity plane’s strength, designated as P1, P2, P3, P4, P5, P6, P7, P8, P9, P10, P11 and P12, respectively. The system of blastability assessment indexes is shown in Fig. 3.
Fig. 2 Principle flow chart for proposed UM-based approach for classification of blastability of rock mass
Fig. 3 Index systems of blastability comprehensive assessment of rock mass
Table 1 Suggested quantitative indications for classi?cation of blastability of rock mass associated with individual factor [3, 8]
And the classification standards for these indices are determined with reference to Refs. [3, 8], as shown in Table 1. The evaluation set is {C1, C2, C3, C4, C5}, where designated classes C1, C2, C3, C4 and C5 are denoted as very easy, easy, moderate, difficult, and very difficult, respectively.
3.2 UM model development and validation
The blastability assessment system developed above has been applied to a case study that assesses the blastability of the rock mass at a highway improvement cutting site in North Wales (hereafter referred to as the G cutting site) [3, 8]. Therefore, UM model was established using the theory discussed. The evaluation indexes of UM model for evaluating of classification of the blastability of rock mass was established based on the unascertained mathematics theory and the actual characteristics in this work. The evaluation indexes of the proposed model are as follows:
1) Construction of UM function of single index
The UM functions of single index were constructed to get the value of the evaluation factors, on the basis of the definition of the UM function and the classification in Table 1. The UM function of each index is illustrated in Fig. 4. Then, the evaluation matrix of UM of six rock mass sample could be obtained, according to the functions in Figs. 4(a)-(i) and values of the factors given in Table 2. Taking sample S1 for example, the values of the eight evaluation indices for S1 in Table 2 were substituted into the corresponding UM functions in Figs. 4(a)-(i), respectively. Then, the evaluation matrix of UM of S1 was calculated as
(10)
2) Composite UM of multiple indices of rock mass blastability
The weights of the indices were determined by Eqs.(1)-(6). So, the weights of Γi denoted as w=(w1, w2,…, w8) were (0.107 0, 0.100 7, 0.109 7, 0.101 2, 0.111 7, 0.176 9, 0.176 9, 0.115 7). Then, the composite UM of multiple indices of S1 were calculated as (0.145 0, 0.082 8, 0.197 3, 0.314 2, 0.260 7).
3) Evaluation results recognizing
In the present study, the grade of blastability of rock mass is divided into five grades: very easy, easy, moderate, difficult, and very difficult, therefore the credible degree identification (CDI) criteria were adopted to judge the grade of Γi instead of the maximum measurement identification rule, due to the sequence of the evaluating grade (C1, C2, C3, C4, C5). And the confidence level λ is taken as 0.6. According to Eq. (7) of composite UM vector of multiple indices and Eq. (8) of CDI, k0 =0.746 5, which is larger than λ in descending order, thus the blastability degree of S1 belongs to grade C3. The final assessment results show that the sample S1 belongs to the third rank, which means “moderate”.
From the above, we can get that the several identified results are in accordance with score value criteria. Therefore, the superiority degree of R1 is determined as grade C3, that is, the blastability of sample S1 is moderate. In the same way, rock mass blastabilities of other samples are evaluated. The composite UM of multiple indices and the evaluated results are shown in Table 3 and Fig. 6. It can be concluded that blastabilities of sample S1-S6 are moderate, moderate, difficult, moderate, moderate, and moderate. This shows that the unascertained superiority degrees of sample S1-S6 are 3.428 5, 3.453 3, 4.058 7, 3.675 9, 3.516 7 and 3.289 7, respectively (Table 4), that is, S3>S4>S5>S2>S1>S6.
3.3 Discussion
In order to compare the difference between our methods and other evaluation methods, the evaluation results of blast design [3], LATHAM and LU method [3] and Fuzzy set model [8] are listed in Table 5. We can see from the UM model evaluation results and actual assessment results that the classification results of both Fuzzy models are accordant with the actual situation basically. Compared with conventional statistics, UM theory is an exclusive theory for the study of machine learning law in the condition of small sample number.
The above-mentioned comparisons indicate that all four models are competitive with each other for blastability of rock mass, but the performance of the UM model is relatively superior to the others. Combining the theory with the actual case will verify the feasibility and practicality of the model. Employing the information entropy and credible identification principal, the shortcoming of fuzzy set assessment is overcome and the result becomes more objective. It is apt to realize scientific and rational decision-making. The reasons are that the UM pays more attention to the order of the assessment space and gives the rational rank and credible identification principles. All of those are not possessed by fuzzy set assessment. Besides, the computation process of the UM is more simple.
Fig. 4 UM functions of evaluation indices: (a) Uniaxial compressive strength; (b) Point-load strength index; (c) Uniaxial tensile strength; (d) Elasticity of rock; (e) Mean in-situ block sizes; (f) P-wave velocity; (g) Schmidt hardness value; (h) Density; (i) Fractal dimension of in-situ rock mass
Table 2 Test results of rock specimens data of evaluation indices (Data from Refs. [3, 8])
Table 3 Distinction results of rock mass samples using UMM with entropy
Fig. 6 Results of UM model by CDI criteria in descending order
Table 4 Score results of rock mass samples
Table 5 Comparasion results of different methods
4 Conclusions
1) An UM model of blastability of rock mass predictions is established based on uncertain mathematics theory. In the UM model approach, information entropy is used to determine the indice weights of the factors, and the field data sets are used to investigate the feasibility in the blastability prediction for rock mass. And then, the evaluation results are obtained according to the CDI criteria. the UM model underestimates the blastability of the studied rock masses (G cutting site) in comparison with the conventional methods and FIS model. The experimental results show that the classi?cation accuracies of UM model are superior to those of FIS model.
2) The evaluated blastability results of samples S1-S6 are moderate, moderate, difficult, moderate, moderate, moderate, respectively. It is shown that the unascertained superiority degrees of samples S1-S6 are 3.428 5, 3.453 3, 4.058 7, 3.675 9, 3.516 7 and 3.289 7, respectively, in the order of S3>S4>S5>S2>S1>S6.
3) Combining the theory with the actual case will verify the feasibility and practicality of the model. Employing the information entropy and credible identification principal, the shortcoming of fuzzy comprehensive assessment is overcome and the result becomes more objective. It is easy to realize scientific and rational decision-making. The application results show that it is easy to realize the assessment without any assumption. The UM approach will be used for prediction of mining, rock or geotechnique engineering in future research.
References
[1] KAUSHIK D, PHALGUNI S. Concept of blastability—An update [J]. The Indian Mining & Engineering Journal, 2003, 42(8/9): 24-31.
[2] XUE Jian-gang, ZHOU Jian, SHI Xiu-zhi, WANG Huai-yong, HU Hai-yan. Assessment of classification for rock mass blastability based on entropy coefficient of attribute recognition model [J]. Journal of Central South University: Science and Technology, 2010, 41(1): 251-256. (in Chinese)
[3] LATHAM J P, LU P. Development of an assessment system of blastability for rock masses [J]. International Journal of Rock Mechanics and Mining Sciences, 1999, 36(1): 41-55.
[4] RAKISHEV B R. A new characteristics of the blastability of rock in quarries [J]. Soviet Mining Science, 1982(17): 248-251.
[5] HEINEN R H, DIMOCK R R. The use of sonic measurements to determine the blastability of rocks [C]// Proceedings of the Second Conference on Explosive and Blasting Techniques. Luisville, Kentucky, 1976: 234-248.
[6] LILLY P. An empirical method of assessing rock mass blastability [C]// Large open pit mines Conference. Newman, Australia, 1986: 89-92.
[7] FENG X T. A neural network approach to comprehensive classification of rock stability, blastability and drillability [J]. International Journal of Mining, Reclamation and Environment, 1995, 9(2): 57-62.
[8] AZIMI Y, OSANLOO M, AAKBARPOUR-SHIRAZI M, AGHAJANI BAZZAZI A. Prediction of the blastability designation of rock masses using fuzzy sets [J]. International Journal of Rock Mechanics and Mining Sciences, 2010, 47(7): 1126-1140.
[9] HAN J, WEIYA X, SHOUYI X. Arti?cial neural network method of rock mass blastability classi?cation [C]// Proceedings of the Fifth International Conference on GeoComputation. London, UK, 2000: 23-28.
[10] WANG Guang-yuan. Mathematics treatment of unascertained information [J]. Journal of Civil Engineering Institute of Harbin, 1990, 23(4): l–9.(in Chinese)
[11] LIU K D, CAO Q K, PANG Y J. A method of fault diagnosis based on unascertained set [J]. Acta Automatic Sinica, 2004, 30(5): 747-756.
[12] SHI Xiu-zhi, ZHOU Jian, DONG Lei, HU Han-yan. WANG Huai-yong, CHEN Shou-ru. Application of unascertained measurement model to prediction of classification of rockburst intensity [J]. Chinese Journal of Rock Mechanics and Engineering, 2010, 29(S1): 2720-2727. (in Chinese)
[13] LIU Ai-hua, DONG Lei, DONG Long-jun. Optimization model of unascertained measurement for underground mining method selection and its application [J]. Journal of Central South University of Technology, 2010, 17(4): 949-961.
[14] SHI Xiu-zhi, ZHOU Jian. Application of uncertainty average clustering measurement model to classification of tunnel surrounding rock [J]. Journal of Civil, Architectural & Environmental Engineering, 2009, 31(2): 62-67. (in Chinese)
[15] Li Y C, SUO J J. Safety assessment of platform loadout procedures based on unascertained measures [J]. Chinese Journal of Oceanology and Limnology, 2007, 25(4): 354-358.
[16] JAYNES E T. Information theory and statistical mechanics [J]. Physical Review, 1957(106): 620-630.
(Edited by YANG Bing)
Foundation item: Project(50934006) supported by the National Natural Science Foundation of China; Project(2010CB732004) supported by the National Basic Research Program of China; Project(2009ssxt230) supported by the Central South University Innovation Fund, China; Project(CX2011B119) supported by the Graduated Students’ Research and Innovation Fund of Hunan Province, China
Received date: 2011-05-12; Accepted date: 2011-08-01
Corresponding author: ZHOU Jian, PhD; Tel: -86-13723887261; E-mail:csujzhou@126.com