LightGBM integration with modified data balancing and whale optimization algorithm for rock mass classification

A modified rock mass classification system for TBM tunnels and tunneling based on the HC method of China

During the construction of tunnel, the observational classification of rock mass quality is the most direct method to evaluate the quality of the tunnel face surrounding rock. It is also an important basis for preventing geological disasters in tunnel construction, deciding construction excavation method and supporting measures. The traditional Q method and BQ method of national standard rock mass quality grading evaluation method require on-site and indoor laboratory tests and analysis, which results in the lag time of rock mass quality evaluation, often reduces the construction efficiency, or misses the window time to prevent sudden construction geological disasters. Rapidly and accurately to rock mass quality classification of tunnel face is the primary problem to be solved in observational rock mass classification of highway tunnel during construction period. The artificial intelligence algorithm provides a method to solve the real-time and accurate evaluation of the rock mass of the tunnel. Taking Yanqing-Chongli highway for Beijing Winter Olympic Games as an example, we adopt a method of using photogrammetry and artificial intelligence rock structure parameter to identify working face. In this method, seven index parameters system are established. We use the KNN intelligent algorithm to evaluate the quality of the tunnel rock and select 150 samples from 40 working faces in 8 tunnels for the training and learning. Another 50 samples are selected for rock mass quality evaluation and verification. Compared with the rock mass quality evaluation results of BQ, the accuracy reachs 90%. Conclusions are as follows. (1)KNN method for observational classification of rock mass quality of highway tunnel is a fast and efficient method for observational classification of rock mass quality by using artificial intelligence technology, which can obtain real-time rock mass evaluation results on site. (2)KNN classification method selects seven discriminant indicators, considering the occurrence of rock environment, geological structure, rock structure and other characteristics, and reflecting the operability and applicability of these indicators in the actual engineering. (3)KNN classification method misjudgment rate is very low. It has strong discriminant ability when marking discriminant classification by excluding the influence of artificial factors. This method provides a new way to solve the rock mass grade determination and classification.

Geotechnical and geological parameters have the greatest impact on the performance of hard rock tunnel boring machines (TBMs). This includes the rock and rock mass properties that affect the rate of penetration (ROP) as well as the machine utilization that is heavily dependent on ground support type and related machine downtime and delays. However, despite the widespread use of TBMs and established track records, accurate estimation of machine performance is still a challenge, especially in complex geological conditions. The past studies have tried to use rock mass classification systems for improving the accuracy of the machine performance prediction. Rock mass classifications has been primarily developed for design of ground support, and as such, have not offered a good fit for estimation of TBM performance. This paper will review performance of a hard rock TBM in a 12.24 km long tunnel and offers analysis of field performance data to evaluate the relationship between various lithological units and TBM operation. The results of statistical analysis of the initial 5.83 km long tunnel indicate strong relationships between geomechanical parameters and TBM performance parameters. Site specific models, including Non-linear regression analysis (NLRA), Classification and regression tree (CART), and Genetic Programming (GP) have been used for analysis of a TBM performance relative to the ground condition data. The current study has looked at the possibility of developing a new rock mass classification system for TBM application by using the above noted analysis. Preliminary results indicate that CART can be used for offering a proper rating scheme for a rock mass classification system that can be used for TBM applications.

Real-time perception of rock–machine interaction information in TBM tunnelling using muck image analysis

Diameter-independent indices for TBM rock-mass classification

Tunnel face and wall collapse are common during excavations performed by tunnel boring machines (TBMs) due to the difficulty of correctly identifying the properties of the excavated rock. This identification, however, can be simplified by using the cutting force to estimate rock strength, a method that has already proved quite successful in Japanese tunnel excavations. This paper summarizes knowledge relating to the cutting force obtained through tunnel excavation experience, and the relationship between rock strength and TBM operation is discussed. Although TBM operators rely on intuition to set the cutter head speed appropriately, this decision process represents a logical method of operation that takes advantage of the variable speed capability of the cutter head. Selection of appropriate support methods for the excavated face is also a critical issue in tunnel excavation. This selection process is based on the condition of the rock, which is difficult to determine quickly and accurately during tunnel excavation. The present paper uses the excavation of two tunnels to demonstrate that it is possible to assign rock mass classifications accurately based on rock strength when boring a uniform rock type. It is also shown that the rock mass can be classified from the rock strength normalized by the uniaxial compressive strength when boring through mixed rock types.

The article presents a new design of a vibrating twin-shaft centrifugal module designed for grinding and classification of rock mass. In this design, in addition to grinding, the operation of classification or screening of the crushed mass was added, which does not allow its regrinding and increases the productivity of the device. This is achieved by installing in the bottom of the chamber grinding mesh with cells in accordance with the required class size. At the same time, the classification process is intensified by the presence of vibration from vibration exciters fixed on the camera body and the installation of the camera on elastic supports. The reciprocating horizontal vibrations of the chamber with a given amplitude and frequency contribute to the segregation of the crushed rock mass in the bed by size, which positively affects the efficiency of classification and grinding. The presence of vibration helps to unload the oversize product from the grinding chamber. Also, the article considers experimental studies performed on a vibrational two-shaft centrifugal module to determine the dependence of the performance of a given design on five variable factors: rotor shaft revolutions (n, rpm), size of the loaded rock mass (Δ, mm), rock mass strength (σ, kg/mm2), camera vibration frequency (ω, rpm) and its vibration amplitude (A, mm). Studies have shown the efficiency and increased productivity of the new design in relation to a centrifugal disintegrator without a classification grid and vibration. The results of the work allow us to recommend the design under study for the manufacture of an experimental sample according to the given initial requirements, and the established dependences (Q = f (n, Δ, σ, ω, A) make it possible to develop a mathematical model of the grinding process in this setup to calculate the required parameters.

Pre-tunneling exploration for rock mass classification is a common practice in tunneling projects. This study proposes a data-driven approach that allows for rock mass classification. Two machine learning (ML) classification models, namely random forest (RF) and extremely randomized tree (ERT), are employed to classify the rock mass conditions encountered in the Pahang-Selangor Raw Water Tunnel in Malaysia using tunnel boring machine (TBM) operating parameters. Due to imbalance of rock classes distribution, an oversampling technique was used to obtain a balanced training dataset for unbiased learning of the ML models. A five-fold cross-validation approach was used to tune the model hyperparameters and validation-set approach was used for the model evaluation. ERT achieved an overall accuracy of 95%, while RF achieved 94% accuracy, in rightly classifying rock mass conditions. The result shows that the proposed approach has the potential to identify and correctly classify ground conditions of a TBM, which allows for early problem detection and on-the-fly support system selection based on the identified ground condition. This study, which is part of an ongoing effort towards developing reliable models that could be incorporated into TBMs, shows the potential of data-driven approaches for on-the-fly classification of ground conditions ahead of a TBM and could allow for the early detection of potential construction problems.

Numerical modeling is a fine tool for the assessment of inaccessible zones or unforeseen conditions in geotechnical engineering. When modeling structures in rocks, the laboratory-determined properties must be factored as per rock mass classification. The joints present in rock mass can be utilized for assessment of wedges forming around the tunnels. Half tunnels or overhanging tunnels are constructed in steep rock slopes through slot cutting. This develops overhanging cliff in hard rocks that looked like a half tunnel. Half tunnels have advantage over conventional tunnels or open excavations as they require less excavation. The current construction practice of half tunnels is solely based on experience of construction in hard rock. This study discusses the stability of half tunnels and analyzes its design in the Himalayas. A road extension case study for existing 7-m span tunnel is modeled and studied. Finite element code is used to assess the stability of half tunnels by analyzing different possible scenarios. In addition, rock mass classification is performed along with wedge analysis, and analytical designs. It was found that massive rock mass is quite suitable for the construction of half tunnels. Also, most economical and stable half tunnel designs were recommended.

Poboya Mountains is one of the gold mines located in Palu city. Miners conduct quarrying and slope slicing to extract gold ores every day, resulting in changes on slope stability. The research is to determine rock mass classification and slope stability of metamorphic rocks. The research conducted at Poboya Gold Mine, it geographically located at central of Sulawesi province. The classification of rock mass carried out through a field survey and a laboratory testing, and then determined according to methods of RMR and SMR. The results showed that for gneiss at location 1, it obtained values of 60-69 for RMR and SMR as well. Meanwhile, for schist at location 2, the obtained values are 56-71 for RMR and 57-79 for SMR. Schist at location 3 has values of 65-75 for RMR and 69-79 for SMR, and schist at location 4 have values of 56-70 for RMR and 59-85 for SMR. Referring to the rock mass classification results, it can be concluded that slopes location at Poboya with a good quality of rock mass is at schist of location 3. Generally, slopes considered stable, hence there will be no landslides, and if landslide occurs, it will merely occurs on small blocks.

Rock slopes stability is important for personnel and equipment safety in the open-pit mine. Instability and failure of slope occur due to many factors such as unsuited slope geometry, geological discontinuities, weak slope material due to weather influences. External loads such as high rainfall and seismicity could play an essential role in slope failure. Consequently, a precise classification of rock mass is needed for the basis of determining technical policy. Rock slopes in open pit coal mining areas, especially in Indonesia, are characterized by applying various rock mass classification systems, such as Rock Mass Rating (RMR) and Geological Strength Index (GSI), because the study area comprises well-exposed rock formations. In the RMR system, there are five main parameters i.e. Rock Quality Designation (RQD), Uniaxial Compressive Strength (UCS) of rocks, discontinuity spacing, discontinuity and groundwater conditions were considered. Several rock mass classification systems developed for the assessment of rock slope stability were evaluated with the condition of rock slopes in the tropics, especially Indonesian region, particularly in sedimentary rocks in the open pit coal mining area in order to get the corrected GSI equation used to characterize rock slopes based on rock mass structure quantitative analysis and discontinuities surface conditions. This paper provides correlation between the GSI and RMR for sediment rock in coal mines.

Quantitative granitic weathering assessment for rock mass classification optimization of tunnel face using image analysis technique

Setting up simple estimating equations of TBM penetration rate using rock mass classification parameters

Applying rock mass classifications to carbonate rocks for engineering purposes with a new approach using the rock engineering system

Among all of the subsystems and components of hydropower, the tunnel is one of the crucial structures that face most of the pre-operation, real-time and post-operation challenges. Tunnels used as a water conveyance system in the majority of the projects lead to significant importance in the hydropower sector. The engineering rock mass classification is a crucial step in the accomplishment of underground constructions, particularly tunnels and caverns within the rock mass. The quality of rock mass parameters must be assessed and predicted with greater accuracy. However, the difficulties lie in the proper assessment of these parameters to define their quality. Studies have shown that despite the use of numerical and empirical methods in construction practices, numerous problems are reported when it comes to dealing with such structures. Over the years, various approaches based on machine learning (ML) techniques have been used for the minimization of these difficulties. As in recent trends, the development of machine learning technologies is currently bringing new methodologies in assessing rock mass classification. Although, because a sufficiently large database is required for such initiatives, there are a number of difficulties in their practical execution. Thus, a review of studies on the use of machine learning techniques in rock mass classification is presented in this paper. To determine the future directions of ML in RMC, it presents a comprehensive assessment to examine how ML models are being used. This study offers recommendations for creating an industry-specific model focusing on the strengths and weaknesses of ML techniques