Bayesian assessment of utility tunnel risk based on information sharing mechanism

1 Introduction

Utility tunnel is used to lay municipal pipelines such as electricity, communication, radio and television, water supply, drainage, heat, gas, etc. in a city underground. It replaces the traditional practice of burying pipelines directly in the ground when laying pipelines, and is an important part of the modernization of urban infrastructure. However, in the process of urban underground infrastructure construction in our country, there have been problems such as insufficient scale of underground pipeline construction and low management level. With the frequent occurrence of heavy rains, waterlogging, pipeline leaks and explosions, and road collapses in recent years, the “underground” problem cannot be ignored [1, 2]. Due to the general lack of forward-looking and systematic construction of municipal pipelines in the past, different pipeline departments are acting independently. Based on their own planning and construction, it has caused problems such as crisscrossing of pipelines, lack of safe distance, and insufficient carrying capacity, resulting in huge pipeline accident losses. Effective monitoring, prevention and control of pipeline risks in the pipeline gallery is essential to the sound operation of the utility tunnel.

To achieve effective monitoring of utility tunnel, we can start with project construction specifications, operation and maintenance, and achieve timely monitoring and early warning. The operation and maintenance management of buildings is the integration of personnel, facilities, technology and management processes, mainly including planning, maintenance, repair, and emergency management of personnel and living spaces. Its purpose is to meet the basic use, safety and comfort requirements of personnel in the building space [3]. During the operation and maintenance of the integrated pipeline gallery, various types of pipelines will suffer fatigue and corrosion in different forms and degrees of damage under the long-term effect of the coupling relationship. At the same time, under the complex external environment and long-term different loads, the utility tunnel structure will also have different forms of safety hazards such as structural cracks, uneven settlement, and groundwater infiltration. Under the long-term impact of the coupling effect, the damaged pipelines and utility tunnel structures will become weak links during the operation and maintenance of utility tunnel, and eventually evolve into safety accidents. After years of operation and maintenance of the integrated pipeline gallery, various pipelines will experience varying degrees of aging, and a complex disaster-causing environment may be formed inside. Once a disaster event occurs in the pipeline corridor, various disasters will not exist in isolation, and there are often inter-connections in time, space and causes, forming a complex disaster chain system or disaster chain process [4]. Therefore, the pipeline risk protocol is an indispensable link in the operation and maintenance management of the integrated pipeline gallery.

Commonly used risk assessment methods in construction projects include risk matrix method, fault tree analysis method, fuzzy comprehensive evaluation method, Monte Carlo simulation method and Monte Carlo simulation method based on discrete event tree. However, the existing methods can only perform qualitative or semi-quantitative analysis, and cannot accurately reflect the level of risk from a quantitative perspective; moreover, the current risk assessment methods cannot integrate existing research results into existing research. That is, the accumulation of historical experience has not been well used, and the uncertain factors cannot be accurately considered [5]. The Bayesian network is proposed to solve the problem of uncertainty and incompleteness, and is considered by scholars at home and abroad to be the most ideal tool for knowledge representation, inference, and prediction in an uncertain environment [6] It has great advantages in solving complex construction safety risks caused by multiple risk factors. In summary, the following author will combine the relevant theories of BN to model the uncertainty of the pipeline operation and maintenance management process, and combine the prior knowledge and observation events to perform multi-angle uncertainty reasoning to realize the management of the pipeline. Comprehensive analysis and assessment of risks of corridors and pipelines.

Canto-Perello’s research has identified the human risk factors in the underground integrated pipe gallery, and proposed planning recommendations that will help employees adapt to the tunnel working environment [7]. Ward explained the most important steps and tasks in risk management, and provided the easiest way to manage risk [8]. Kate made a risk management analysis for urban construction projects, pointed out that specific risk management methods were divided into technical means and non-technical means, and provided risk response strategies [9]. Gachie clarified the main body of construction project risk management, that is, the owner. Its role in project construction is particularly obvious. It needs to carry out risk management for the whole life cycle of the project, grasp the project progress, and strictly control the project risk [10]. Chen et al.’s study developed a simplified risk value (SRV), which was derived using a risk assessment simulation tool named Areal Locations of Hazardous Atmospheres (ALOHA), version 5.4.7 and geography information systems including SuperGIS, version 10.1, and Surfer, version 10 to produce a potential risk map (PRM) for underground pipeline leakage in Taipei [11], thus many information methods were used to controle the risks [12–15], so as to ahieve the sustainable development mode [16].

So, the study for exploring the cooperative mechanism of utility tunnel company and pipeline company has been a hot place. This paper uses the Stackelberg game model to analyze the resource sharing among utility tunnel institutions within the utility tunnel association, which helps to understand the decision-making process in the game, and its equilibrium results can guide the decision-making behavior of people seeking utility tunnel treatment. And then, the potential disaster risk factors in the operation process of utility tunnel, and it constructs the risk early warning model of integrated pipeline corridor based on Bayesian network. Bayesian networks can construct the causal relationship of utility tunnel operation safety risk accidents from the perspective of utility tunnel company and pipeline company error.

2 Cooperative game for risk treatment

In the centralized risk decision-making, there is only one decision-maker that decides the price and the scale of utility tunnel risk treatment, that is, the pipeline company and the utility tunnel company seek utility tunnel risk treatment are regarded as an overall benefit, and the optimal scale and price of utility tunnel risk treatment are decided by maximizing the overall benefit of the pipeline company and the utility tunnel company seek utility tunnel risk treatment. It can be seen that: the optimal scale of utility tunnel risk treatment in centralized decision-making is Q centralized * = 1 a [ ( b + h + c ) - ( b + h + c ) 2 - 2 ab + a 2 - 2 ah ]

The optimal risk service benefit of utility tunnel company and pipeline company is ∏ centralized * = 1 6 aQ centralized * 3 - 1 2 ( b + h + c ) Q centralized * 2 + ( b - 1 2 a + h ) Q centralized * - 1 2 h - k

In the following analysis, compared with a Stackelberg game, can centralized decision-making bring higher benefits to utility tunnel company? Firstly, the overall benefits of centralized decision making and decentralized decision making are compared and analyzed as follows: ∏ centralized * - ∏ l arg e s * - ∏ small s * = [ 1 6 aQ centralized * 3 - 1 2 ( b + h + c ) Q centralized * 2 + ( b - 1 2 a + h ) Q centralized * ] - [ 1 6 aQ s * 3 - 1 2 ( b + h ) Q s * 2 + ( b - 1 2 a + h ) Q s * ] > 0

Then, the optimal scale of utility tunnel risk treatment for centralized and decentralized decision making is compared and analyzed as follows: Q centralized * - Q s * = 1 3 a [ ( b + h + c ) - 3 ( b + h + c ) 2 - 2 ab + a 2 - 2 ah + 4 ( b + h + c ) 2 - 3 a ( 2 b - a + 2 h ) ] Q centralized * - Q s * > 0 1 3 a [ ( b + h + c ) - 3 ( b + h + c ) 2 - 2 ab + a 2 - 2 ah + 4 ( b + h + c ) 2 - 3 a ( 2 b - a + 2 h ) ] > 0 ( b + h + c ) - 3 ( b + h + c ) 2 - 2 ab + a 2 - 2 ah + 4 ( b + h + c ) 2 - 3 a ( 2 b - a + 2 h ) > 0

So, it is concluded that Q concentrated * - Q s * > 0 is constantly established. Therefore, it can be seen that centralized decision-making can increase the total revenue of utility tunnel company and pipeline company, and the scale of utility tunnel risk treatment for centralized decision-making is larger than that for decentralized utility tunnel risk treatment.

From the previous section, we can see that the optimal scale of utility tunnel risk treatment for utility tunnel company and utility tunnel decision-making is Q s * = 1 3 a [ 2 ( b + h + c ) - 4 ( b + h + c ) 2 - 3 a ( 2 b - a + 2 h ) ] . The optimal price is w s * = 1 2 aQ s * 2 - ( b + h ) Q s * + ( b - 1 2 a + h ) . Study the sensitivity of the optimal utility tunnel risk treatment price and optimal utility tunnel risk service scale with respect to b, a, c, h.

Theorem 1.The scale of consultation is a monotonically increasing function of b.

Proving ∂ Q s * ∂ b = 1 3 a [ 2 - 1 2 * 8 ( b + h + c ) - 6 a 4 ( b + h + c ) 2 - 3 a ( 2 b - a + 2 h ) ] = 1 3 a [ 2 - 4 ( b + h + c ) - 3 a 4 ( b + h + c ) 2 - 3 a ( 2 b - a + 2 h ) ]

When 4 (b + h + c) -3a ≤ 0, then ∂ Q s * ∂ b = 1 3 a [ 2 - 4 ( b + h + c ) - 3 a 4 ( b + h + c ) 2 - 3 a ( 2 b - a + 2 h ) ] ⩾ 2 3 a > 0 .

Simplify

4 ( b + h + c ) 2 - 3 a ( 2 b - a + 2 h ) > 2 ( b + h + c ) - 3 2 a . From 4 (b + h + c) -3a > 0 to know 2 ( b + h + c ) - 3 2 a > 0 . Since 6 c + 3 4 a > 0 is constant, it can be seen that ∂ Q s * ∂ b > 0

Therefore, the scale of utility tunnel risk service is a monotonically increasing function of b.

Theorem 2. The scale of risk treatment is a monotonically increasing function of h.

Proving: ∂ Q * ∂ h = 1 3 a [ 2 - 1 2 * 8 ( b + h + c ) - 6 a 4 ( b + h + c ) 2 - 3 a ( 2 b - a + 2 h ) ] = ∂ Q s * ∂ b > 0

Therefore, the risk service level is a monotonically increasing function of h.

Theorem 3. The scale of utility tunnel risk service is about c being a monotonically decreasing function.

Proving: ∂ Q * ∂ c = 1 3 a [ 2 - 4 ( b + h + c ) 4 ( b + h + c ) 2 - 3 a ( 2 b - a + 2 h ) ]

Simplify 4 ( b + h + c ) 2 - 3 a ( 2 b - a + 2 h ) < 2 ( b + h + c ) . From b > a, it can be concluded that is constant, in summary, ∂ Q s * ∂ b < 0 .

Therefore, the risk service scale of utility tunnel is about c being a monotonically decreasing function. The research analyzes the decentralized and centralized decision-making between utility tunnel company and pipeline company under Stackelberg.

4 Bayesian network risk warning based on information sharing

The utility tunnel company is the leader, the pipeline company is the follower, and the decision is made by the leader, the utility tunnel company first, and the follower company based on the risk service choice of the utility tunnel company. decision making. Similarly, centralized decision-making can increase the total income of utility tunnel company, and the scale of utility tunnel risk treatment for centralized decision-making is larger than that for decentralized decision-making.The risks related to utility tunnel can be divided into three categories, which are numbered: natural disaster risk A1, direct pipe corridor risk A2, and man-made operation risk A3. The specific risk factors are listed in Table 2.

In the Bayesian network directed acyclic graph, nodes represent random variables, and the edges between nodes represent the logical dependence between variables. Each node is attached with a probability distribution. Parent node C is attached to its edge distribution P(C), and child nodes B, A, R are attached to conditional probability distributions P (B|C) , P (B|C) , P (A|R). Bayesian network is a representation of joint probability distribution. Taking C2, C3⟶B2⟶A2⟶R chain as an example, the joint probability distribution function including all nodes is: P ( C 2 , C 3 , Bds 2 , A 1 , R ) = P ( R | A 1 ) P ( A 1 | B 1 ) P ( B 1 | C 2 , C 3 ) P ( C 2 ) P ( C 3 )(13)

After determining the number of experts and the weights of experts, the expert questionnaire survey results are calculated according to the following formula, the probability distribution of the probability levels of disaster risk factors in the operation and maintenance of utility tunnel can be obtained, that is, the probability distribution of the parent node risk factors in the Bayesian network model probability. P ( C i = j ) = ∑ k = 1 n w k P ijk / ∑ k = 1 n w k(14)

In the formula, i = 1,2,...,9; j = 1,2,...,5; P(Ci = j) is the probability that the risk factor Ci is at level j; n is the number of questionnaires or experts Number; w _k is the weight of the k-th expert; P_ijk is the probability that the k-th expert believes that the risk factor Ci is at level j, P_ijk = 0 or 1.

As the utility tunnel is put into use, the risk profile changes and evolves dynamically with the timeline of use, with different geological conditions, pipeline settings, management issues and other objective factors causing the risk sources to vary. The dynamic element incorporated by dynamic Bayes does not mean that the network structure changes over time, but that the sample data, or observations, change over time. A Bayesian network can be defined as BN=(G, θ), where G is a directed acyclic graph of the joint probability distribution over X and θ denotes the parameters of the network. where the joint probability distribution over X is defined as P ( X 1 , X 2 ⋯ , X n ) = ∏ t = i n P ( X i | P a ( X i ) )(15)

Dynamic Bayesian network models extend this formulation to model stochastic processes with a time component. In order to represent a stochastic process in terms of a Bayesian network, a probability distribution over the random variables X₁,  X₂, …, X _n needs to be obtained, but such a distribution is very complex. Therefore, in order to be able to study and model complex systems accordingly, some assumptions and simplifying conditions need to be made.

Assumptions.

assume that the conditional probability change process is uniformly smooth for all t in a finite time.

Based on the above assumptions, a dynamic Bayesian network with a joint probability distribution built on the time trajectory of a stochastic process then consists of two parts: a priori net B₀defined on the joint probability distribution of the initial state X_[1]; and a transfer net B_→→defined on the transfer probabilities P (X_[t+1]|X_[t]) of the variables X_[1] and X_[2]. Thus, its joint probability distribution is. P ( X [ t ] , X [ 2 ] , ⋯ X [ t ] ) = P B 0 ( X [ 1 ] ) ∏ t = 1 T P B → ( X [ t + 1 ] | X [ t ] )(17)

With reference to the International Tunnel Association (ITA), the probability of disaster events related to the risk of integrated pipeline corridors is divided into 4 levels (Rare, Occasionally, Possible, Frequent) and the classification standards (Ignorable, Need to consider, Alert, Danger). For negligible risk factors, the probability of risk occurrence and adverse effects are very small, and the main energy can be focused on daily management and operation to maintain and ensure the orderly operation of the project; when risk factors are in a state of alert, closely monitor the changes in such risk factors, especially risk factors that deteriorate the situation and risk factors that hinder the orderly operation of the pipeline gallery. Formulate preventive measures according to the specific situation; what needs to be focused on is the risk factors of the dangerous state, which have a high probability of inducing disaster risks and must be effectively controlled.

Mainly for the pipe gallery design, construction, operation and maintenance and scientific research personnel of universities and colleges, for the Beijing comprehensive pipe gallery, questionnaires were issued to relevant personnel, and 12 copies were returned, of which 10 were valid questionnaires, with an effective rate of 83.3%. Apply the expert survey method to collect experts’ opinions on the levels of the above risk factors and the weights of the risk factors. Due to the dynamics and uncertainties of risk factors related to pipe corridors and pipelines, the sample data often has some unobservable hidden variables. Therefore, this paper adopts an iterative convergence algorithm for samples with missing values-EM algorithm for parameter learning. Iteration makes the model parameters tend to the maximum likelihood estimation, and finally obtains the conditional probability distribution, thereby determining the probability of the level distribution of each disaster risk factor, more details are shown in Table 3.

With the help of Bayesian network realization platform Netica software, the risk level distribution of integrated pipeline gallery is systematically evaluated. It can be found that (1) Earthquake disaster C1 and geological disaster C3 are at level 2 risk, which can be ignored. Daily monitoring of such natural disaster risks is enough; (2) Meteorological disaster C2 and gas pipeline deformation, leakage C4 and Deformation and leakage of the water pipeline C5 is at level 3 risk. Pay attention to strengthen corresponding management and inspections, and pay attention to the conversion of risks; (3) Electricity leakage, overheating C6 and thermal pipeline deformation, leakage C7, and irregular personnel dressing and operation C8 and external construction infringement C9 are at a level 4 risk.

It is necessary to closely monitor the changes of such risk factors and formulate preventive measures for specific situations. It can also be seen that the occurrence probability of natural disaster risk A1 is level 3, and the occurrence probability of pipeline gallery and pipeline direct risk A2 and man-made operation risk A3 is level 4. According to the principle of maximum membership in fuzzy comprehensive evaluation, comprehensive pipeline gallery. The overall risk probability level is 5, and the risk probability is relatively high. Therefore, the risk control during the operation and maintenance of the pipe gallery should be strengthened.

Using the reverse reasoning ability of the Bayesian network, two critical paths leading to the occurrence of disasters in the operation and maintenance of the integrated pipeline corridor can be found, namely C6⟶B5⟶A2⟶R and C8⟶B7⟶A3⟶R, namely During the operation and maintenance of the integrated pipeline gallery, the key risk factors leading to the occurrence of disasters are the leakage of power lines, overheating C6, and personnel dressing and irregular operation C8. The risk assessment results are in line with the overall status quo of Beijing’s integrated pipeline gallery. In the pipeline operation and maintenance process of the integrated pipeline gallery in Beijing, including the pipeline corridors that have been built and under construction, the probability of occurrence of various potential risk factors should be controlled according to the actual situation on the site, and the above two risk factors should be focused on controlling the power supply. For cable leakage protection, pay attention to supervising and managing the dressing and operation specifications of the operators, do a good job of disaster event risk control at the source, ensure the safety of pipeline gallery and pipeline operation and maintenance, and provide guarantee for the development and operation of the city.

Utility tunnel replaces the traditional practice of burying pipelines directly in the ground when laying pipelines, and is an important part of the modernization of urban infrastructure. However, in the process of urban underground infrastructure construction, there have been problems such as insufficient scale of underground pipeline construction and low management level. Meanwhile, the main body of the utility tunnel is all kinds of pipelines, so the normal operation of the pipelines is very important for the operation benefit of the comprehensive utility tunnel; meanwhile, identifying the potential risks of the pipelines in time can reduce the losses caused by the uncertain risks to the comprehensive utility tunnel.

Based on the characteristics and advantages of Bayesian network, a method for risk analysis and early warning of urban underground comprehensive pipeline gallery based on Bayesian network is proposed. In the process of risk assessment and analysis, the expert survey method is used to calculate the probability of occurrence of various potential disaster risk factors in the pipe gallery, which can not only play the role of expert experience, but also reduce the bias caused by subjective human factors and bring the risk assessment results closer In reality. The risk analysis results show that leakage, overheating of power pipelines, and irregular personnel dressing and operation are the key factors leading to disasters. Corresponding risk prevention and control should be emphasized. The application of the Bayesian network model can reflect the disaster risk level of the integrated pipe gallery operation and maintenance, which verifies the feasibility of the method, and also provides a basis for the safe operation and maintenance of the integrated pipe gallery and disaster prevention and mitigation in the process of operation and maintenance.

This research is of positive significance for dealing with the complexity and uncertainty of the safety risk of urban underground comprehensive pipeline gallery. The contributions are:on the one hand, in view of the complex composition of the risk elements related to the integrated pipeline corridor and the complex relationships between the elements, the use of rich text materials, combined with expert consultation, analyzes the main composition of information security risk sources, and clarifies the hierarchical structure relationship between the risk elements. A clear, structured, and systematic smart city information security risk assessment index system has been constructed; on the other hand, in view of the uncertainty in the evolution and derivation of smart city information security risks, a probabilistic perspective is adopted and Bayesian The applicability of the network in dealing with the problem of uncertainty is applied to the quantitative study of the safety risk assessment of urban underground comprehensive pipeline gallery.