Risk assessment for tunnel portals in the construction stage has been widely recognized as one of the most critical phases in tunnel construction as it easily causes accident than the overall length of a tunnel. However, the risk in tunnel portal construction is complicated and uncertain which has made such a neural network very attractive to the construction projects. This paper presents a risk evaluation model, which is obtained from historical data of 50 tunnels, by combining the fuzzy method and BP neural network. The proposed model is used for the risk assessment of the Tiefodian tunnel. The results show that the risk evaluation level is IV, slope instability is the greatest impact index among four risk events, and the major risk factors are confirmed. According to the evaluation results, corresponding risk control measures are suggested and implemented. Finally, numerical simulation is carried out before and after the implementation of risk measures, respectively. The rationality of the proposed risk evaluation model is proved by comparing the numerical simulation results.
Over the past few decades, construction of highways has been developing quickly in China. Tunnel construction has become the first choice for highway alignment because of its advantages of optimal alignment, reduced travelling time, and enhanced operation efficiency. Meanwhile, tunneling is a dangerous occupation owing to its complicated construction technology, uncertainties risk factors, and complicated geological conditions. There are numerous casualties and millions of economic losses caused by tunnel accidents. Compared with the overall length of a tunnel, portals usually have a limited area of influence. The weathering extent of surrounding rock is heavier, the buried depth is shallow, and it is vulnerable to the impact of rainfall. In the entrance section, the construction is difficult, and these unfavorable factors easily lead to engineering accidents, such as slope instability, large deformation, tunnel collapse, and other accidents [
Risk assessment for highway tunnel portals during the construction stage has been widely recognized as one of the most critical phases. However, risk factors of tunnel portal construction are complex and uncertain. At the same time, most of the information in risk evaluation comes from the expert’s subjective judgement which is usually imprecise and subjective in the decision-making process. Handling vagueness and subjectivity becomes a primary task in risk assessment.
The fuzzy system describes a class of extrapolated blurring without explicit boundaries and establishes a correspondence between uncertainties and membership functions so that favorable mathematical tools can be used to analyze many inaccurate vague phenomena in nature. Fuzzy theory has found in-depth research and application in the mining, nuclear, petrochemical, and construction industries. van Laarhoven and Pedrycz introduced the concept of fuzzy set theory into the traditional analytic hierarchy process (AHP) and originally proposed the fuzzy analytic hierarchy process (FAHP) [
Artificial neural network (ANN) is also known as a neutral network which is a mathematical model for finding patterns among datasets where there are complex relationships between the inputs and outputs. ANN attempts to simulate the structure and operation of the human neural network system. Since ANN acts like a “black-box” and cannot explain the reasoning process, it can well achieve the self-adaptation through the learning function and can acquire the fuzzy data expression knowledge accurately and automatically. The ability to learn from examples has made this technique a very useful tool in data modeling [
ANN and fuzzy theory are complementary technologies. Fuzzy theory tries to describe and deal with the ambiguity concept in human language and thinking. The artificial neural network is based on the human brain’s physiological structure and information-processing process. With the rapid development of fuzzy system and artificial neural network research, it has been found that the original independent field can be compensated and fused together, which leads to a new field—fuzzy neural network (FNN). FNN provides effective tools for addressing uncertainties in decision-making [
For the purpose of handling the vagueness and subjectivity in risk evaluation of highway tunnel portal construction, this paper proposes a risk evaluation model by combining fuzzy theory with the neural network. There are many kinds of neural network types. This paper uses the BP network—a multilayer feedforward neural network—which can achieve any nonlinear mapping from the input to output. 80% to 90% of the neural network model uses the BP network or its change form. The feasibility and effectiveness of this model are proved by an engineering case.
The remainder of this research is organized as follows: Section
The BP neural network generally has three or more layers of neurons. There are input layer, hidden layer, and output layer, respectively. According to the Kolmogorov theorem, this model uses a three-layer BP neural network with a single hidden layer [
Structure of the BP neural network (
This topology is divided into four parts: the first part is the input layer and each of its nodes represents an input variable. The second part is the fuzzy process. The neural network input value, which is the input layer of the neural network structure, can be obtained by the fuzzification of the risk factor. The third part is the fuzzy reasoning layer, which is the hidden layer. It can complete the mapping between the input variable and the fuzzy value of the output variable. The fourth part is the output layer, which is the result of the risk level.
The evaluation of the index system is the basis and key of the risk assessment research. It directly affects the objectives, accuracy, and results of the evaluation. From risk evaluation theory of the highway tunnel construction, any risk factors associated with a highway tunnel project are likely to evolve into a risk event, leading to the occurrence of construction safety accidents. This paper collected the data of the 50 highway tunnel samples from the literature.
All these 50 tunnels are located in China. Among them, 12 tunnels are located in Fujian Province, 10 in Shanxi Province, 8 in Zhejiang Province, 6 in Shaanxi Province, 5 in Hebei Province, 3 in Shandong Province, 3 in Hubei Province, and 3 in Hunan Province. There are 39 two-lane and 11 three-lane tunnels among 50 tunnels. There are 15 tunnels with a length of 3000 m or more, 22 tunnels with a length of 1000–3000 m, and 13 tunnels with a length of less than 1000 m. The longest one is the Mayazi tunnel which is 9007 m. The Wulidun tunnel is the shortest one which is about 980 m. According to the safety status and the screening of risk factors, the risk assessment index system of the 50 highway tunnel portal construction was built, as shown in Figure
The hierarchical structure of risk evaluation in highway tunnel portal construction.
It is composed of three layers. On the top, the goal of this paper is to assess the risk level of highway tunnel portal construction. The first index is four risk events of the highway tunnel portal construction, which include tunnel entrance collapse (
It is necessary to quantify these risk indexes of the established index system because the BP neural network requires quantitative data in risk evaluation of construction. As a result, the fuzzy evaluation method was used to quantify the degree of risks in this research. Take one of the 50 tunnels as an example. Step 1: the set of the comments level
In this paper, the risk of highway tunnel portal construction was divided into five levels according to 50 sample tunnels, where level I, level II, level III, level IV, and level V represent less risk, low risk, general risk, high risk, and higher risk, respectively. The reviews set is expressed by Step 2: factors domain of the evaluation object
The number of evaluation objects is Step 3: membership matrix
Take Tunnel 1 which belongs to the 50 tunnels as an example.
Each risk factor holds a degree of membership to every risk level for different risk events, with the summation of all of its degrees of membership being 1. The membership matrices of the secondary indicators at each level are as follows: Step 4: the weight of the risk level
Each risk level has different impacts on the highway tunnel portal construction; therefore, the weight of each risk level in the judging set is different in the set of the comments level. Determine the standard risk probability level Step 5: calculate modeling input
The input data of the risk evaluation model can be obtained by combining
The input data of Tunnel 1 are calculated as shown in Table
The input data of Tunnel 1.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
0.136 | 0.200 | 0.344 | 0.360 | 0.328 | 0.256 | 0.344 | 0.360 | 0.256 | 0.280 | 0.352 | 0.360 | 0.232 | 0.360 | 0.168 | 0.224 | 0.336 | 0.280 | 0.280 | 0.208 | 0.328 | 0.360 | 0.360 | 0.360 | 0.328 | 0.296 | 0.304 | 0.288 |
Similar to the abovementioned calculation process, the input data of the other 49 highway tunnels can be obtained. The calculation process is omitted here due to the space constraints.
The expected output corresponding to this paper is the risk evaluation level of highway tunnel portal construction. In this section, FAHP is used to calculate the risk evaluation level of Tunnel 1 portal construction. The calculation process of the output data for Tunnel 1 will be described in detail as follows. Step 1: calculate the weight
AHP is employed to calculate the weight of the risk index of highway tunnel portal construction. According to the abovementioned hierarchical structure, individual judgements are collected and comparison matrices of the risk factor are constructed. Calculation of the comparison matrix requires consistency checking. The expression’s consistency ratio (
The average random consistency indicator.
Matrix order | 3 | 4 | 5 | 6 | 7 | 8 | 9 |
---|---|---|---|---|---|---|---|
RI | 0.58 | 0.90 | 1.12 | 1.24 | 1.32 | 1.41 | 1.46 |
Through expert surveys and associated risk theories, a judgement matrix for the construction stage of the cavern section was established for each of the identified seven risk factors. The comparison matrix
Weight vectors of the risk factor are calculated using the eigenvalue method which are as follows: Step 2: identify the membership degree matrix
The membership degree matrix Step 3: single risk factor fuzzy evaluation
Combining the weight of single risk factors with Step 4: multifactor fuzzy comprehensive evaluation
Taking
This research chooses the maximum membership degree principle to identify the risk level because it is generally applicable in the course of engineering risk assessment. The risk level of the construction stage in highway Tunnel 1 portal is most likely to be level I. Step 5: risk probability score
Each risk evaluation level is represented by a digit, which is converted to a corresponding probability score according to Table
The correspondence between the risk level and score.
Risk level | I | II | III | IV | V |
---|---|---|---|---|---|
Score | 10000 | 01000 | 00100 | 00010 | 00001 |
As a result, the risk score of Tunnel 1 is 10000.
Similarly, the probability of construction risk of the other 44 tunnel portals and the risk score of overall 45 tunnels can be obtained as shown in Table
The risk probability score of sample tunnels.
Tunnel | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 |
Score | 10000 | 00100 | 00100 | 01000 | 00100 | 00010 | 01000 | 01000 | 00100 |
|
|||||||||
Tunnel | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 |
Score | 01000 | 00010 | 00100 | 00010 | 01000 | 00100 | 00100 | 01000 | 00100 |
|
|||||||||
Tunnel | 19 | 20 | 21 | 22 | 23 | 24 | 25 | 26 | 27 |
Score | 00100 | 00100 | 00010 | 00100 | 00010 | 00100 | 00010 | 00100 | 00100 |
|
|||||||||
Tunnel | 28 | 29 | 30 | 31 | 32 | 33 | 34 | 35 | 36 |
Score | 00100 | 00100 | 01000 | 00010 | 01000 | 01000 | 00010 | 00010 | 00100 |
|
|||||||||
Tunnel | 37 | 38 | 39 | 40 | 41 | 42 | 43 | 44 | 45 |
Score | 00100 | 01000 | 00010 | 00010 | 00100 | 00100 | 10000 | 00100 | 00010 |
The BP neural network is a kind of a network with a teacher. The teachers are actually a training sample for training the network. The input and output mappings of the network are obtained from training samples. The sample input and the target output required by the network must be known, and the weight coefficient between each layer is determined by the input after the exact output value can be derived. In the BP neural network, the data are propagated backwards from the input layer through the hidden layer. If the output layer does not have the desired output value, the connection weight of the network is corrected from the output layer in the direction of reducing the error. The error will gradually decrease with the continuous learning of the network until the error is no longer down, and then the network training is completed.
With different learning rates, there is great influence on the performance of the established BP neural network model. The smaller the learning rate is, the slower the convergence rate is. If the learning rate is too large, the training is prone to oscillation. At present, we can only roughly determine the learning rate through experience for different issues, and the selection range is generally in the range [0.01, 0.8].
The performance of the BP neural network is also related to the number of hidden neurons. In general, the bigger the number of hidden neurons is, the better the network performance is. However, if the number of hidden neurons is too much, the training time may be too long. Currently, there is no ideal analytical formula for determining the number of hidden neurons. Generally, the following empirical formula is used to obtain the estimated value:
In the three-layer BP neural network, assuming that the number of input neurons is Step 1: determine the structure of the network and then initialize all of the network weights and the threshold neurons in the hidden layer and output layer. Step 2: the fuzzy sample dataset Step 3: calculate the output Step 4: calculate the value of each layer. The error of the output layer is expressed as Step 5: adjust the weight value by using Step 6: calculate the error by using
The flow diagram of the BP algorithm is shown in Figure
The flow diagram of the BP algorithm.
Forty-five samples data from the 50 tunnel samples data are selected, which have been obtained in Section
Comparison of network test results.
Tunnel | Test results of the fuzzy neural network | Model result | Cumulative error | Realistic risk level | ||||
---|---|---|---|---|---|---|---|---|
Output 1 | Output 2 | Output 3 | Output 4 | Output 5 | ||||
46 | 0.0000 | 0.0001 | 1.0000 | 0.0000 | 0.0000 | 0 0 1 0 0 | 0.01% | III |
47 | 0.0000 | 0.0032 | 0.9992 | 0.0000 | 0.0000 | 0 0 1 0 0 | 0.40% | III |
48 | 0.0000 | 0.0001 | 0.9497 | 0.0016 | 0.0000 | 0 0 1 0 0 | 5.20% | III |
49 | 0.0000 | 0.0003 | 0.9854 | 0.0001 | 0.0000 | 0 0 1 0 0 | 1.50% | III |
50 | 0.0000 | 0.9847 | 0.0149 | 0.0001 | 0.0015 | 0 1 0 0 0 | 3.18% | II |
By comparing the test results with the expected results, the difference between the two values is taken as the error, and the error of each object is obtained. The cumulative error of the five subjects is 2.06%, and the prediction accuracy of the network model is 97.94%, which indicates that the BP fuzzy neural network model has a high prediction accuracy.
The Tiefodian highway tunnel is connected to Baoji and Hanzhong in Shaanxi Province of China located in the south of Tiefodian Town, the west of the National Highway 316. The tunnel is composed of two separate tunnels, which belong to short tunnels. The left tunnel being 135 m long has a mileage pile number of ZK164 + 570–ZK164 + 705. The right tunnel is 185 m long and has a mileage pile number of YK164 + 550–YK164 + 735.
The tunnel is perpendicular to the slope surface with no bias, and the portal of this tunnel lies in the valley foot on the right bank of Bao River. The depth of the tunnel portal is lower than 36.0 m, which belongs to the shallow tunnel. The tunnel span is 16.0 m. The inclination angle of the stable natural slope is 41°. The lithology along the tunnel is completely weathered gneiss. During the construction of tunnel portal excavation, the collapse and the slope instability are easily caused. Therefore, there is a certain degree of difficulty in excavating the tunnel portal.
The distance between the left tunnel exit and right tunnel exit is 30 m. The surrounding rock type is completely weathered gneiss. The surrounding rock of the right tunnel portal and left tunnel portal is both of level V. Take the portal section of the left tunnel as the research object for risk assessment. The longitudinal profile of the left tunnel is shown in Figure
Longitudinal profile of the left tunnel.
Tunnel portals are constructed with the bench method, which is shown in Figure Step 1: in the arch of the tunnel, the advanced small pipe with grouting reinforcement strata is used. Step 2: the upper bench is excavated. Step 3: the initial support of the bolt, steel frame, and shotcrete is constructed at the upper bench. Step 4: the lower bench is excavated. Step 5: the initial support of the bolt, steel frame, and shotcrete is constructed at the lower bench. Step 6: the invert arch is excavated. Step 7: the concrete of the invert arch is constructed. The excavation distance of each part is 3 m.
Construction steps of the bench method.
The construction risk of the Tiefodian tunnel portal is evaluated according to the proposed model in Section
The input data of the Tiefodian tunnel.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
0.224 | 0.280 | 0.088 | 0.296 | 0.168 | 0.096 | 0.152 | 0.232 | 0.360 | 0.232 | 0.304 | 0.168 | 0.144 | 0.152 | 0.248 | 0.280 | 0.136 | 0.160 | 0.336 | 0.144 | 0.136 | 0.160 | 0.280 | 0.152 | 0.256 | 0.136 | 0.176 | 0.184 |
Taking those input data into the evaluation model which has been trained completely in Section
Obviously, the risk probability level of the Tiefodian tunnel is level IV, which belongs to high risk.
In order to verify the accuracy of the proposed model, FAHP is used to calculate the risk probability level of Tiefodian tunnel portal construction. According to the FAHP method described in Section
The total ranking of risk factors.
Risk factors |
|
|
|
|
Total ranking |
---|---|---|---|---|---|
0.5962 | 0.2616 | 0.0989 | 0.0434 | ||
|
0.2569 | 0.1467 | 0.4307 | 0.2522 | 0.2451 |
|
0.1450 | 0.0143 | 0.2591 | 0.0435 | 0.1177 |
|
0.4337 | 0.0423 | 0.0255 | 0.0802 | 0.2756 |
|
0.0148 | 0.0240 | 0.0141 | 0.0143 | 0.0171 |
|
0.0250 | 0.0753 | 0.0764 | 0.1365 | 0.0481 |
|
0.0444 | 0.4354 | 0.1500 | 0.0243 | 0.1563 |
|
0.0802 | 0.2619 | 0.0441 | 0.4490 | 0.1402 |
The risk value with fuzzy comprehensive evaluation is as follows:
The result shows that the risk level is level IV, which belongs to high risk.
According to Table
The BP fuzzy neural network method and FAHP method are used to calculate the risk probability level in Tiefodian tunnel portal construction. The results are shown in Table
Comparison of evaluation methods.
Evaluation target | BP fuzzy neural network method | FAHP method |
---|---|---|
Risk result | (0.0000, 0.0000, 0.0084, 0.9403, 0.0513) | (0.0126, 0.2332, 0.2608, 0.3798, 0.1134) |
Risk level | IV (high risk) | IV (high risk) |
The BP fuzzy neural network method and FAHP method are the same as the evaluation results of the risk probability grade in the construction stage of the Tiefodian tunnel. The advantages of the BP fuzzy neural network method are as follows: the BP fuzzy neural network method has the characteristics of strong fault tolerance and self-adaptability, which can overcome the inaccuracy of the result quantification of each risk factor in FAHP. Coupled with the fuzzy algorithm, the BP fuzzy neural network method is applied to form a composite evaluation system, which makes the results more reasonable and closer to the actual situation.
The risk probability level of the construction of Tiefodian tunnel portals is grade IV, which indicates high risk. Risk control measures should be applied to the construction to ensure the stability of the tunnel. In view of the risk assessment results in Section Adjust the construction method: the original construction method is the bench method, which can cause the surrounding rock disturbance. Therefore, we exchange the bench method for the centre diaphragm (CD) method. The excavation footage is no greater than 3.0 m, and the distance between different excavation parts is at 9.0 m to maintain tunnel stability. Meanwhile, the anchorage, primary lining, and invert arch are constructed in the tunnel. When the length of the invert arch is 12 m, the secondary lining is installed. The blasting charge should be strictly controlled in the portal section to reduce the blasting excavation risk. Finally, the dynamic monitoring of tunnel deformation should be strengthened during the construction process. The construction of the CD method is shown in Figure Reinforcement of surrounding rock: ensuring the stability of the tunnel portal is difficult because of the poor quality of the rock mass, the thinner cover depth, and rainfall in the Tiefodian highway tunnel. As a result of this, the presupport measures should be improved before the excavation of the tunnel. Actually, the pipe roof has a good effect as a presupport measure in the construction of the tunnel portals. A 30 m long pipe roof, which consists of a total of 44 seamless steel pipes (Φ108 × 6 mm) with a construction scope of 2 × 57°, is suggested as a presupport measure of the tunnel portal on the basis of the abovementioned analysis. The length of single pipe sections is 3 or 6 m. They are connected at the joints by 150 mm long threaded sections. With the use of the umbrella arch as a guide wall, the pipe roof is constructed along the outer contour line of the open cut. Strengthening the supporting parameters: the supporting parameters are important in avoiding tunnel face instability and great deformation of the tunnel portals. Therefore, the diameter of the rock bolt is improved to Φ28, and the length is increased to 4.5 m. The I22b steel frame in H shape is adopted, and the length of the pregrouting bolt is added to 5.0 m. In the construction process, the monitoring work should also be carried out simultaneously, and the measured data should be backanalyzed. The stability of the surrounding rock is determined by the analysis results of monitoring data. And the support parameters are adjusted to ensure the safety of construction.
Construction of the left tunnel portal by the CD method.
The deformation of the surrounding rock and the force of the supporting structure were simulated based on the construction conditions of the Tiefodian tunnel. The rationality of the proposed model and the risk measures was proved by simulating and analyzing with the MIDAS/GTS finite-element analysis software. The ZK164 + 570 section and ZK164 + 600 section are selected as the object of numerical simulation which belong to the Tiefodian tunnel portal.
According to the Saint-Venant principle, tunnel excavation has little effect on the surrounding rock which is located in 3–5 times the diameter. The model size is taken as
The model parameters after optimization.
Formation or structure | Elastic modulus |
Poisson’s ratio ( |
Bulk density (kN/m3) | Cohesion (kPa) | Internal friction angle (°) |
---|---|---|---|---|---|
Fully weathered gneiss | 0.2 | 0.4 | 18 | 50 | 28 |
Strong weathering gneiss | 0.5 | 0.35 | 20 | 200 | 32 |
Weathering gneiss | 1 | 0.3 | 22 | 320 | 35 |
This simulation is based on the bench method before the risk control measures in the construction. The model meshing and the supporting structure simulation are established as shown in Figures
Model meshing before risk controls.
Spray anchor support before risk controls.
The stress state of the surrounding rock is redistributed and deformed toward the temporary surface with the excavation process of the tunnel cavern. The deformation of the cavern can be the most intuitive and convenient response to the deformation state of the surrounding rock. The value of vault settlement and horizontal convergence is a typical observation project to reflect the deformation of cavern. The final deformation of the tunnel portal with the bench method before the risk control is shown in Figures
Vault settlement value.
Horizontal convergence value.
Obviously, the deformation of the surrounding rock after tunnel excavation is symmetrical, which is mainly reflected in the vault, and the sidewall converges inward. The maximum vault settlement value is about 54.91 mm, and the maximum arched part has a certain amount of spring back about 24.95 mm due to the unloading effect. The horizontal convergence value of the surrounding rock at the left and right sidewalls is about 46.22 mm. The vault settlement value and the horizontal convergence value of surrounding rock are larger than the specification requirement [
Adjust the construction method and support parameters according to the importance of risk factors; this simulation is based on the CD method after the risk control measures in the construction. The model meshing and the support structure simulation are established as shown in Figures
Model meshing after risk controls.
Spray anchor support after risk controls.
Through the numerical simulation analysis, the vault settlement and the horizontal convergence cloud diagram of surrounding rock at each construction stage obtained are shown in Figures
Vault settlement value.
Horizontal convergence value.
It can be seen from Figures
The numerical simulations of two cases are listed and analyzed in order to verify the effectiveness of the proposed measures. The details are shown in Table
The comparison of the deformation value.
Deformation value | Evaluation index | ||
---|---|---|---|
Vault settlement (mm) | Horizontal convergence (mm) | Specification allowed (mm) | |
Before the risk control | 54.91 | 46.22 | 40 |
After the risk control | 6.27 | 13.32 | 40 |
It can be seen from Table
The historical data of the 50 tunnels.
Number | Tunnel name | Number of one-way lanes | Location | Surrounding rock level | Tunnel length (m) | Maximum depth (m) |
---|---|---|---|---|---|---|
1 | Aofeng Mountain Tunnel | Three lanes | Fuzhou | IV | 1867 | — |
2 | Lake Yun No. 1 Tunnel | Two lanes | Mianzhu | V | 3555 | 400 |
3 | Mountain Hu Tunnel | Two lanes | Nanjing | IV | 1645 | 466 |
4 | Cang Ridge Tunnel | Two lanes | Taijin | IV | 7530 | 768 |
5 | Jiuling Mountain Tunnel | Two lanes | Wuji | V | 5473 | 887 |
6 | Qingshan Hillock Tunnel | Two lanes | Changsha | V | 1245 | 80 |
7 | Xuefeng Mountain Tunnel | Two lanes | Anhua | IV | 7039 | 780 |
8 | Anyuan Tunnel | Two lanes | Anyuan | V | 6868 | 470 |
9 | Lujialing Tunnel | Two lanes | Chongqing | IV | 6664 | 701 |
10 | Wujian Ridge Tunnel | Two lanes | Yongzhang | V | 1050 | 197 |
11 | Mayazi Tunnel | Two lanes | Wuguan | V | 9007 | 714 |
12 | Shanziding Tunnel | Two lanes | Meida | V | 500 | 110 |
13 | Luwan Tunnel | Three lanes | Lishui | IV | 708 | 170 |
14 | Mountain Mao Tunnel | Three lanes | Changning | V | 1628 | 50 |
15 | Tongzhou Tunnel | Two lanes | Yongjia | IV | 493 | 102 |
16 | Wuzhi Mountain Tunnel | Three lanes | Leshan | V | 3923 | 800 |
17 | Xiang River Tunnel | Two lanes | Huangyuan | IV | 1858 | 300 |
18 | Zhongtiao Mountain Tunnel | Two lanes | Jiezhou | V | 7423 | 605 |
19 | South Village Tunnel | Two lanes | Nancun | IV | 6787 | 153 |
20 | Wolonggang Tunnel | Three lanes | Beijing | V | 420 | 35 |
21 | Liujiapai Tunnel | Two lanes | Liujiapai | V | 1233 | 136 |
22 | Nan Yanmenguan Tunnel | Two lanes | Shanyin | IV | 5247 | 600 |
23 | Foling Tunnel | Two lanes | Linfen | IV | 8803 | 762 |
24 | Xilingjing Tunnel | Two lanes | Taijia | V | 6555 | 711 |
25 | Yanmenguan Tunnel | Two lanes | Xizhou | V | 5183 | 600 |
26 | Yangtou Mountain Tunnel | Two lanes | Qianjiang | V | 5385 | 466 |
27 | Sumu Mountain Tunnel | Three lanes | Huhetaote | IV | 3213 | — |
28 | Xuefeng Moutain Tunnel | Two lanes | Shaoyang | V | 6958 | 840 |
29 | Taining Tunnel | Two lanes | Taining | IV | 7039 | 485 |
30 | Leigong Mountain Tunnel | Three lanes | Xiamen | V | 3433 | 253 |
31 | Zhengjiayuan Tunnel | Two lanes | Zhashui | V | 2037 | 189 |
32 | Foyangling Tunnel | Four lanes | Binzhou | IV | 3904 | 70 |
33 | Baiyangwan Tunnel | Three lanes | Hangzhou | V | 1400 | 52 |
34 | Wanxichong Tunnel | Three lanes | Kunming | V | 7980 | 754 |
35 | Mountain Tiger Tunnel | Four lanes | Jinan | V | 1880 | 276 |
36 | Jianping Tunnel | Two lanes | Tongchuan | IV | 1287 | 240 |
37 | Yanling Mountain Tunnel | Two lanes | Hangzhou | IV | 1250 | 178 |
38 | Jinzhuwan Tunnel | Three lanes | Chongqing | V | 1322 | 255 |
39 | Yangzong Tunnel | Three lanes | Yuxi | V | 2727 | 141 |
40 | Magongci Tunnel | Four lanes | Zibo | IV | 655 | 86 |
41 | Wulidun Tunnel | Two lanes | Rucheng | V | 2380 | 980 |
42 | Zhenbao Tunnel | Two lanes | Boshan | IV | 2880 | 32 |
43 | Daiyuling No. 2 Tunnel | Two lanes | Zhuanghe | V | 2930 | 262 |
44 | Shimenya Tunnel | Two lanes | Yichang | V | 7524 | 894 |
45 | Queer Mountain Tunnel | Two lanes | Ganzi | IV | 7079 | 700 |
46 | Xueshanliang Tunnel | Two lanes | Abazhou | IV | 6950 | 598 |
47 | Ziyang Tunnel | Two lanes | Ziyang | V | 7938 | 904 |
48 | Baihua Mountain Tunnel | Three lanes | Wuding | V | 1620 | — |
49 | Jiaodongao Tunnel | Two lanes | Ningbo | IV | 2185 | 410 |
50 | Shigu Tunnel | Three lanes | Dongguan | IV | 4011 | 500 |
This paper develops a BP fuzzy neural network model based on fuzzy theory and BP neural network to handle the vagueness and subjectivity in risk evaluation of highway tunnel portal construction. A risk evaluation model which is obtained from historical data of 50 tunnels is established by combining the fuzzy method with the BP neural network. The proposed model is applied for the risk assessment of the Tiefodian tunnel. The results show that the risk evaluation level is IV and slope instability is the greatest impact index among four risk events. Based on above analysis, we conclude that supporting parameters, rainfall and groundwater, and surrounding rock level precede the others. At the same time, this model is confirmed to be available in risk evaluation of the tunnel portal by using the fuzzy analytic hierarchy process (FAHP). According to the evaluation results, corresponding risk control measures are suggested and taken. Besides, numerical simulation is carried out before and after the implementation of risk measures, respectively. The rationality of the proposed risk evaluation model is proved by comparing the numerical simulation results.
However, the proposed method is still a semiquantitative risk evaluation method. Because of the complicated geological conditions and uncertainties of tunnel construction, the risk identification is based on the rich experience of experts in this paper. Although the main risk factors can be identified, some less influence of the risk factors has the probability be missed. Moreover, the number of tunnel samples for the proposed model needs to be further expanded to improve the accuracy of risk evaluation.
The authors declare that there are no conflicts of interest regarding the publication of this paper.
This research was financially supported by the National Natural Science Foundation of China (Grant nos. 51408054 and 51678500); the Natural Science Foundation (2017JM5136) by the Science and Technology Department of Shaanxi Province; the Housing and Urban-Rural Construction Foundation (2017-K55) by the Housing and Urban-Rural Department of Shaanxi Province; the Scientific Research Program (KLTLR-Y14-15) for Technology of Highway Construction and Maintenance Technology of National Transportation Industry Key Laboratory; Higher Education Foundation (2050205) by the Financial Department of Shaanxi Province; and the Research Program (XAGDXJJ16003) sponsored by Xi’an Technological University.