Risk Management Methodology for Transport Infrastructure Security

Abstract

The development of transport infrastructure is associated with risks, expressed in the likelihood of harm to the road users’ health during road accidents and their consequences. The risk management process is aimed at reducing the influence of factors that contribute to the occurrence of an accident and increase the consequences’ severity after it. This article proposes a risk management methodology within five stages: identification, analysis and evaluation, processing, development of recommendations, and monitoring. For each step, we describe the methods and models that allow us to effectively solve the problem of risk management. We proposed a risk management algorithm based on feedback. We tested the adequacy of the methodology on a specific example: we conducted an analysis, an assessment, and proposed risk management measures in the field of ensuring road safety in a small town.

1. Introduction

Ensuring road safety is one of the priorities of state policy. Road accidents pose a serious threat to the sustainable functioning of the transport complex and cause significant social and economic damage. The number of deaths as a result of road accidents globally has already exceeded 1.35 million people a year and, unfortunately, by 2030, this indicator is predicted to triple in growth due to the countries that do not pay much attention to this problem.

To reduce the road accident rates, first of all, it is necessary to conduct scientific research aimed at developing a comprehensive mechanism for preventing accidents as well as analyzing, assessing, and managing the risks of road safety.

Since many traffic parameters (flow intensity in the first place) are stochastic in nature, and the functioning of the transport system is associated with risks (material, human, timing), any change in the traffic situation can lead to the system’s malfunction (deterioration of the traffic situation, and in the most serious cases—to a transport collapse). Therefore, any actions to change the infrastructure or the management system must be scientifically justified. In such situations, it is always important to calculate the possible risks of road safety, their probable harm to the road users’ health and material losses during road accidents and their consequences as well as also have risk management tools.

Therefore, the purpose of this article was to develop a comprehensive methodology aimed at managing the risks to ensure the security of the transport infrastructure.

2. Literature Review

The assessment of transport and social risk and their predictive values are defined in the Federal Target Program “Improving Road Safety in 2013–2020.” [1], however, the list of factors causing risks is incomplete.

To classify the road safety risks, the author developed a concept in [2]. According to this concept, road traffic is a complex socio-technical system consisting of three subsystems (levels): (1) a subsystem, the functioning of which directly ensures the satisfaction of the transport need, namely the spatial movement of people and goods; (2) a subsystem for the traffic preparation and maintenance; and (3) a subsystem formed by executive authorities that implement the functions of state administration of a sectoral and intersectoral nature in relation to the subsystems of the first and second levels. Each level has its own risks. He classified the risks of each group in accordance with the selected criteria: risk cause (kind); place of risk; risk subject and source; objects to which the risks are directed; and the size and extent of the damage [3].

An important direction in ensuring road safety, as the author notes in his work [4], is the management of road safety risks, which are expressed in the likelihood of causing harm to road users from the road accidents and their consequences. The process of traffic safety risk management, in his opinion, can be defined as reducing the influence of factors that contribute to the appearance of the accidents’ causes. These factors are risk factors, circumstances that affect the risk likelihood or consequences, but are not their direct cause. A group of researchers [5] developed characteristics by which risks were classified in accordance with the road safety requirements. The risk classification was carried out by taking into account the program of the driver’s behavior, which was presented in the paper as the maximum entropy of their perception field. The choice of this criterion was determined by the possibility of a comprehensive assessment of many road environment factors that affect the driver.

The authors of [6] analyzed an approach to calculate the car collision probability (collision detection assessment) in the context of their prevention. The authors proposed a study of the collision probability without temporary collision measures as an intermediate or necessary value, which makes it possible to obtain the collision probability over a long period of time by integrating it over time. According to the authors in [7], rapid prediction of the accident severity allows trauma and emergency centers to assess the potential damage from an accident and, accordingly, send the appropriate emergency rescue units to provide appropriate emergency care. The authors proposed a two-level ensemble machine learning model to predict the crash severity. The first level combined four basic machine learning models: the k-nearest neighbor, decision tree, adaptive boost, and support vector machine; the second layer classified the accident severity based on a feed-forward neural network model. The authors used as input data only those accident signs that could be obtained instantly and easily. In [8], to predict the occurrence of accidents, vehicle failures, and driver errors, a fuzzy cognitive model was built. This allowed them to take into account a large number of heterogeneous factors. The predictive model to estimate the transportation duration was based on the results of the expert survey, and fuzzy logic was used to formalize the expert estimates.

The authors of [9] pointed out that compared to vehicle collisions, collisions with pedestrians were less studied due to insufficient data, but the development of image processing technology contributed to solving this problem through the use of video data. They used video data to analyze pedestrian–vehicle conflicts.

The study in [10] was devoted to a risk-based analytical framework to estimate the number of fatalities in a crash. It combined the probabilities of both an accident and the fatalities in the event of an accident. It was assumed that the frequency of crashes between the different road users was proportional to their roadway usage. The authors estimated the fatality rates for various combinations of vehicles and pedestrians and concluded that in the absence of road safety measures, there was first an upward trend and then a downward trend in mortality. The lowest road death rate was observed in the bus-dominated scenario. At the same time, it was also necessary to take into account the risks of a long-term period, when the consequences of accidents manifest themselves with a delay in time. This approach to assessing risks in manufacturing was developed by the authors in [11].

Although public transport is considered the safest, assessing the risk of accidents on bus routes can improve the safety of transport operators. In [12], the authors corrected the existing methods for assessing the safety of bus networks by integrating the safety factors, predictive models, and risk assessment methods. The experiment showed that transport managers can use this methodology to accurately analyze safety on each route.

We cannot ignore the problems of the technical component degradation, where an additional problem is that the security barriers in risk management cannot always be classified as equipment, since they often consist of operational and organizational elements. In [13], the authors combined system modeling with the monitoring of these elements.

Infrastructure is also an important component. In [14], the authors assessed the relationship between the infrastructure deficiencies and the frequency of crashes and severity and found that a high density of access roads strongly affected the collision frequency.

To manage and minimize risks, researchers have offered various methods. In [15], the authors developed a method for using subjective video annotations to assess the risk likelihood in avoiding a vehicle collision as well as to characterize the dynamics of a vehicle and the events in classifying the accident risks. The authors in [16] developed a unified approach to substantiate the requirements for an automated traffic control system as a system of “measuring equipment—identification—vehicle traffic control”. They showed that it is advisable to use multi-criteria optimization. In the article by [17], the authors discussed topical issues of vehicle traffic control when using operational and technical control systems. The authors substantiated promising directions to improve the traffic control system using automated means of monitoring compliance with traffic rules. Thus, in [18], it was found that the installation of speed cameras in various sections showed a decrease in the total number of accidents by 70%, accidents with property damage by 53%, a decrease in accidents with injuries by 84%, and a complete absence of fatal accidents.

The relevance of research on road risk management is increasing due to the possible emergence of autonomous vehicles (AV) on public roads. The authors of the study in [19] calibrated and evaluated the responsibility-sensitive safety (RSS) model developed by Intel using possible crash situations where the minimum time to collision was less than 3 s. In each scenario, the human driver was replaced by an AV controlled by the adaptive cruise control (ACC) system based on the model predictive control (MPC) built into the RSS model. They evaluated and compared three types of safety performance: human driver, ACC model, and ACC model with built-in RSS. As a result, they found that the third one optimally improved safety performance in emergency situations. The driver has a great influence on the intention to adopt, use, and adapt AV. The study in [20] proposed an objective method to assess and predict driver confidence in AV technologies. Data on the frequency of use of advanced driver assistance systems (ADAS) and self-reported confidence ratings were collected and combined to classify driver confidence levels as low, medium, and high based, which found that the driver’s hand position during ADAS activation (which means during lane departure warning and lane keeping assist) was closely related to their confidence level.

Increasing intellectualization helps to solve the problems of road safety. The growth data collected in transport systems can greatly improve the mobility research and provide valuable mobility insights for commuters, data analysts, and transport operators. In this case, it is advisable to use machine learning methods [21,22,23].

Despite the diversity and extensiveness of the research on risk assessment in various aspects, there is a need to develop a comprehensive methodology for managing the transport system risks. The analyzed research did not consider the risks as a complex, highlighting one or more sides. First of all, they singled out a person or the influence of other factors through them indirectly. Without excluding the dominance of human factors, it is necessary to remember the importance of the vehicle technical condition, infrastructure, and organizational measures. Several studies have fully presented the classification of risks, sources, and consequences, but did not provide mechanisms by which the root causes of accidents can be influenced. At the same time, the growing array of the initial information about the accident rate requires appropriate powerful processing methods. Therefore, this study was aimed at developing a comprehensive multi-stage methodology to manage the transport system risks, based, among other things, on the modern methods of machine learning and simulation, which ensures road safety.

3. Research Methodology

In this research, we propose the use of a five-stage process of traffic safety risk management:

• Risk identification: classification and identification of potential road safety risks.
• Risk analysis and assessment: determination of the risk’s likelihood identified during the risk identification stage as well as their consequences. To achieve this goal, the statistical data of past years as well as previous experience are widely used.
• Risk treatment: choice of risk management methods. The main risk management methods include risk minimization, risk acceptance, risk transfer, and risk rejection.
• Development of risk management activities, which includes the direct planning of activities as well as the appointment of the so-called “owners” of risks.
• Permanent control over risks: risk monitoring, timely adequate response to changes in the system, and the assessment of the risk management effectiveness.

The risk identification stage systematizes the identified risks for further study. As is stated in [24]: “management decisions effectiveness in the development of a system for improving traffic safety largely depends on forecasting and correct risk classification”. Most often, at this stage, quantitative methods are used that are based on the mathematical modeling of risk events for their possible forecasting.

Risk analysis and assessment is part of the risk management process. As a rule, it is a structured process including the identification of ways to achieve the set goals, the analysis of the consequences, and the likelihood of the occurrence of hazardous events in order to decide on the need for risk treatment. Risk assessment is a process that combines the identification, analysis, and comparative risk assessment. At this stage, qualitative methods are used and are based on expert opinion to assess risks. The peculiarity of these methods is to use the experts’ creativity and experience through questions rather than strict mathematical models and proof.

At the processing stage, a choice of risk management methods is carried out, which corresponds to risk minimization, risk acceptance, risk transfer, and risk rejection. One of the visual methods of risk treatment is the bow-tie analysis [25,26,27], which allows the relationship between the risk sources and the implementation consequences to be shown. This method displayed risk (risk event) in one diagram as well as all of its sources, possible consequences, and other related entities such as the key risk indicators (KRIs) or mitigation measures from the risk realization. The method is mainly applied in situations where it is difficult to conduct a complete fault tree analysis or when the study is more focused on creating barriers or controls for each risky situation path. The method can be useful in a situation where there are precisely established independent paths leading to a risky situation.

When constructing a bow-tie diagram, a dangerous event is determined as a central node. Then, using the study of risk sources, a list of causes influencing the event under consideration is compiled. After that, the mechanisms for the development of danger to a critical event are identified. It is necessary to draw a line separating the cause from the event. This allows the bow-tie diagram to be formed on the left side, where additional factors that can lead to the mitigation of a dangerous event and its consequences can be identified and included in the diagram. On the right side of the diagram, the various hazardous event consequences are identified and a line connecting the central event to each possible consequence is drawn. Finally, the factors are depicted as barriers to the consequence. This design approach is used for positive impacts, where barriers reflect the controls that reduce the adverse impacts.

Although the method combines the investigation of the causes of the event using a fault tree and the analysis of the consequences using an event tree, its main focus is on the barriers between the causes and hazardous events, hazardous events, and consequences. The bow-tie method provides a visual, simple, and clear graphical representation of the problem, and is focused on the controls to prevent and reduce the consequences of hazardous events and evaluate their effectiveness.

One of the best tools for assessing the decision adequacy is simulation modeling, which has already been proven to be an effective tool for road transport expertise [28]. In the case of risk management, it allows one to simulate the operation of a real system with different parameter values and study options for managing the traffic situation in the “what-if” format. At the same time, the micro-simulation method, in which the vehicle is the modeling unit, has proven its effectiveness. This makes it possible to take into account the individual characteristics of the traffic flow units.

To implement continuous control over the risks based on feedback, we developed an algorithm, shown in Figure 1. It includes risk monitoring, timely response to changes in the system, and risk management efficiency assessment. To identify significant factors influencing the accident rate and the consequence severity, we used one-way and multi-factor analysis of variance using Fisher’s F-test and the determination coefficient. The proposed algorithm makes it possible to evaluate the decisions’ effectiveness in the traffic safety management system by analyzing the impact of adopted laws, regulations, and various preventive measures to reduce the accident likelihood and the consequence severity.

4. Research Materials

We chose the town of Yelabuga (Figure 2) as the object of this study and is the administrative center of the Yelabuga region. It belongs to a medium-sized town with a population of 74,181 people. People of working age predominate. There is a positive dynamics of the city population, however, over the past 5 years, the number of Yelabuga region permanent residents has decreased. Therefore, there are prospects for the development of the town transport infrastructure.

The town is characterized by a high the road network density. Currently, there is no cycling infrastructure. A feature of the city’s transport system is the organization of one-way traffic on three major highways. This helps to increase throughput as well as to eliminate the conflict of oncoming traffic flows with an insufficient roadway width. This is because the Elabuga town has a thousand-year history, which leaves an imprint on its architecture and the impossibility of expanding the roadway. For the analysis, we used accident statistics, collected by the traffic police for Yelabuga town for 2017–2019, and located in the public domain on the official website [30]. Along with the descriptive statistics and analysis of variance, we applied machine learning methods: decision trees and association rules. We used the Deductor Studio analytical platform.

5. Research Results

5.1. Risk Identification

Since the increase in the safety level consists of identifying the factors and causes that affect the accident likelihood and the consequence severity, it is necessary to identify the most critical factors. Through the processing of statistical data (the results of which are described in our previous works [31,32,33,34,35,36]), we obtained such factors and have presented them in Figure 3 as a factors tree.

5.2. Risk Analysis and Assessment

In our study, when developing a system for increasing road safety, we analyzed and evaluated the risk levels. Since all factors are gathered by traffic police officers only for traffic accidents with victims, we decided not to limit ourselves only to the results of the numerical analysis, but also to obtain an expert opinion. We presented the results of the processing of the statistical data to an expert group. This consisted of people involved in the road accident expertise. They formulated the types of risks (Table 3, column Risk) and possible consequences (Table 3, column Risk consequences) and further evaluated each risk occurrence probability (Table 3, column Probability) in accordance with the values in

Table 1. Assessment of the risk probability.

Risk Probability	Highly Unlikely	Unlikely	Maybe	Likely	Very likely
Probability score	1	2	3	4	5

, and each event consequence severity (Table 3, column Consequences severity) in

Table 2. Assessment of the consequence severity.

The Consequence Severity	Without Consequences	Small	Significant	Critical	Catastrophic
Severity assessment	1	2	3	4	5

. We assessed the degree of expert agreement using the concordance coefficient and Pearson’s chi-squared test.

The column Risk level in

Table 3. Summary table of the risks that must be taken into account to increase traffic safety.

№	Risk	Risk Consequences	Probability	Consequences Severity	Risk Level	Way of Influence
Driver/pedestrian
1.	Violation of traffic rules by the driver/pedestrian	Danger of an accident	5	5	25	Availability of automatic photo-video recording of traffic violations
		Decreased road safety		5	25	Compliance with traffic rules
						Tougher penalties for non-compliance with traffic rules
2.	Driver category (nonprofessional, professional)	Danger of an accident	3	3	9	Driver training
		Violation of traffic rules		3	9
3.	Age/driving experience of the driver/pedestrian	Danger of an accident	2	4	8	Implementation of an e-learning system for drivers with the most frequent accidents
		Inexperience		3	6	Preventive work with pedestrians who most often violate traffic rules
4.	Poor psychological/physical condition of the road user	Danger of an accident	3	4	12	Psychological state control
		Conflict situation on the road		5	15	Prohibition of driving in a state of deep fatigue
5.	The degree of alcohol or drug intoxication of the driver/pedestrian	Danger of an accident	4	5	20	Alcohol and drug control
		Decreased road safety		5	20	Deprivation of a driver’s license for driving under the influence of alcohol or drugs, up to and including imprisonment
6.	Social status/intelligence level of the driver/pedestrian	Danger of an accident	2	2	4	Development of the driving culture/pedestrian behavior level
		Conflict situation on the road		3	6	Development of the intelligence level (mandatory passing of IQ tests)
Vehicle
7.	Faulty vehicle technical condition	Danger of an accident	4	5	20	Monitoring vehicle technical condition
		Threat to the life and health of road users		5	20	Increase in penalties for the faulty condition of the vehicle
						Mandatory maintenance of the vehicle
8.	Lack of active and passive safety systems	Danger of an accident	5	5	25	Improvement of active and passive safety systems
		Threat to the road users life and health		5	25	Strict control of the use of active and passive safety systems
9.	Lack of ADAS	Danger of an accident	4	3	12	Mandatory availability of ADAS for each vehicle
		Threat to the road users life and health		3	12
10.	Speed characteristics	Danger of an accident	4	4	16	Control of speed characteristics
Infrastructure
11.	Unsatisfactory condition of the roadway	Danger of an accident	3	4	16	Mandatory control of the road surface condition
		Travel time reduction		3	9
12.	Incorrect location of road network objects	Places of road accidents concentration	4	4	16	Planning the structure of the road network in accordance with regulatory documents
		Decreased vehicle throughput		3	12
13.	Disadvantages of the transport and operational state of the road network	Danger of an accident	2	3	6	Maintain optimal infrastructure condition
		Decreased road safety		2	4
14.	Unregulated pedestrian crossing	Danger of an accident	5	5	25	Additional lighting of pedestrian crossings and approaches to them
		Decreased road safety		5	25	Combination of unregulated pedestrian crossings with artificial road bumps
				5	25	Installation of duplicate road signs “Pedestrian crossing” over the carriageway with LED backlight
15.	Unregulated intersection of unequal streets (roads)	Danger of an accident	5	5	25	Installation of traffic light regulation
		Decreased road safety		5	25	Correction of traffic lights phases
		Places of road accidents concentration		5	25	ADAS based on onboard information systems
Information technology
16.	Not using information technology for decision making	Lack of automatic recording of the accidents number	3	3	9	Introduction of modern information technologies
		Difficulties in identifying places of road accidents concentration		4	12
		Problems with identifying the most frequent accidents types		3	9
		Low efficiency in detecting persistent traffic offenders		3	9
17.	Insufficient investment in innovation and IT	Low process efficiency due to outdated technology	2	2	4	Improvement of information policy
18.	Unauthorized access and damage to information	Lack of reliable analysis of traffic accident statistics	1	3	3	Security Boost
						Threat blocking
19.	Lack of widespread use of intelligent transport systems	Lack of intelligent traffic planning	3	2	6	Implementation of intelligent transport systems
		Lack of development and popularization of public transport		2	6
		Low communication between road users in the city infrastructure		2	6
Traffic flow
20.	Increasing the density and intensity of the traffic flow	Danger of an accident	4	3	12	Maintain optimal infrastructure conditions
		Negative emissions of pollutants from exhaust gases into the environment		3	12
21.	Insufficiency of speed regulation	Danger of an accident	4	5	20	Competent placement of traffic signs, installation of traffic lights at unregulated intersections
		Decreased road safety		5	20	Increasing the penalties for speeding; availability of photo and video recording
Environment, ecology
22.	Bad weather conditions	Danger of an accident	3	5	15	Timely clearing of snow drifts on the roads
23.	Negative impact on the environment	Deterioration of the ecological situation	4	3	12	Increase throughput
						Decrease in vehicle mileage
						Maintain optimal infrastructure conditions

is defined as the product of the risk probability and the consequence severity. Since some risks have several consequences, the severity of which is different, we took the maximum value as the final assessment of the risk level.

As can be seen from the summary table, in order to reduce the consequences of the most probable and dangerous risks, it is necessary to introduce some significant measures (Table 3, column Way of influence) that are related, first of all, to the prevention of traffic rule violations, strict control of the state of active and passive safety systems, both vehicles and the street-road network including the necessary infrastructural changes for unregulated pedestrian crossings and the intersections of unequal streets (roads).

5.3. Risk Treatment

To build the bow-tie diagram, we selected the risks from the summary in Table 3 that are highly probable and lead to severe consequences (Figure 4). On the left side of the diagram, we show the event causes that can be avoided by using active safety measures (before the accident). On the right side, we show the consequences, the severity of which can be minimized by applying passive safety measures (after the accident). Barriers between the causes and dangerous events, dangerous events and consequences in our case were active and passive safety measures.

An analysis of the statistical data of traffic accidents in Yelabuga town made it possible to choose one of the most emergency intersections—the T-shaped intersection of Neftchilar Avenue, Okruzhnoye and Tanaevskoe highways, which have a different number of traffic lanes in each direction (Figure 5). From the Tanaevsky highway, there is also an exit to a tire store and a gas station. This intersection is unregulated. Since the intersection has a different number of lanes, and there is no traffic signal regulation, traffic jams and conflict situations arise and traffic participants create obstacles for each other, so the probability of an accident and the time of crossing the intersection increase.

Traffic intensity is often considered as one of the most important factors that can affect the number and severity of road traffic accidents. Network technologies and the development of Internet platforms (remote work practices, online trading platforms, distance learning technologies) contribute to a decrease in traffic. However, in order for online platforms to be able to replace regular trips to work and school, good communication quality is necessary. If we talk about the organization of a single information interaction between infrastructure, vehicles, and pedestrians, then the tools and technologies that provide them, due to their high cost, will be implemented primarily in megacities, where mobility issues are more significant. The conducted studies [31,32,33,34,35,36] have shown that in order to reduce the accident likelihood in small- and medium-sized towns, measures to improve the road infrastructure and management efficiency are needed. We found that a significant part of the accident was associated with the incorrect vehicle location on the roadway, therefore, control measures in the form of installing photo and video recording cameras are ineffective, so it is necessary to look for alternative low-cost measures to expand the parking space.

Since the budgets of small- and medium-sized towns are small, organizational and managerial measures to optimize the existing infrastructure parameters such as adjusting the traffic lights phases can be recommended as an effective tool.

5.4. Development of Risk Management Measures

For this problematic section of the Yelabuga town road network, we proposed the introduction of traffic light regulations. For this object, we developed a simulation model (SM).

To build the model and conduct optimization experiments, we used the AnyLogic software package, which has a built-in traffic library. The road network section the SM structure and its animation before the changes are shown in Figure 6.

When conducting the computer experiments, realized with in-building the OptQuest optimizer, we determined rational traffic control parameters (the number of phases and their duration), ensuring the minimum travel time for the vehicle section. The road network section SM animation after the changes and the results of the optimization experiment are shown in Figure 7.

5.5. Implementation of Continuous Control

The advantage of the SM is also that it can be used for a preliminary assessment of the proposed solution effectiveness. With the optimal parameters obtained as a result of the SM run for a three-phase control cycle, we found that traffic jams at this intersection do not practically form (Figure 7a), thus the number of conflict situations, and, accordingly, the accident likelihood are reduced. This will probably increase the level of traffic safety in this city.

6. Conclusions

This paper presents a risk management methodology to ensure the security of the transport infrastructure. Risk analysis when implementing a traffic safety management system includes the identification, risk assessment, risk treatment, development of risk management measures, and risk control. We developed a risk management algorithm, the key feature of it is the system feedback. We also proposed methods that are appropriate to use at each stage of the risk management methodology. We tested the adequacy on the example of the small town of Yelabuga. We identified the factors influencing the road accident frequency and severity and assessed their probability and consequence severity. For the most critical of the identified risks, we carried out a more detailed analysis to formulate active and passive safety measures. We found that in order to reduce the likelihood of the most dangerous risks, it is necessary to introduce comprehensive measures related to the prevention of traffic violations, the strict control of active and passive safety systems of both the vehicle and the road network including the necessary infrastructural changes for unregulated pedestrian crossings and intersections of unequal streets (roads). We demonstrated that simulation modeling is an effective tool for developing measures to optimize the transport infrastructure. It is also a tool for the fifth step—preliminary control. This is especially relevant, since the transport system refers to the types of systems, experiments with which are potentially associated with the possibility of a threat to human life and health. The developed simulation model showed a decrease in the number of congestion with the simultaneous achievement of a minimum target indicator—the time it takes a vehicle to travel through the road network section. Further research will be related to the development of an automated risk management system based on the algorithm proposed by the author.