An Intelligent Advanced Classification Method for Tunnel-Surrounding Rock Mass Based on the Particle Swarm Optimization Least Squares Support Vector Machine

Round 1

Reviewer 1 Report

The manuscript “Intelligent advance classification method for tunnel surround- 2 ing rock mass based on particle swarm optimization least 3 squares support vector machine” describes a method for classifying rock mass based on particle swarm optimization(PSO)-least squares support vector machine(LSSVM). The research idea is interesting and innovative.

 

Some correction to do:

Lines 35 and 44: et al.. changed with etc.

 

Line 66: what does it means “t” ?

 

Line 68: body the feature, It then… maybe you have to add . not ,

 

Figure 2. enlarge the figure to the whole page.

 

The title of the paragraph 3.1 “Digital rig” is the same of the sub-paragraph 3.1.1 “Digital rig”. Change one of the two or add more information in one of the two. Or:

3.1. Digital rig

RPD-150C is the primary model of the mine research, using a diesel engine as the power unit; the rig realizes a wide range of operations through 5 hydraulic cylinders, 150 m drilling and coring, and the maximum diameter is 225 mm[20-23].

Induction system. They are installed on a drill or rig to obtain drilling parameters for the rig, including position sensor, rotation sensor, and pressure sensor flow sensor.

Data conversion system. The system… and so on

 

Figure 3. The words in the figure are in Chinese? Change in English.

 

Figure 3. Which software do you use to obtain figure 3? Add more information in the text and into the caption of the figure.

 

Line 287: algebra(parameter… add space

 

Conclusion: delete the number 1, 2 and 3.

 

References. The references are not closed to recent years. Change with references more recently.

 

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

acceptable in this form

Author Response

Thank you very much for your approval of this manuscript, and we will continue to work on it in our future research efforts.

Reviewer 3 Report

General comments

Authors put an effort to prove that numerical drilling results can be used for rock mass characterization and classification.

The rock mass classification is very complex. There are many factors affecting the tunnelling.

After reading the manuscript, some questions arise:

- what is the rational for the selection of the drilling parameters (speed, torque, speed, propulsion, striking energy, …)? Other parameters can be used….

- are the results comparable to the others classification methods (Q classification method, MRM classification method, SRC classification method, …)?

- can this method be used in all geotechnical/geological settings or is more proper to massive and homogeneous rock masses? The geological setting is unknown.

 

I suggest applying this method together with the traditional rock mass classifications. Only this way can their applicability be proved.

 

The language/organization needs improvements.

 

Some figures are not readable.

Author Response

请参阅附件。

Author Response File: Author Response.docx

Reviewer 4 Report

Fast and accurate classification of surrounding rock mass is the basis for tunnel design and construction and has significant engineering application value. This paper proposes  a method for classifying and predicting surrounding rock mass based on particle swarm optimization(PSO)-least squares support vector machine(LSSVM). The research idea is that based on digital drilling technology, the acquired data is divided into a training group and a test group, and the training group continuously optimizes the algorithm of particle swarm optimization least squares support vector machine and then uses the test group to check. Fast searching ability of the particle swarm could significantly accelerate the computational power and computational accuracy of the least squares support vector machine, making it a high-speed analog search capability. Taking Jiaozhou Bay undersea tunnel in China as an example, a comparison of evaluation results of PSO-LSSVM and QGA-RBF(quantum genetic algorithm-radical basis function neural network) is made, and the results show that PSO-LSSVM, which matches well with field measured surrounding rock grade. The engineering application proves that the method has good self learning ability even when the sample size is small and the prediction accuracy is high, which meets the engineering requirements.

Conclusions

 In this paper, based on the classification problem of surrounding rock mass classification, a method of pre-classification of surrounding rock mass based on particle swarm least squares support vector machine is proposed. Based on the results of digital drilling detection, this method extracts useful information, uses the particle swarm optimization least squares support vector machine parameters, establishes the prediction model, and classifies the surrounding rock mass categories in front of the face. The accuracy of the particle swarm least squares support vector machine is improved with the increase in the number of learning samples. The particle swarm can quickly find the characteristics, overcome the blindness and randomness of artificial selection, and has high precision

The article is very interesting in its subject matter and has been prepared in a correct manner. With the above statements in mind, I propose and request that the article be published in its current form.

Author Response

Thank you very much for your approval of this manuscript, and we will continue to work on it in our future research efforts.

Author Response File: Author Response.docx

Round 2

Reviewer 3 Report

General comments

The description of the geological setting is messy. More careful is mandatory in a scientific paper.

 

The main objection remains: is this method better than the traditional ones?

I suggest applying this method together with the traditional rock mass classifications. Only this way can their applicability be proved.

 

The language/organization needs improvements.

 

 

Some figures are not readable.

 

Author Response

The authors have carefully gone through the reviewers' comments, these comments are very valuable and helpful for the improvement of the manuscript's quality. We have made the following responses. The dedicated works by reviewers are deeply appreciated.

①The reviewer’s comments: The description of the geological setting is messy. More careful is mandatory in a scientific paper.

The authors’ Answer:

Thanks to the reviewers' comments, we have revised the geological description of the construction site in a more critical way. Please check out lines 253-282 on page 7-8 of the revised article.

 

General description of the geological setting:

The tunnel site area of the Jiaozhou Bay Cross Harbour Tunnel is stable, with the development of small and medium fracture structures, mainly hard rocks, with relatively intact rock masses, small permeability coefficients, more developed adverse geological effects and generally good rock conditions. The basement rocks in the site area are mainly volcanic rocks of the Lower Cretaceous Qingshan Group and Late Yanshan Laoshan Super Unit intrusive rocks, which are hard rocks with high weathering resistance. The geological structure of the site is mainly fracture structure, in which the marine section crosses 4 groups and 14 fracture zones, along which there are intrusions of rock veins, with crushed rocks, fractured rocks and vesicular rocks in the fracture zone.

 

Description of the geological setting of the test area:

FK4+375.5 The palm wall is dominated by tuff with rhyolite. The rock is slightly weathered and relatively intact, with a predominantly blocky structure, locally with a blocky mosaic structure. The rock is relatively hard, with low permeability, and is a Class III enclosing rock.

FK4+375.5 to FK4+363.5, this section is basically the same as the current palm wall, with low strength and developed joints and fractures.

FK4+363.5~FK4+347.5, section FK4+363.5~356 has developed joints and fractures, the surrounding rock is fractured, and accompanied by strong weathering soft and weak fracture filling, the surrounding rock contains water, especially around FK4+353 develops penetrating joints.

FK4+347.5～FK4+334.5, this section of the surrounding rock is broken, the strength of the surrounding rock is poor, in the fault fracture zone, there is mud trapping at the fault, longitudinal fissure development is more.

FK4+334.5 to FK4+320.5, the section is slightly more complete than the previous section, the strength has increased, the integrity is slightly better and the rock body is tuff.

FK4+320.5～FK4+312.5, this section still belongs to the fault fracture zone, the surrounding rock is fractured, has low strength and poor integrity, especially at FK4+341 and FK4+345 there are more obvious fractures and contains fractured weak fillings.

 

 

②The reviewer’s comments: Is this method better than the traditional ones?

The authors’ Answer:

We would like to inform the reviewers that the method proposed in this paper has advantages over the traditional method.

Firstly, in terms of the results of surrounding rock classification, the method proposed in this paper remains largely consistent with the traditional method (Table 1). Secondly, in terms of applicable conditions, the Q classification method is applicable in both soft and hard rock masses, and the advantages of the Q classification method are more obvious in dealing with extremely soft rock formations; the RMR method is more applicable to hard rocks affected by joints and is not suitable for use in soft rock formations; the BQ classification method is applicable to all kinds of rocks except for expansive enclosing rocks, and has a wider range of use; the above traditional methods, among the rock formations to which they are applicable, the values derived from the above traditional methods have a high degree of confidence in the rock formations to which they are applicable.

所提出的基于钻探参数的围岩预测分类方法适用于硬质岩石的均质和非均质岩体，但适用于所有软岩效果较差。 在硬岩层中，对钻井参数的感知更为显著，其实验结果精度较高，而在软岩层中其对钻井参数的感知减弱，从而影响实验结果。最后，在试验参数获取方面，该方法检验的过程（推导围岩分类结果的过程）比传统方法简单，需要大量的计算参数，下面我们列出了各种方法的计算方程，以及该方法所需的计算参数， 这需要在现场花费大量时间才能获得;本文提出的方法所需的参数只需要从数字钻机系统中导出钻探参数，然后就可以进行岩石分类。

结合这三点，我们得出结论，本文提出的方法比传统方法具有优势。

 

传统方法Q法计算公式。[1]

Q法的参数为：岩石质量指数RQD，节理组数J _n，节理粗糙度系数J_r，节理蚀变影响因子J _a，节理水分折现因子J _w，应力折现因子SRF。

 

传统方法RMR法计算公式。[2]

RMR方法要获得的参数：岩石强度为₁，RQD值为₂，节理间距为₃，不连续结构面特征为4，地下水条件为₅，接头方向校正因子为B。

 

传统方法BQ法计算公式。[3]

BQ方法得到的参数：R_c表示岩石的单轴饱和抗压强度，K_v表示岩石完整性指数。

R_c：;I_s（50）的计算公式表示实测岩石点荷载强度指数。

K v：;_{v pm}的计算公式表示岩体的弹性纵波速度;V_pr表示岩石的弹性纵波速度。

 

表1 PSO-LSSVM与传统方法相比的预测结果

序号	PSO-LSSVM	Q 法	RMR 方法	BQ 方法
1	四.4（3.98）	四.4 （0.28）	四.4（33.16）	IV.4（335.33）
2	三.3（3.02）	三.3 （5.31）	三.3 （59.01）	III.3（359.39）
3	三.3（3.26）	三.3 （4.95）	三.3 （58.3）	III.3（360.26）
4	四.4（3.98）	四.4 （0.46）	四.4（37.76）	IV.4（330.89）
5	四.4（3.98）	四.4 （0.73）	三.3 （40.79）	IV.4（333.95）

 

 

1. 北卡罗来纳州巴顿;连，R.;Lunde， J. 隧道支护设计的岩体工程分类.岩石力学 1974， 6， 189–236.
2. 比尼亚夫斯基，Z. 工程岩体分类;纽约：威利-跨科学出版社：1989年。
3. 水利部中华人民共和国.岩体工程分级标准GB50218-94。1995年。

Author Response File: Author Response.docx

Round 3

Reviewer 3 Report

The comments in the reply should be incorporated into document. Please, improve the manuscript.

Author Response

①The comments in the reply should be incorporated into document. Please, improve the manuscript.

The authors’ Answer:

Thank you for your comments， we have incorporated the comments in the reply into the manuscript. Please check out lines 383-427 on page 12-13 of the revised article.

The method proposed in this paper is compared with traditional methods (Q method, RMR method, BQ method)：

Firstly, in terms of the results of surrounding rock classification, the method proposed in this paper remains largely consistent with the traditional method (Table 3). Secondly, in terms of applicable conditions, the Q classification method is applicable in both soft and hard rock masses, and the advantages of the Q classification method are more obvious in dealing with extremely soft rock formations; the RMR method is more applicable to hard rocks affected by joints and is not suitable for use in soft rock formations; the BQ classification method is applicable to all kinds of rocks except for expansive enclosing rocks, and has a wider range of use; the above traditional methods, among the rock formations to which they are applicable, the values derived from the above traditional methods have a high degree of confidence in the rock formations to which they are applicable. The proposed method of predictive classification of surrounding rocks based on drilling parameters is applicable in both homogeneous and non-homogeneous rock masses of hard rocks, but is less effective in all soft rocks. In hard rock formations, the perception of drilling parameters is more significant and its experimental results are highly accurate, while its perception of drilling parameters is weakened in soft rock formations, which can affect the experimental results. Finally, in terms of test parameter acquisition, the process of this method test (the process of deriving the results of surrounding rock classification) is simpler than the traditional method, which requires a large number of calculation parameters, and we list below the calculation equations for the various methods, as well as the calculation parameters required for this method, which require a lot of time in the field to obtain; the parameters required for the method proposed in this paper only require the drilling parameters to be derived from the digital drilling rig system, and then the rock classification can be carried out.

Combining these three points we conclude that the method proposed in this paper has advantages over traditional methods.

Traditional method Q method Calculation formula[37]:

(9)

Parameters to be obtained for the Q method: RQD for rock quality index, J_n for number of joint groups, J_r for joint roughness coefficient, J_a for joint alteration impact factor, J_w for joint water discount factor and SRF for stress discount factor.

Traditional method RMR method Calculation formula[38]:

(10)

Parameters to be obtained for the RMR method: A₁ for rock strength, A₂ for RQD values, A₃ for joint spacing, A₄ for discontinuous structural face characteristics, A₅ for groundwater conditions and B for the joint orientation correction factor.

Traditional method BQ method Calculation formula[39]:

(11)

Parameters to be obtained for the BQ method: R_c represents the uniaxial saturation compressive strength of the rock and K_v represents the rock integrity index.

Formula for calculating R_c：；I_s（50） represents the measured rock point load strength index.

Formula for calculating K_v：；V_pm represents the elastic longitudinal wave velocity of the rock mass; V_pr represents the elastic longitudinal wave velocity of the rock.

Table 3 Prediction results of PSO-LSSVM compared with traditional methods

Serial number	PSO-LSSVM	Q method	RMR method	BQ method
1	Ⅳ4(3.98)	Ⅳ4 (0.28)	Ⅳ4 (33.16)	Ⅳ4(335.33)
2	Ⅲ3(3.02)	Ⅲ3 (5.31)	Ⅲ3 (59.01)	Ⅲ3(359.39)
3	Ⅲ3(3.26)	Ⅲ3 (4.95)	Ⅲ3 (58.3)	Ⅲ3(360.26)
4	Ⅳ4(3.98)	Ⅳ4 (0.46)	Ⅳ4 (37.76)	Ⅳ4(330.89)
5	Ⅳ4(3.98)	Ⅳ4 (0.73)	Ⅲ3 (40.79)	Ⅳ4(333.95)

 

37. Barton, N.; Lien, R.; Lunde, J. Engineering classification of rock masses for the design of tunnel support. Rock Mechanics 1974, 6, 189–236.
38. Bieniawski, Z. Engineering Rock Mass Classifications; New York: The Wiley-Interscience Publication: 1989.
39. Ministry of Water Resources of the People’s Republic of China.Standard for Engineering Classification of Rock Masses GB50218-94. 1995.

 

Author Response File: Author Response.docx

Round 4

Reviewer 3 Report

Please, improve the Fig 4. The lettering is too small.

 

Author Response

1. Response to Reviewer

Title: An intelligent advanced classification method for tunnel-surrounding rock mass based on the particle swarm optimization least squares support vector machine

Authors: Jie Lu, Weidong Guo, Jinpei Liu, Ruijie Zhao, Yueyang Ding and Shaoshuai Shi

Manuscript ID: applsci-2133723

 

2. Cover Letter

Dear reviewers,

 

Thank you for allowing us to submit a revised draft of the manuscript “An intelligent advanced classification method for tunnel-surrounding rock mass based on the particle swarm optimization least squares support vector machine” for publication in the Applied Sciences. We appreciate the time and effort that you and the reviewers dedicated to providing feedback on our manuscript and are grateful for the insightful comments on and valuable improvements to our paper.

 

We have incorporated the suggestions made by the reviewers. Those changes are highlighted within the manuscript, for a point-by-point response to the reviewers’ comments and concerns. Here below is our description of the revision according to the reviewers’ comments.

 

Sincerely

 

Weidong Guo

 

3. Response to the reviewers’ comments

The authors have carefully gone through the reviewers' comments, these comments are very valuable and helpful for the improvement of the manuscript's quality. We have made the following responses. The dedicated works by reviewers are deeply appreciated.

①The reviewer’s comments: Please, improve the Fig 4. The lettering is too small.

The authors’ Answer:

Thank you for your comments, we have enlarged the lettering in Figure 4.

Figure 4. Digital drilling rig parameter acquisition interface.

Author Response File: Author Response.docx