† Corresponding author. E-mail:
Project supported by the National Natural Science Foundation of China (Grant No. 71473207) and China Fundamental Research Funds for Central Universities (Grant No. 2682016cx082).
Crowd force by the pushing or crushing of people has resulted in a number of accidents in recent decades. The aftermath investigations have shown that the physical interaction of a highly competitive crowd could produce dangerous pressure up to 4500 N/m, which leads to compressive asphyxia or even death. In this paper, a numerical model based on discrete element method (DEM) as referenced from granular flow was proposed to model the evacuation process of a group of highly competitive people, in which the movement of people follows Newton’s second law and the body deformation due to compression follows Hertz contact model. The study shows that the clogs occur periodically and flow rate fluctuates greatly if all people strive to pass through a narrow exit at high enough desired velocity. Two types of contact forces acting on people are studied. The first one, i.e., vector contact force, accounts for the movement of the people following Newton’s second law. The second one, i.e., scale contact force, accounts for the physical deformation of the human body following the contact law. Simulation shows that the forces chain in crowd flow is turbulent and fragile. A few narrow zones with intense forces are observed in the force field, which is similar to the strain localization observed in granular flow. The force acting on a person could be as high as 4500 N due to force localization, which may be the root cause of compressive asphyxia of people in many crowd incidents.
Major events, such as the Olympic Games, the World Expo, rock concerts, and large religious events, etc., can attract more than 5 × 105 people. A proper crowd control and management strategy is essential for the safety of the people. In general conditions, a large population gathering should pose no serious problem. However, a combination of inadequate facilities and deficient crowd management may cause chaos or even catastrophe. Crowd disasters happen in different areas across the globe almost every year and take approximately 2000 lives per year on average.[1,2] These crowd accidents within recent decades have evoked a lot of scientific interest in the study of crowd flow. Crowd simulation models included macroscopic models, cellular automata, social force models, velocity-based models, continuum models, hybrid models, behavioral models, and network models. Macroscopic models describe the movement of pedestrians at a high level of aggregation as flow, whilst microscopic models describe the behavior of each individual and their interactions in detail. Lachapelle and Wolfram[3] presented a crowd model based on the mean field games theory. This macroscopic approach is based on a microscopic model, which considers smart pedestrians who rationally interact and anticipate the future. Based on concepts from the kinetic theory for active particles Dogbe[4] developed a model, in which pedestrians are regarded as entities whose microscopic states include a geometric variable, a mechanical variable, and a variable to describe the rate of development of individual strategies that influence the collective dynamics. Zhang et al.[5] numerically studied the pedestrian push-force in evacuation with arched congestion before an exit based on cell automata.
Helbing et al.[6–9] proposed a social force model and the movement of people is simplified as a kind of self-driven particle in which the driving force is not external, but from each particle itself and used a computer model to quantitatively explain the behavior of panicking individuals. They created a generalized force model using Newton’s formula, force equals mass multiplied by acceleration. The social force model has been widely adopted to study the evacuation behavior. In Refs. [9] and [10] based on suitable video recordings of interactive pedestrian motion and improved tracking software, the authors proposed an evolutionary optimization algorithm to determine optimal parameter specifications for the social force model. The calibrated model is then used for large-scale pedestrian simulations of evacuation scenarios, pilgrimage, and urban environments. They found the phenomenon of intermittent flows and stop-and-go wave in crowd flow at extreme high densities. This crowd dynamics is dangerous and it may cause people to fall. It is noted that the reaction of pedestrians to what happens in front of them is much stronger than to what happens behind them and a centrifugal social force model[11,12] was proposed by taking into account the distance between pedestrians as well as their relative velocities. Wang et al.[13] improved the social force model by considering the impacts of interaction among companions and further developed a comprehensive model by combining that with a multi-exit utility function. Guo et al.[14] extended the original cost potential field cellular automata to study more general evacuation scenarios. Tang et al.[15] adopted a cellular automaton (CA) model to study the passengers’ motion at the hall of HSR station during the check-in process.
Aftermath investigations have shown that the physical interactions in the crowd add up and cause dangerous pressures up to 4500 N/m, which leads to compressive asphyxia of people. Death occurred 15 seconds after a load of 6227 N was applied or 4 min–6 min after a load of 1112 N was applied;[16] a force of 3 × 103 N can break the ribs of a human from medical bone fracture analysis.[17] The interaction forces of human bodies play a dominant role in resulting in the death of people in these accidents. However, little work has been conducted on how the forces acting on people spread in the crowd flow and why such deadly forces arise. Lee and Hughes[18] studied data on accidents in human crowds that occurred over the past decade showing that one of the fatal consequences of crowding was asphyxia, that is by crushing, by utilizing the standard forward–backward autoregressive modeling techniques for spectral analysis of a measured signal to predict pressures generated by very high densities of pedestrians. It is difficult to directly measure the contact forces in crowd flow, particular in extremely high densities. Discrete element method (DEM)[19–22] is a numerical method for computing the motion and effect of a large number of particles. It can provide the local information (particle position and velocity, inter-particle contact forces) needed to investigate the relationship between microscopic and macroscopic behavior. The forces acting on each particle are calculated and force balance is integrated explicitly and acceleration velocity, velocity, and the coordinate at each time step are deduced accordingly by applying Newton’s second law. Lin et al.[20,23] proposed a discrete element method to study the crowd through an exit at different desired velocities and found that clogging occurs more easily and the exit may be totally blocked (i.e., deadlock situation) when the desired velocity is high enough. Lin et al.[24,25] also conducted a series of experiments by using mice driven by a varying number of joss sticks and the experiment found that the escape times significantly increased with the increased levels of stimulus.
Discrete element method is adopted to study the contact force in crowd flow passing through a bottleneck. This paper is structured as follows. The next section details the formulation of the proposed DEM model. In Section
In DEM simulation,[17–20] a granular material is modeled based on a finite number of discrete, semi-rigid, spherical particles interacting by means of contact or non-contact forces, and the translational and rotational motions of every single particle are described by Newton’s laws of motion. Lin et al.[20,23] proposed a discrete element method (DEM) to study the crowd through an exit. In DEM, the human body is modeled based on a finite number of discrete, soft, spherical particles interacting by means of contact or non-contact forces. The differences of the proposed DEM model with the social force[6–9] are in the three aspects: (i) Hertz contact model is incorporated and the normal push-back force for two overlapping particles is proportional to the area of the overlapped two particles, and is thus a nonlinear function of overlapped distance; (ii) an anisotropic social force model is proposed; and (iii) both translational movement and rotational movement are considered. A detailed description of the proposed DEM model is included in Refs. [20] and [23].
The translational and rotational motions of each person are described by Newton’s laws of motion. The dynamical movement of people is time-dependent and can be described as follows:
The sum of force Fi is expressed as:
The sum of the contact forces which act on a person and lead to the deformation of his body is called the composition of contact forces. The composition of contact forces, coupling the self-driven force and the social force, provides the particle with the acceleration velocity following Newton’s second law of motion. It is a vector quantity with a magnitude and a direction. The composition contact force on each person by its neighboring people due to physical deformation is expressed as:
The scalar contact forces, i.e., the sum of absolute value of the contact forces acting on a person i in all directions, is defined. The scalar contact forces represent the degree of body deformation due to external forces and it is expressed as:
A comparison of the two contact forces is shown in Fig.
Studies are conducted for 300 people getting out of a room with an exit of 1-m wide as shown in Fig.
The flow is continuous in the first 50 seconds and it clogs from 50 s to 120 s and the throughput is zero. From 120 seconds, the flow resumes until all the people get out. The flow rate is shown in Fig.
The vector contact force field at different time is shown in Fig.
The strain localization due to nonhomogeneous deformation of elastoplastic materials is observed on a wide class of engineering material in soils and rocks. In this study, the particles of movement are the self-driven people with a desired velocity of 1.5 m/s. The actual velocity ranges from 0 m/s to 0.1 m/s as observed in the simulation. The self-driven force of a person is around 180 N∼ 240 N, i.e., approximately 3 times the mass of the person. The forces from an assembly of people are accumulated and concentrated through the interactions of particles, and a spike force of 1500 N is observed at critical zones, where people should endure higher pressure and are more vulnerable to compressive asphyxia.
The desired velocity is an important parameter to represent the mental state of the crowd during an evacuation. Further study is conducted for 300 people getting out at a higher desired velocity, i.e., 2.5 m/s, which represents an evacuation of high competition. The increase in the desired velocity leads to stronger interactions of human bodies. Figures
Figure
The contact forces acting on an individual person during evacuation process at a desired velocity of 2.5 m/s are shown in Fig.
In this work, we study crowd flow by using the theory from granular dynamic. The human body is treated as a granular particle and dynamical movement of people is studied by applying Newton’s second law. Hertz’s contact model is incorporated to describe the forces among human bodies when they are physically contacted. The velocity of desired (VOD) represents the mental state of crowd. The larger the VOD, the higher degree of aggression to move forward, which results in the flow rate fluctuating greatly and the total freezing of the crowd at the bottleneck.
Force localization due to nonhomogeneous deformation of elastoplastic materials is observed on a wide class of engineering material in soils and rocks. For the self-driven human bodies with high densities, forces are predominately the result of bodies’ contact mechanics and the contact force is inherently fragile and susceptible to reorganization. The force is localized to narrow domains called force localization. People at the zone of high force endure a higher pressure and are more prone to compressive asphyxia if the pressure continually lasts for a few minutes.
At a desired velocity of 1.5 m/s, a spike force of 1000 N could be incurred. At a desired velocity of 2.5 m/s, a spike of 4500 N could be incurred. It is noted that a desired velocity of 2.5 m/s is only slightly higher than normal walking velocity and it can hardly be defined as a panic velocity. However, it could produce an alarming force of 4500 N and cause people to lose consciousness or be compressed to death. Clearly, the nonhomogeneous deformation of an assembly of human bodies leads to the force localization and concentration, which results in a number of critical zones with extremely high force.
After the study of the crowd incident in the love parade on 24th July 2010 in Duisburg, Germany, Helbing[9,10] concluded that there was no sign of panic in the crowd flow during the incident and the crowd turbulence was the cause of the loss of 21 lives in the tragedy. The trivial behaviors of people interactions, e.g., rushing, pushing, could lead to the generation of turbulent flow, whilst the force localization leads to forces at some areas being uncontrollable and unbearable. The force localization could be the root cause of compressive asphyxia of people in many crowd incidents. Further study of the characteristics of the contact forces in the crowd flow (e.g., the magnitudes and their frequency) will be conducted and further compared with the results in gravity-driven silo flow.
[1] | |
[2] | |
[3] | |
[4] | |
[5] | |
[6] | |
[7] | |
[8] | |
[9] | |
[10] | |
[11] | |
[12] | |
[13] | |
[14] | |
[15] | |
[16] | |
[17] | |
[18] | |
[19] | |
[20] | |
[21] | |
[22] | |
[23] | |
[24] | |
[25] |