Development of an expert system for rock mass classification

This paper documents the development of an expert system for rock mass classification for use on microcomputers. The expert system called. EMC was. developed using an expert system shell FLOPS. FLOPS is a shell based on the Al language 0PS5 with many unique features. The system RMC utilizes most of these unique features. Among the features used are approximate reasoning with fuzzy sets, the blackboard architecture and the emulated parallel processing of fuzzy production rules. Details of the development of RMC including discussion of some relevant fuzzy set backgrounds are described in this paper.     

基于Mamdani模糊推理的山岭隧道围岩RMR14分级

岩体分级方法(RMR14)可充分评价岩体力学性质和结构面的条件,且可考虑初始地应力场和开挖方式等工程因素的影响,非常适用于山岭隧道围岩的评价。但RMR14评级方法的阶梯状评分会引起评分区间边界出现模糊不确定性,为更加准确地评价岩体质量,基于Mamdani模糊推理方法,引入模糊隶属函数来解决模糊不确定性问题。采用"if-then"推理法则实现多输入参数多推理规则的模糊推理,提出基于Mamdani模糊推理的RMR14岩体分级方法,并将此方法应用于独平高速漂里隧道的围岩质量评价。研究结果表明,基于Mamdani模糊推理的RMR14岩体质量分级方法既可借鉴既有工程经验,又可恰当处理评分边界的模糊不确定性问题;与直接采用RMR14分级方法相比,本文提出的模糊分级方法能更加准确评价围岩质量。     

Estimation of deformation modulus of rock masses by using fuzzy clustering-based modeling

The paper describes a method which incorporates Takagi–Sugeno (TS) fuzzy modeling with two data clustering approaches including fuzzy-c-means (FCM) clustering and subtractive clustering to estimate the rock mass modulus of deformation. For this aim, a database including 120 cases collected from several galleries of dam sites locations was established. The information returned by fuzzy clustering was initially used to define the number of rules and antecedent membership functions and afterwards linear least squares estimation implemented to obtain fuzzy consequent parameters. An adaptive neuro-fuzzy inference system (ANFIS) was applied to modify the pre-determined TS clustering-based model structures to improve the generalization performance of those. For evaluation of the performance, root mean square error (RMSE) and variance account for (VAF) values have been utilized as performance criteria. It can be said, that ANFIS approach enhances the performances of fuzzy clustering-based models in predicting modulus of deformation of rock masses successfully.     

Engineering geological evaluation and preliminary support design for the metro extension tunnel,Ankara, Turkey

The paper reports an assessment of the engineering geological characteristics of the rock mass to be encountered between Mecidiye and Gazino stations on the new extension of the Ankara metro and the determination of appropriate support and excavation methods. The rock mass quality was estimated using the rock mass rating (RMR), geological strength index (GSI) and rock mass quality (Q) systems and the tunnel divided into sections. The RMR, Q and NATM systems were used to determine the support and excavation methods in these areas. The deformations and stress concentrations around each tunnel section were investigated and the interaction of the support systems with the rock mass was analyzed using finite element software. It is concluded that rock mass classification systems should be used in tandem with numerical tools, although it is emphasized that the estimation of rock mass properties is not an exact science and both rock properties and numerical models should be refined based on observations and the results of instrumentation installed during the construction of a tunnel.      

Prediction of the blastability designation of rock masses using fuzzy sets

The main objective of rock blasting design is to achieve a balance among optimum powder factor, proper fragmentation, throws, ground vibration, etc. The in-situ rock mass properties are among the most important contributory factors in fragmentation. The term blastability is used to indicate the susceptibility of the rock mass to blasting and its characterization has become a pressing task for blasting operations. Several approaches have been used for estimating blastability. Despite their widespread use in practice, they have some common deficiencies leading to uncertainties in their practical applications through sharp transitions between two adjacent rating classes and the subjective uncertainties on data, which are close to the range boundaries of rock classes. In this study, the fuzzy set theory was applied to blastability designation (BD) classification systems. Furthermore, a new methodology in terms of "Effective Rules" is developed in construction of rule base part of the Mamdani fuzzy inference system structure, to efficiently solve fuzzy inference systems with a large number of fuzzy rules (e.g. nearly 400,000 rules). In comparison with the conventional methods, it was seen that the fuzzy model operated more consistently. Moreover, it was shown that the fuzzy set theory could effectively overcome the uncertainties encountered in the practical applications of conventional classification systems.     

基于人工智能专家系统的岩爆烈度分级预测研究

岩爆是岩石的一种复杂动力失稳破坏。由于影响岩爆发生的因素众多,因此很难做到对其发生的可能性及烈度进行准确预测。本文基于人工智能专家系统方法,针对岩爆预测过程中出现的模糊性和随机性特点,建立了岩爆烈度分级模糊综合预测模型。通过岩爆烈度分级预测的知识获取,首先选取岩石脆性系数σc/σt、岩石应力系数σθ/σc、初始应力水平σ1/σc、弹性能量指数Wet及岩石脆性指数Is五个因素作为岩爆评价指标,采用模糊数学方法确定各指标的模糊隶属函数及权重|然后以规则编程的形式对所获取的知识进行表示,建立了岩爆烈度分级预测专家系统的知识库,并编制了相应的分析和计算程序|最后利用国内外岩爆工程实例对所提方法进行验证,预测结果与实际情况相吻合,从而为岩爆的预测提供了一种新的有效途径。     

Developing a fuzzy model based on subtractive clustering for road header performance prediction

Road header performance prediction plays a significant role in the successful implementation of a tunneling project; so that, there is a need for accurate prediction of the advance rate of tunneling. However, there is relatively less study on predicting the performance of such machinery by using soft computing techniques although they have some advantages over the other methods. On the other hand, often models applied for road header performance prediction neglect interaction between machine and rock mass parameters. The Takagi–Sugeno (TS) fuzzy system model, one of the most popular fuzzy models, can be applied to solve complex problems by transferring a nonlinear system into a set of linear subsystems. However, in many situations, it is not convenient to identify all the rules; so, using the fuzzy clustering techniques in which the rules are resulted from measured data can be useful and valuable. In this paper, a new model based on the geological and geotechnical site conditions is developed to predict the road header performance. The model is developed using soft computing technique that applies the concept of fuzzy logic to take into account the uncertainty and complexity derived from the interaction between rock properties and road header parameters. The prediction capabilities offered by TS fuzzy model based on subtractive clustering method are demonstrated by using field data of obtained from Tabas coal mine in Iran.     

Geotechnical aspects of rock tunnelling in China

To meet the huge demands of transportation, energy and other infrastructure projects, a large volume of rock tunnelling is being carried out across the country as China embarks on her modernization plan. This paper summaries the geotechnical issues encountered on such projects, including the various modes of rock mass instability and geological hazards. Methods of evaluating rock mass quality and ground conditions as developed in the West are now widely used on tunnelling projects in China. As well Chinese engineers and researchers have also developed geomechanical systems for rock mass classification, failure modes and support requirements. These systems have been proven to be effective in dealing with the complex geologic and tunnelling conditions in China. Also described in the paper are analytical techniques currently used in China for tunnel performance prediction, including canonical analysis, expert systems, block vector analysis and advanced numerical modelling techniques.     

岩石地下工程锚喷支护设计的人工智能方法及其集成

提出了一种可用于岩石地下工程锚喷支护设计的人工智能综合设计方法。这种方法是首先采用专家系统、人工神经元网络和案例设计3种人工智能方法对同一岩石工程分别进行支护设计，然后再用分区多专家决策算法将所得3种设计结果加以综合，形成最终设计方案，它充分利用了各种人工智能方法的特长，有效地避免了单一方法所容易出现的偏见。应用表叫，人工智能综合设计方法能把工程类比法提高到一个新高度。     

On the use of the RMR system for estimation of rock mass deformation modulus

Estimation of rock mass deformation modulus is the subject of many studies in rock engineering research work. Although numerous predictive models have been developed for the estimation of the deformation modulus, they cannot be generalized for other sites because of inadequate accuracy. Furthermore, it is very valuable that the predictive models involve some accessible input parameters. The rock mass rating (RMR) is a well-known geomechanical parameter, which is usually determined to describe the quality of rock mass in rock engineering projects. In this study, five parameter ratings of the RMR classification system are used to predict the deformation modulus of rock mass in the abutment of the Gotvand earth dam. Statistical analysis and an artificial neural network are employed to present two new predictive models. Finally, probabilistic analysis is used to predict the rock mass deformation modulus, which overcomes the low accuracy caused by the inherent uncertainty in prediction. The results indicated that the parameter ratings used in the RMR classification system can predict the rock mass deformation modulus with a satisfactory correlation. However, the parameters don’t have the same influence on the rock mass deformability with the joint condition and the groundwater as the major and minor influencing parameters, respectively.     

裂隙岩体质量三维精细分级S-RMRmbi方法及可视化

传统岩体质量分级方法级别较少,其评分标准常呈阶梯状,应用于裂隙岩体质量评价时,其精度无法满足岩体质量出现复杂多变情况时的评级要求,使得同级别岩体质量可能出现显著差异,因此有必要将岩体质量级别做进一步精细分级处理,并为评价岩体工程稳定性和制定支护方案提供依据。基于前期开展的岩体质量RMR_mbi分级方法和结构均质区三维分显研究工作,以广西铜坑矿锌多金属矿体3^#矿块为试验区,进行RMR_mbi质量评级;结合综合模糊判别理论,提出岩体质量精细分级S-RMR_mbi方法,对试验区分级结果做进一步细化;并以均质区三维分显模型为载体,借助3DMine可视化功能,构建岩体质量精细分级可视化模型;从定性以及定量两方面对RMR_mbi和S-RMR_mbi法进行对比分析后,进一步提出与精细分级后的岩体相对应的支护方案。研究结果表明:(1)质量5级分档对凸显裂隙岩体质量层次性效果不够理想,基于综合模糊判别理论的岩体质量精细分级S-RMR_mbi方法有效细化了岩体质量分级结果;(2)赋有S-RMR_mbi分级属性值的各均质区及重组模型展示了岩体质量的空间分布特征;(3)S-RMR_mbi和RMR_mbi两分级结果模型显示其岩体质量分布状况大体一致,基于相似度的两方法λ12值为0.999 2,验证了S-RMR_mbi分级方法的合理性;(4)精细化分级后的岩体可采取更为细致、有效的支护手段。研究成果对于提升裂隙岩体质量评价精度与制定具体施工方案具有重要的指导作用。     

Strength classification of rock material based on textural properties

Rock, as a construction material, has great importance during the construction and service phases in a rock environment. The classification of rock materials based on their strength behavior provides a simple and fast solution to determine the type and application of support system as well as the method for opening underground structures. Intact rock materials are generally classified with regard to the strength, such as uniaxial compressive and point load strength. Rock texture, which consists of grains and matrix, directly affects the strength. The relation between the textural and mechanical properties of rock materials has been investigated, and rock texture was quantified from the texture coefficient (TC). The coefficient can be used to put a number on rock textures with experimental studies carried out on thin sections of rock material using image analysis. The main scope of this research is to classify the rock material according to its TC values based on the binary and fuzzy domain. In this study, TC is divided into five classes from very low to very high, and a fuzzy model is proposed to predict the uniaxial compressive strength from TC. A dataset is prepared to construct an objective study with 12 litho-type rock materials from 19 locations in Turkey. The binary and fuzzy classification as well as fuzzy model for the prediction of compressive strength is also applied to the dataset to illustrate the use of the proposed classification and model for underground construction in rock engineering. The model is applied to determine the intact rock material’s rating in rock mass rating classification (RMR) from the proposed classification as well as from the fuzzy model. The results of the example encourage the application of the proposed methods, especially for pre-feasibility studies of rock engineering projects.     

Safe rapid drifting – Support selection

Current drift advance rates in mining fall short of expectations with advances in drilling and blasting technologies. Quick access to orebodies improves their net present value (NPV). This is critical for block cave mining where several kilometers of drift network is initially required at high capital cost. Many mining companies are now planning block caving because of its long-term low production cost. This paper critically reviews the developments in tunnelling and mine drift development rates with advances in drilling and explosives technologies. Current drift support practice during development is also critically reviewed together with the rock mass classification systems. These reviews show that, while drilling and explosive technologies have drastically improved since 1850, current drift advance rates in the Canadian metalliferrous mining industry have either remained stagnant or dropped below the1960 advance rate levels and in comparison to advance rates in civil tunnelling. It is also established that a major cause for this stagnation is the use of long-term support in good ground conditions where only temporary support is required near-face for worker safety in the short-term. Long-term support takes up 46% of development cycle time. The paper presents a methodology for drift support design for underground hard rock conditions typically found in current mining practice in the Canadian Shield and discusses the rationale for optimizing ground support systems installed near-face during drift development. An improved Q-system called Q-star (Q^*) that accounts for discontinuous joints (rock bridges) and roughness of short joints in the rock mass is developed to more reliably estimate the self-support capacities of rock masses. It is recommended that construction damage be accounted for in the rock mass rating for safe support selection during development. A procedure is developed for the adjustment of Q^* to account for construction damage. Perimeter blasting is recommended as pre-requisite for rapid drift development in order to minimize construction damage, reduce support demand and scaling and mucking times. A support matrix is presented based on rock mass quality and stress level for safe rapid drift development. Two case histories in active mines are presented to validate the procedure. The methodology is applicable to stress-induced damage and may not apply where complete relaxation occurs. While the procedures presented are focused on typical conditions in the Canadian Shield underground mines, they may be applicable in civil engineering tunneling and other underground mines where drill-and-blast is used as the excavation method.     

Some new Q-value correlations to assist in site characterisation and tunnel design

The rock mass quality Q-value was originally developed to assist in the empirical design of tunnel and cavern reinforcement and support, but it has been used for several other tasks in rock engineering in recent years. This paper explores the application of Q and its six component parameters, for prediction, correlation and extrapolation of site investigation data, and for obtaining first estimates of some input data for both jointed distinct element and continuum-approximation modelling. Parameters explored here include P-wave velocity, static modulus of deformation, support pressure, tunnel deformation, Lugeon-value, and the possible cohesive and frictional strength of rock masses, undisturbed, or as affected by underground excavation. The effect of depth or stress level, and anisotropic strength, structure and stress are each addressed, and practical solutions suggested. The paper concludes with an evaluation of the potential improvements in rock mass properties and reduced support needs that can be expected from state-of-the-art pre-injection with fine, cementicious multi-grouts, based on measurements of permeability tensor principal value rotations and reductions, caused by grout penetration of the least favourable joint sets. Several slightly improved Q-parameter ratings form the basis of the predicted improvements in general rock mass properties that can be achieved by pre-grouting.     

Fuzzy approach for preliminary design of weak rock slopes in lignite mines

Rock slope failure modes are either driven by structurally controlled mechanisms (planar, wedge, or toppling failure) or lean upon the highly jointed or weak rock mass (circular failure). Developing practical tools for preliminary slope design is a popular topic among geotechnical society. This study proposes a practical methodology to predict a safe overall slope angle for weak rock slopes of lignite mines. Rock mass rating (RMR) and slope mass rating (SMR) classification systems can be remarkably misleading for weak rocks. Fuzzy modification was proven to improve rock and slope quality predictions. In addition, structurally controlled failures can be estimated more precisely. Later, a popular slope performance chart of Bieniawski that is based on RMR was modified by the fuzzy approach, and the computer models of a weak rock mass failure in a Turkish lignite mine are presented. The modified methodology was proven to be better suited to the weak rock conditions.     

富溪双连拱隧道围岩强度及稳定性评价

围岩强度指标是各种隧道工程设计中不可缺少的重要参数,准确预测围岩强度对隧道支护设计和稳定计算十分重要。针对富溪隧道地质条件复杂、围岩稳定性差等问题以现场岩石点荷载试验推测岩石强度,以隧道围岩的Q指标统计为基础,利用基于Q指标的经验公式推测围岩抗压强度,给出了富溪隧道各个断面围岩的抗压强度,据此评价各段围岩的稳定性。研究表明,采用的3个经验公式中,无论平均值,还是分散度,Yudhbir公式预测结果都最接近规范标准值,说明本文所用方法用于隧道工程是可行的。该方法简单、实用,预测结果能够反映围岩工程特性,符合实际,可以作为隧道工程中定量评价围岩强度的有效方法,为隧道设计、施工提供可靠依据。     

动力影响下山区露天矿合理边坡角的确定

针对山区露天开挖和岩石边坡工程实际，根据极限平衡理论和模糊数学理论，建立了相应的理论分析模型。考虑矿山生产爆破振动动力因素作用和影响，利用所建立的数学模型对某大型金属露天矿山边坡稳定性进行了具体计算分析，并依此对矿山边坡总体边坡角进行了评价。将理论计算分析结果与露天矿山工程实际情况进行了对比，表明所采用的理论模型和计算分析方法符合矿山工程实际。     

Fuzzy model for the prediction of unconfined compressive strength of rock samples

A data driven approach to the modeling of unconfined compressive strength of rock samples is presented. Fuzzy logic approach is used to represent a nonlinear relationship as a smooth concatenation of local linear submodels. The partitioning of the input space into fuzzy regions, represented by the individual rules, is obtained through fuzzy clustering. The numerical results are compared with a conventional statistical (multi-linear) model. It is shown that the fuzzy model is not only more accurate but as opposed to other black-box approaches (such as neural networks), it also provides some insight into the nonlinear relationship represented by the model.