Ontology-based inference decision support system for emergency response in tunnel vehicle accidents

Abstract

Emergency response plans for tunnel vehicle accidents are crucial to ensure human safety, protect critical infrastructure, and maintain the smooth operation of transportation networks. However, many decision-support systems for emergency responses still rely significantly on predefined response strategies, which may not be sufficiently flexible to manage unexpected or complex incidents. Moreover, existing systems may lack the ability to effectively respond effectively to the impact different emergency scenarios and responses. In this study, semantic web technologies were used to construct a digital decision-support system for emergency responses to tunnel vehicle accidents. A basic digital framework was developed by analysing the knowledge system of the tunnel emergency response, examining its critical elements and intrinsic relationships, and mapping it to the ontology. In addition, the strategies of previous pre-plans are summarised and transformed into semantic rules. Finally, different accident scenarios were modelled to validate the effectiveness of the developed emergency response system.

Highlights

• •Established an ontology for emergency response in tunnel vehicle accident.
• •Digitization of tunnel vehicle accident emergency response rules.
• •Implemented rule-based self-inference output for dynamic accident scenarios.
• •Improved efficiency compared to paper-based emergency response plans.

1. Introduction

Modern transportation networks heavily rely on seamless operation of tunnels to facilitate the rapid passage of vehicles. In emergency situations, such as accidents, natural disasters, and other unforeseen events, it is critical for decision makers to quickly and accurately formulate sound emergency response strategies. This aspect plays a key role in enhancing public safety, reducing environmental impacts, and maintaining the overall functionality of transport infrastructure. Bjelland et al. [1] suggested that the emergency response to a road tunnel emergency is a complex task that requires a systems-theoretical approach to the complexity of the tunnel system from the design stage to the actual emergency. Several researchers have investigated emergency decision making for tunnel emergencies at different levels, including the quantitative evaluation of accident risks [[2], [3], [4]], summarisation of disaster evolution patterns [[5], [6], [7], [8]], and proposals of management and control strategies [[9], [10], [11], [12]]. Despite the significant progress in the development of advanced control technologies, few innovations have been effectively implemented in emergency response plans.

Currently, emergency response strategies employed by operating units are mainly based on traditional fixed response plans. The conventional emergency response plan includes the development of response strategies for electromechanical (E&M) facilities and emergency responders. Borghetti et al. [13] focused on 15 interventions implemented on some Italian motorway tunnels between 2019 and 2021, assessing road tunnel risk in terms of the effectiveness of the emergency team response. In the event of an emergency, tunnel operation and maintenance managers as the first echelon of response in the actual emergency response, generally in accordance with the emergency response system prompt information, to achieve a one-click electromechanical emergency response to the incident, to achieve effective control of emergencies. However, the preparation of an emergency response plan is an elaborate task, particularly for E&M response in tunnels. This requires the separate development of appropriate E&M response strategies for each control unit of the tunnel and application to the operation and maintenance (O&M) control platform of the tunnel in the form of a plan card during the actual operation [[14], [15], [16]]. Moreover, revisions and updates to paper plans can be cumbersome and time-consuming. It is often necessary to manually distribute new versions, which can lead to potential delays in implementing critical changes.

With the introduction of the concept of intelligent tunnel operation and maintenance [[17], [18], [19]], tunnel operations are progressively shifting towards the application of advanced technologies and intelligent systems to improve the efficiency, safety, and overall performance of tunnel infrastructure. This includes accessing real-time tunnel environment data using the Internet of Things (IoT) technology and sensing the tunnel environment to support tunnel O&M decision making [20,21]. In terms of IoT technology applications, many researchers have combined IoT technology with the Semantic Web to achieve multi-source heterogeneous data representation, interpretation, and utilisation. By applying Semantic Web principles such as linked data and ontological modelling to IoT ecosystems, it is possible to derive nuanced insights, enhance data interoperability, and facilitate smart decision making in various domains, ranging from smart cities to industrial automation [22]. This makes it possible to digitise tunnel emergency plans. For the tunnel scenarios, Khademi et al. [23] proposed an architecture for a decision support system in the dynamic environment of tunnels using artificial intelligence, sensor fusion, and communication technologies. Hence, with the advancement of smart tunnelling technology, the disadvantage of poor targeting of fixed protocol-based decision making can be balanced.

This study is focused on intelligent emergency response strategies for systems at all levels. An information semantic web was used as a carrier based on ontological reasoning systems to encapsulate generic electromechanical responses and personnel response strategies in a flexible and dynamic framework. The proposed framework allows for the automatic generation of response actions based on the conditions that characterise the accident scene.

The remainder of this article is structured as follows. Section 2 presents the state of the art related to emergency responses in tunnels. Section 3 provides specific details of this study, including the research framework, generic response knowledge system for tunnels, methodology for mapping the ontology to the emergency response system for tunnels, and semanticisation of response rules. In Section 4, the reasoning capability of the decision support system is verified by analysing an example of a tunnel accident.

2. State of the art

At the most basic level, applications of the Semantic Web require the support of ontologies, which provide the conceptual framework and structure required to realise the vision of the Semantic Web, in which data are not only linked but also have a semantic meaning. The use of ontologies contributes to the creation of smart, connected, and machine-understandable web environments. In the field of decision control, building ontology decision-support systems have been investigated to enhance decision-making processes. Hu et al. [24] constructed an automatic diagnostic system for structural defects in tunnels by converting IFC files and items in a database into a web ontology language to achieve a unified representation of the information in the building information model and database. Cui et al. [25] established a comprehensive evaluation framework for the seismic risk of metro stations based on Monte Carlo simulations and ontology theory, considering economic losses, rescue costs, casualties, and post-disaster recovery time. Based on the ontology, Peng et al. [26] developed a decision-support methodology for operational hazard management of public tunnels. The above studies addressed scenarios for the daily operation and maintenance of tunnel facilities; however, ontology-based architectures for decision-support systems in emergencies during tunnel operations have not yet been covered by relevant studies.

For other emergency scenarios, researchers have proposed ontology-based decision-making systems for emergency responses that validate data-driven generation for dynamic disaster scenarios. Jiang et al. [27] focused on building fire scenarios and fused static building geometric information and dynamic sensory data into a building fire protection ontology. They merged them into a digital twin data model and developed rule and process models to achieve intelligent control. Ge et al. [28] proposed a generic DPKG-based disaster prediction method for data-driven dynamic prediction of disasters in complex disaster environments, such as forest fires and geological landslide risks. They achieved this by constructing a public semantic ontology of disasters and a unified spatio-temporal framework benchmark, integrating remotely sensed information with geographic information, and correlating dynamic data with static knowledge. Guyo et al. [29] established an ontology model that defined the building and environmental data required by firefighters during a building fire emergency. Khalid et al. [30] proposed an intelligent building evacuation ontology in the case of an emergency, considering three perspectives: the user, building, and environment. To effectively manage flood disasters, Daher et al. [31] integrated flood-related heterogeneous data using an ontology and constructed a knowledge map for flood disaster management.

Current research on emergency response strategies for tunnel emergencies remains predominantly at the level of the independent response of each system, such as the early warning and prediction of accidents [[32], [33], [34], [35]], disaster control [36,37], evacuation of personnel [11,38], and on-site rescue [10,39,40]. There is a lack of research on integrating the equipment in a tunnel, coordinating it into a coherent response strategy, and aggregating it into an intelligent system. The purpose of this study was to establish a decision-support system for tunnel vehicle accident emergency responses based on ontological reasoning. The ontology-based inference approach facilitates seamless integration of diverse data sources, enhancing the overall interoperability of the emergency response system. This ensures that all pieces of relevant information are considered, facilitating effective decision making.

3. Materials and methods

3.1. Theoretical framework

Fig. 1 shows the framework of the self-inference decision-making system of the vehicle accident emergency response in tunnels. The basic layer of the framework consisted of four aspects: the tunnel vehicle accident, tunnel entity, tunnel electromechanical system, and emergency response organisation. These four aspects are integrated into knowledge ontology for emergency responses to tunnel vehicle accidents. Based on the ontology and dynamic information of actual vehicle accidents, it is fused with the rule model defining the situation assessment, electromechanical strategy, and organisational response to achieve intelligent control and decision making oriented to complex accident scenarios.

Fig. 1.

Self-inference decision-making system for emergency response to tunnel vehicle accidents.

In addition, these four aspects of the data for actual tunnels were stored in the data layer. It is also necessary to identify and collect relevant knowledge and basic data for the emergency response of tunnel vehicle accidents from the literature and relevant planning guidelines, and to analyse the response strategies according to the multi-step emergency response process of receiving, confirming, and responding to an alarm. The ontology of the tunnel accident emergency response was constructed based on the owlready2 development package in Python [41], and the validity of the ontology construction was verified by interacting with the Protégé software [42]. Furthermore, based on the knowledge ontology of emergency response to tunnel vehicle accidents, the semantic web rule language (SWRL) was used to develop semantic rules for situational assessment, electromechanical, and personnel responses involved in the control and handling process of tunnel vehicle accidents. In practical applications, responders can enter information about tunnel vehicle accidents through the platform to represent accident scenarios. Through the input of basic event information and the inference of response rules, the system can automatically generate mechanical and E&M facility response strategies for different scenarios and recommend evacuation and rescue routes.

3.2. Tunnel emergency response ontology

The ontology is developed using a seven-step approach [43,44]. It is necessary to separate the elements of vehicle accidents, tunnel entities, E&M systems, and emergency response organisation, list and define key concepts, and obtain the domain vocabulary related to the emergency response to tunnel vehicle accidents. The first task in establishing the ontology is to sort the elements oriented toward vehicle accidents, tunnel entities, E&M systems, and emergency organisations and define the key classes. Next, a top-down approach [45] is used to organise the identified concepts into a hierarchical structure. Furthermore, the relationship between two classes is characterised by defining the object property, and the data characteristics of a single class can be expressed qualitatively and quantitatively based on the data properties [46].

Fig. 2 shows a schematic of the classes and attributes of the tunnel emergency response ontology for vehicle accidents (TVAOnto), which consists of seven classes: ‘tunnel entity’, ‘device’, ‘vehicle accidents’, ‘degree of vehicle accident risk’, ‘tunnel manager’, ‘response organisation’, and ‘response’. ‘Tunnel entity’ represents the physical space of the actual tunnel, such as the left and right tunnel caverns, entrances, crossing passages, carriageways, and control units, as a spatial dimensional representation of the accident site and the distributions of mechanical and electrical installations. The tunnel is equipped with different sizes of electromechanical ‘device’ based on its characteristics, such as the tunnel video monitoring, lighting, ventilation, fire alarm, and broadcasting systems. During the process of handling vehicle accidents in tunnels, the electromechanical systems perform ‘response’, release information regarding the accidents, and guide the vehicles and passengers inside and outside the tunnels in an orderly manner to achieve the evacuation. The ‘degree of vehicle accident risk’ reflects the case of the ‘vehicle accident’, such as the type of incident, time and space, and the severity of the incident to provide basic reference information for decision making by responders. ‘Emergency organisation’ is mainly focused on the unit in charge of the incident at the tunnel. When a vehicle accident occurs, various emergency response groups perform ‘response’ work in a reasonable manner, depending on the emergency organisation and the division of responsibilities. These jobs are based on the principle of the hierarchical response of the primary unit, secondary unit, and unit in charge of the tunnel where the incident occurs.

Fig. 2.

Concept relationship in TVAOnto.

3.3. Response rules based on incident information

3.3.1. Accident situational assessment

In emergency response management, decision makers are tasked with the critical responsibility of swiftly gathering relevant information relating to the site of a vehicle accident for a comprehensive assessment. This process requires a meticulous examination to determine the precise type of incident, such as collision, fire outbreak, chemical spillage, or any other form of disaster within the tunnel infrastructure. Furthermore, it is crucial to ascertain the exact location and scope of an accident within a tunnel network and meticulously analyse the temporal dimension to gauge the elapsed time since its occurrence. Moreover, assessments of the number and conditions of the vehicles and casualties are paramount for formulating an effective response strategy. Traditionally, accident severity has been categorised based on consequences; however, this approach may not always be suitable for emergency responses. Instead, the assessments should focus on real-time accident information. For vehicle accidents, we classified the types of accidents into vehicle accidents, fire accidents, and dangerous goods fires. In this study, the situation assessment rules for vehicle accidents and fires were developed using SWRL, with a focus on the rescue ability and quality of personnel in the operating unit.

The first step is to define the degree of vehicle accident risk based on the location, number of vehicles, number of injuries, and number of deaths. For example, when a vehicle accident occurs in the middle of a tunnel without a fire, with the number of vehicles ≥7, 3 < number of injuries ≤5, and 1 < number of deaths ≤3, it is considered to be risk degree 2. The SWRL description of the accident classification rule is as follows.

VehicAccident(?x), fire(?x, false), hasVaLocation(?x, ?val), swrlb:contains(?val, "middle"), nb_vehicles(?x, ?nv), swrlb:greaterThanOrEqual(?nv, 7), nb_injured(?x, ?ni), swrlb:greaterThan(?ni, 3), swrlb:lessThanOrEqual(?ni, 5), nb_death(?x, ?nd), swrlb:greaterThan(?nd, 1), swrlb:lessThanOrEqual(?nd, 3) - > hasVaRiskDegree(?x, VaRisk2)

When a fire event occurs, it is converted into a fire-specific emergence plan. An example of this is as follows.

VehicAccident(?x), fire(?x, true) - > hasFireResponse(?x, ConvertedFireResponse)

In addition, with reference to 12 curves for typical tunnel fire loads defined by the Ministry of Transport of France [47], the tunnel fire load curves were converted into situation semantic rules for the heat release rate for different vehicle types. It is convenient for decision makers to query the fire scale at the current incident time and predict the scale of fire evolution. For example, when a highly combustible HGV burns and the decision maker receives a message that the vehicle has been burning for 5 min, information regarding the current heat-release rate can be obtained by inputting ‘burning time’ as 5 according to the following rule. Moreover, the decision maker can predict the evolution of the fire within 20 min if rescuers cannot arrive at the scene within 20 min. The SWRL for the evaluating the rule of the heat release rate is expressed as follows.

VehicAccident(?fa), hasFireType(?fa, HGV_10t_highlyCombustion), BurningTime(?ft, ?bt), swrlb:lessThan(?bt, 10), FireGrowthFactor(?ft, ?A), swrlb:multiply(?Q, ?A, ?bt) - > HRR(?ft, ?Q)

Furthermore, based on the degree of risk, response degree rules are set up for the managerial level of the tunnel operating unit. An SWRL example is a scene in which, in the event of a degree 1 vehicle accident, the highest response organisation is the emergency response leadership team of the primary unit.

VehicAccident(?va), hasVaRiskDegree(?va, VaRisk1) - > byResponsed(?va, EmergencyLeadingGroup1)

By categorising accidents based on on-site situational information, emergency responders and authorities can efficiently allocate resources, prioritise actions, and manage the aftermath of incidents.

3.3.2. Initial E&M response

The primary E&M response in tunnel emergencies goes beyond routine vehicle control and encompasses a comprehensive set of functionalities geared towards enhancing safety, facilitating evacuation, and supporting effective emergency response efforts. The primary E&M response strategies for tunnel emergencies involve several key aspects.

• 1)Vehicle diversion devices guide vehicles away from the affected area, preventing further congestion and ensuring the smooth flow of emergency vehicles. The response strategies are listed in Table 1.
• 2)Communication systems facilitate real-time communication and provide instructions to tunnel users and emergency responders. The response strategy is that accident tunnel broadcasting should fully respond to broadcast the accident information.
• 3)During emergencies, such as severe vehicle congestion or fire incidents, ventilation systems manage the air quality, control smoke, and create safe zones for evacuation. Generally, all jet fans in incident and non-incident tunnels blow in the positive direction.
• 4)Emergency lighting and signage are essential to safely guide individuals through tunnels, particularly in situations where visibility is compromised. When an accident occurs, the tunnel lights are fully turned on.
• 5)Surveillance and monitoring include the use of cameras, sensors, and other monitoring devices to assess the situation, identify potential hazards, and guide emergency response efforts. All closed-circuit televisions (CCTV) within the control area where an accident occurs are used for recording.

Table 1.

The response strategy of vehicle diversion devices.

Vehicle Diversion Device	Location	Response Strategy
The variable message board	In the accident tunnel inside	Sends a warning message
	In the accident tunnel entrance	Sends a warning message
The traffic lights	In the dual carriageway entrance of the accident tunnel	Turned “red light”
The variable speed limit sign	In the dual carriageway entrance of the accident tunnel	Speed limit “0”
The lane indicator	In the accident tunnel	①The lane indicator in the control area where the accident occurred: the front side “closed”
		②The upstream lane indicators of the control area where the accident occurred: the front side “red light × ”
		③The downstream lane indicators of the control area where the accident occurred: the front side “green light ↑”

3.3.3. E&M response synergy during on-site investigations

The synergy between the E&M responses enhances the overall effectiveness of investigations of a tunnel accident site, providing a safe environment and knowledge for on-site investigation teams. Key response strategies highlight the importance of E&M response synergy in tunnel accident site investigations.

• 1)For the vehicle diversion part, it is necessary to control the accident and non-accident tunnels to form a reasonable emergency lane and enable the site investigation personnel to smoothly enter the accident tunnel. The response strategies for the on-site investigation are listed in Table 2.
• 2)Well-lit environments enhance the safety of investigation teams, enabling them to assess the damage, identify potential risks, and collect data effectively. Therefore, all accident and non-accident tunnel lights are turned on.
• 3)It is also necessary to ensure the air quality in tunnels to create a safe environment for on-site investigators. Hence, jet fans are turned on in accident and non-accident tunnels.

Table 2.

The response strategy of vehicle diversion devices for on-site investigation.

Vehicle Diversion Device	Location	Response Strategy
The variable message board	In the accident tunnel inside	Sends a vehicle control message
	In the accident tunnel entrance	Sends a vehicle control message
The traffic lights	In the dual carriageway entrance of accident tunnel	Turned “red light”
	In the left carriageway entrance of the non-accident tunnel	Turned “red light”
	In the right carriageway entrance of the non-accident tunnel	Turned “green light”
The variable speed limit sign	In the dual carriageway entrance of the accident tunnel	Speed limit “0”
	In the left carriageway entrance of the non-accident tunnel	Speed limit “0”
	In the right carriageway entrance of the non-accident tunnel	Speed limit “40”
The lane indicator	In the left carriageway of the accident tunnel	The front side “red light × ”
		The back side “green light ↑”
	In the right carriageway of the accident tunnel	The front side “green light ↑”
		The back side “red light × ”
	In the non-accident tunnel	①The lane indicator in the non-accident tunnel on the back side all show “red light × ”
		②Symmetrically upstream of the accident control unit front side: “red light × ”
		③Symmetrically downstream of the accident control unit: front side “green light ↑”
		④Symmetrically of the accident control unit: front side closed

3.3.4. E&M response synergy during evacuation

The emergency evacuation of tunnels necessitates a well-coordinated E&M response to ensure swift and orderly evacuation of individuals. By following existing tunnel evacuation techniques, E&M devices can include the following components.

• 1)In-tunnel broadcasting devices can provide directive orders to persons trapped in tunnels regarding evacuation, guide them through the evacuation process, provide details of the type of emergency, and direct the individuals to safe locations.
• 2)In addition to the clear visibility provided by the lighting system, flashing reminders from indicators on cross passages help enhance evacuation and ensure visibility even under low-light conditions.

After on-site investigations, one- or two-way closure measures are implemented based on site conditions. The response strategies for the evacuation guidance are presented in Table 3.

Table 3.

The response strategy of vehicle diversion devices for evacuation guidance.

Vehicle Diversion Device	Location	Response Strategy
The variable message board	In the accident tunnel inside	Sends an evacuation guide message
	In the accident tunnel entrance	Sends a vehicle control message
The traffic lights	In the dual carriageway entrance of the accident tunnel	Turned “red light”
	In the left carriageway entrance of the non-accident tunnel	Turned “red light”
	In the right carriageway entrance of the non-accident tunnel	Turned “yellow light” (one-way closure); “red light” (two-way closure)
The variable speed limit sign	In the dual carriageway entrance of the accident tunnel	Speed limit “0”
	In the left carriageway entrance of the non-accident tunnel	Speed limit “0”
	In the right carriageway entrance of the non-accident tunnel	Speed limit “40” (one-way closure); “0” (two-way closure)
The lane indicator	In the left carriageway of the accident tunnel	(Same as on-site investigation)
	In the right carriageway of the accident tunnel
	In the non-accident tunnel
Pedestrian crossing signs	In accident Unit	flashing tip
	Upstream of the accident unit
Vehicular crossing signs	Upstream of the accident unit

4. Case of study

4.1. Instantiation of classes

4.1.1. Tunnel information

Fig. 3 illustrates an example of the tunnel. The number of tunnel caverns and lanes was 2 × 2. The tunnel must be divided into manageable subareas to optimise vehicle management during an emergency response, improve emergency preparedness, and ensure efficient tunnel operations, and targeted control measures must be implemented. Therefore, the first priority is to segment the tunnel to analyse specific vehicle organisation and control strategies based on the different classifications and locations of incidents. The tunnel is equipped with five pedestrian crossings and two vehicular crossings, and the left and right tunnels are divided into six control units bounded by vehicular crossings, as shown in Fig. 3.

Fig. 3.

Schematic diagram of the tunnel system.

4.1.2. Properties of E&M devices

It is necessary to determine the placement of E&M installations in tunnels to achieve localised and targeted response measures to improve safety, efficiency, and coordination in the event of accidents or emergencies. The positional attributes of the various types of E&M devices were defined to describe their relative positions in the tunnel (Fig. 3). Table 4 lists the locations of the individual pieces of E&M devices.

Table 4.

The location of the tunnel E&M devices.

E&M Devices	Individuals	Location in Tunnel Entity
Broadcasting	B01∼B05	Control Unit 1
	B06∼B11	Control Unit 2
	B12∼B17	Control Unit 3
	B18∼B22	Control Unit 4
	B23∼B28	Control Unit 5
	B29∼B34	Control Unit 6
CCTV	C01-C06	Control Unit 1
	C07-C14	Control Unit 2
	C15-C21	Control Unit 3
	C22-C27	Control Unit 4
	C28-C34	Control Unit 5
	C35-C41	Control Unit 6
illumination	T2∼T18	youxian
	T21∼T36	zuoxian
Jet fan	Y11F, Y12F, Y13F	youxian
	Z11F, Z12F, Z13F	zuoxian
Lane indicator	LS01A, LS02A	youxian entrance
	LS03A, LS04A	Control Unit 1
	LS05A, LS06A	1# vehicular crossing
	LS07A, LS08A	Control Unit 2
	LS09A, LS10A	2# vehicular crossing
	LS11A, LS12A	Control Unit 3
	LS13A, LS14A	youxian exit
	LS15A, LS16A	zuoxian entrance
	LS17A, LS18A	Control Unit 4
	LS19A, LS20A	2# vehicular crossing
	LS21A, LS22A	Control Unit 5
	LS23A, LS24A	1# vehicular crossing
	LS25A, LS26A	Control Unit 6
	LS27A, LS28A	zuoxian exit
Pedestrian crossing sign	PCS_Y1, PCS_Y2	Control Unit 1
	PCS_Y3	Control Unit 2
	PCS_Y4, PCS_Y5	Control Unit 3
	PCS_Z1, PCS_Z2	Control Unit 4
	PCS_Z3	Control Unit 5
	PCS_Z4, PCS_Z5	Control Unit 6
Traffic lights	Y_TL1, Y_TL2	youxian entrance
	Z_TL1, Z_TL2	zuoxian entrance
vehicular crossing sign	VCS_Y1, VCS_Z2	1# vehicular crossing
	VCS_Z1, VCS_Y2	2# vehicular crossing
Variable message board	VMB_Y1, VMB_Y2	youxian
	VMB_Z1, VMB_Z2	zuoxian
	Y_VMBset	youxian entrance
	Z_VMBset	zuoxian entrance
Variable speed limit sign	VSLS_Y1, VSLS_Y2	youxian entrance
	VSLS_Z1, VSLS_Z2	zuoxian entrance

The definition of E&M individuals and their attributes by owlreay2 are as follows.

# Create individuals of Broadcasting, and assert its property.

B01 = Broadcasting("B01", BroadcastingLocation = ControlUnit1)

B02 = Broadcasting("B02", BroadcastingLocation = ControlUnit1)

…

# Create individuals of LaneIndicator.

LS01A = LaneIndicator("LS01A″, LIUpstreamOf = [LS03A, LS04A], LILocation = [Y_Entrance1])

LS02A = LaneIndicator("LS02A″, LIUpstreamOf = [LS03A, LS04A], LILocation = [Y_Entrance2])

LS03A = LaneIndicator("LS03A″, LIUpstreamOf = [LS05A, LS06A], LILocation = [ControlUnit1])

LS04A = LaneIndicator("LS04A″, LIUpstreamOf = [LS05A, LS06A], LILocation = [ControlUnit1])

…

4.1.3. Approach and evacuation route

The focus of this study was not on push display visualisation for scenario instructions but on the efficient and effective establishment of rules. Furthermore, based on the established rule base, a pre-set emergency approach and evacuation routes can be saved in the system, as the underlying data can be saved. In addition, the instruction results for the approach and evacuation routes obtained by reasoning link this data terminal for calling. A predefined scenario for the approach and evacuation routes is as follows.

Depending on the type and location of the accident in the tunnel, the present on-site disposal group selects routes into the tunnel in order of preference and should comply with the following provisions.

A vehicle accident occurred at ‘Control Unit 1’ (Fig. 4(a)). If it is a non-fire accident, then the on-site disposal group can select ‘Approach Route 1_1’ to enter the accident tunnel in the direction of the vehicle, or ‘Approach Route 1_2’ to enter the tunnel from the non-accident tunnel via 1# vehicular crossing. In the case of a fire incident and the jet fan starts blowing downstream of the accident tunnel, the on-site disposal group enters only upstream of the accident point in the tunnel ‘Approach Route 1_1’ for safety.

Fig. 4.

Approach route with different accident points.

As shown in Fig. 4(b), an accident occurred in ‘Control Unit 2’. If the vehicle accident occurred without fire, the on-site disposal group can approach the accident point from the same direction as vehicle flow, that is, ‘Approach Route 2_1’ and ‘Approach Route 2_2’. In this case, the on-site disposal group may also approach from the opposite tunnel through the adjacent carriageway crossings into the accident tunnel, that is, ‘Approach Route 2_3’ and ‘Approach Route 2_4’. When a fire accident occurs in the tunnel and the jet fans have been turned on, ‘Approach Route 2_2’, which is downwind of the fire, is neglected, or the accident tunnel is reached from both sections of the opposite cavern through 1# vehicular crossing, that is, ‘Approach Route 2_4’ and ‘Approach Route 2_5’. ‘Approach Route 2_3’, which is downwind of the fire, is ignored.

The same principle applies to the accident in ‘Control Unit 3’ (Fig. 4(c)). ‘Approach Route 3_1’, ‘Approach Route 3_2’, ‘Approach Route 3_3’, and ‘Approach Route 3_4’ are used as the tunnel route for general vehicle accidents, and ‘Approach Route 3_2’ is ignored when a fire event occurs because of already exhaust smoke downstream of the fire point.

The general principle of safe evacuation is that vehicles downstream of the accident site should be evacuated directly from the tunnel along the carriageway. Vehicles upstream of the accident point are evacuated to a non-accident tunnel through a vehicle crossing upstream of the accident point. An accident in the right tunnel is used as an example for illustration. In this case, the accidents that occurred in ‘Control Unit 1’, ‘Control Unit 2’, and ‘Control Unit 3’ evacuation routes are shown in Fig. 5.

Fig. 5.

Evacuation route with different accident points.

4.2. Applications

4.2.1. Accident situational assessment phase

When a vehicle accident occurs in a tunnel, the road network centre or monitoring centre receives the accident alarm and converts the accident alarm information into semantic information. Based on the semantic rules of the accident situational assessment set by SWRL, the degree of risk of the intelligent evaluation of different vehicle accident scenarios can be realised based on the degree of accident risk. In addition, for fire accidents, the assessment of the fire situation can be achieved using an inference system. The Protégé software was used to verify the accuracy of the inference results for some typical vehicle accident scenarios.

Fig. 6 shows the results of accident-level reasoning for accidents at different locations and casualty scenarios. Accident Case 1 (Va1): Control Unit 2 had a serial collision of 10 cars with no vehicle fire; the number of injuries was 10, with 10 fatalities. The accident information provided could be converted into the properties of vehicle accidents, and the system reasoned by itself, as shown in Fig. 6(a). The reasoning of ‘hasVaRiskDegree’ results in ‘VaRisk1’, indicating that the degree of accident risk is degree 1, and the primary-unit emergency-leading group should respond to the accident, that is, ‘Va1’ ‘byResponsed’ ‘EmergecyLeadingGroup1’. Accident case 2 (Va3): a four-car post-collision fire in Control Unit 2 with five injuries and three fatalities at the current scene. The property of the vehicle fire accident is shown in Fig. 6(b), and the reasoning result of ‘hasVaRiskDegree’ is ‘ConvtertedFireResponse’, which indicates that the incident should be transferred to a fire special-handling unit. As shown in Fig. 6(c), accident case 3 (Va2) was a small car fire that had been burning for 1 min, and the heat release rate for the current period was 1.17 MW as the inference result. This is consistent with the rules for calculating the heat release rate mentioned in Section 3.3.1.

Fig. 6.

Accident degree and response organisation inference.

4.2.2. Incipient E&M response phase

Consider accident Case 1 as an example (Va1) to verify whether the reasoning results of the initial E&M response strategy are correct in this accident scenario. When the monitoring centre receives the accident information, it should turn on the corresponding E&M devices to take initial vehicle control. At this time, the on-site investigation and disposal work of the on-site disposal group has not commenced; hence, as in Fig. 6(a), the ‘OnsiteResponse’ is ‘false’.

The self-inference results of the initial E&M response for this accident are shown in Fig. 7, and all these inference results are consistent with the response strategy presented in Table 1. All broadcasting (B1-B17) contained in the accident tunnel (right tunnel cavern) responds to the warning message notification, whereas the non-accident tunnel (left tunnel cavern) does not have to respond. The CCTVs in ‘Control Unit 2’ (C07-C14), where the accident occurred, perform the recording action, whereas the CCTVs in the other control units are not required to perform the response. All lights (T02-T19) in the accident tunnel (right tunnel cavern) are switched on. Moreover, all fans (Y11F-Y13F, Z11F-Z13F) blow positively in the accident and non-accident tunnels. Variable message boards inside and outside the accident tunnel (right-lane cavern) transmit vehicle accident alert messages (VMB_Y1, VMB_Y2, and Y_ VMB set). The traffic lights (Y_TL1 and Y_TL2) at the entrance of the accident tunnel (right-tunnel cavern) turn red. As shown in Fig. 8, the lane indicators (LS09A-LS14A) in the downstream section of ‘Control Unit 2’ have green lights on the front side and red lights on the back side. The indicators in the upstream section of ‘Control Unit 2’ (LS01A-LS06A) have red lights on the front side and red lights on the back side. The front of the lane indicators (LS07A-LS08A) in ‘Control Unit 2’, to which the accident belongs, are turned off on the front side.

Fig. 7.

E&M response results at the incipient phase (part of).

Fig. 8.

Lane indicator response results at the incipient phase.

4.2.3. Access to site phase

Consider accident case 1 as an example (Va1). For the on-site disposal group to arrive at the accident sites on time and successfully, E&M cooperation is required, and the ‘OnsiteResponse’ in Fig. 6(a) is defined as ‘true’. The responses of broadcasting, CCTV, jet fan and lane indicators, variable message boards, and traffic lights in the accident tunnel are still performed, in addition to the activation of non-incident tunnel-related devices. Based on this accident information, the lane indicators inside the non-incident tunnel, vehicle signals outside the cave, and variable message boards perform the actions shown in Fig. 9, which satisfy the response strategy listed in Table 2.

Fig. 9.

E&M response at the approach phase (part of).

The approach routine of the on-site disposal group for vehicle and fire accidents is presented in Fig. 10. The routine satisfies the rules for the approach process mentioned in Section 4.1.3.

Fig. 10.

Approach routine inference result.

4.2.4. Evacuation induction phase

Accident case 1 is used again as an example (Va1). A new attribute for Va1 is defined to enter the evacuation phase; that is, ‘EvacuationResponse’ is defined as ‘true’. Fig. 11(a) illustrates the reasoning results of the evacuation routes under Va1 conditions, which are consistent with the evacuation strategy under Control Unit 2, as discussed in Section 4.1.3. As shown in Fig. 11(b)–11(e), the reasoning results of the E&M responses are consistent with the rules listed in Table 3. For the two control measures in particular (Fig. 11(d) and (e)), the vehicle signals for the outer carriageway of the non-accident tunnel are ‘yellow’ and the variable speed limit is ‘40’ when the one-way closure is in place. When the two-way closure is in place, the vehicle signals for the outer lanes of the non-accident tunnel are ‘red’, and the variable speed limit is indicated as ‘zero’.

Fig. 11.

Evacuation induction phase inference result.

5. Conclusions

This study revealed the significant benefits of digitally constructing a response strategy for tunnel traffic accidents. Embracing digital solutions offers several advantages, with a primary focus on reducing labour costs associated with plan preparation, as well as ensuring scalability and adaptability. These ensure that contingency plans remain dynamic and responsive to changing circumstances.

First, by structuring the response strategies into a knowledge ontology, relevant concepts, relationships, and rules are coherently organised. This structured representation enhances clarity and facilitates understanding among emergency response operators. Furthermore, the dynamic nature of the SWRL rules allows response strategies to adapt to evolving conditions and contextual factors. As new information becomes available during a tunnel traffic accident, SWRL rules can be invoked to update response plans and tailor actions based on the specific circumstances encountered, thereby optimising the resource allocation and minimising the response time. Different scenarios were used as case studies to verify the validity of the ontology and semantic rules for the practicality of the decision-making model for emergency response to tunnel vehicle accidents.

In summary, transforming E&M response strategies for tunnel traffic accidents into a knowledge ontology and SWRL empowers organisations to leverage the full potential of semantic technologies to enhance preparedness, responsiveness, and resilience during emergencies. By harnessing structured knowledge representation, automated reasoning, and collaborative knowledge sharing, tunnel operators can collectively strive for safer and more efficient tunnel operations.

The aim of this study was to investigate the feasibility and effectiveness of ontology-based reasoning in tunnel emergency response scenarios. This initial step is crucial, as it lays the basis for further developments in the field. However, the system should address several issues to bridge the entire chain, ranging from tunnel incident sensing to the analysis and decision-making output. The integration of incident information obtained from front-end sensing and intermediate data-processing systems into ontology-based decision-support systems has not been studied in detail. Future studies should focus on the specification of input forms, integration of different event data, and the development of flexible interfaces for predictive models to improve the accuracy and robustness of the system.

Data availability statement

Data will be made available on request.

CRediT authorship contribution statement

Gongyousheng Cui: Writing – original draft, Software, Methodology, Conceptualization. Yuchun Zhang: Writing – original draft, Data curation. Haowen Tao: Writing – review & editing, Supervision. Xineng Yan: Visualization, Investigation. Zihao Liu: Validation, Software.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.