Application of Computer Simulation Technology for Structure Analysis in Disaster

Automation in Construction 13 (2004) 597¨C 606

By LU Xinzheng, YANG Ning & JIANG Jianjing

Department of Civil Engineering, Tsinghua University, Beijing, 100084

Times non-self cited: 1

Abstract: With the development of mechanics and computer technology, computer simulation has become an important tool in the structure analysis and design. Especially when the structures are under disaster load such as blast, penetration, impact of collapse or typhoon, it is difficult to analyze with test method, while the advantages of computer simulation method such as safe, efficient and cheap, are shown obviously in these problems. Various simulation systems are classified and discussed in this paper generally. And four practical examples are presented to demonstrate the function of simulation system. The first one is the analysis for the safety of blast-resist doors under blast load. The second one is the study on new earth-penetration-weapons. The third one is the simulation on the collapse of World Trade Center in New York. And the forth one is the real-time display system for bridge under typhoon. These examples are used to illuminate the advantage of computer simulation technology on disaster load conditions and emphasis some problems in the simulation.

Keywords: simulation; disaster; dynamic FEA



1 Introduction

Computer has been an important analysis tool since finite element method (FEA) was developed in 1960s[1]. With the development of computer science, especially with the multimedia technology which was developed in 1990s, computer simulation now can take the place of experimental research in some aspect, or can reduce the experimental workload[2]. To those special load problems, such as blast, penetration or collapse, it is very difficult to study in the laboratory, while the advantages of computer simulation, such as safe, efficient and cheap, are shown more obviously on these problems. The damage process under those types of disaster can be recurred on the computer with proper parameter and numerical model. And the results can be displayed clearly with the computer graphic simulation. The advantages of simulation technology are discussed in this paper with several practical examples, together with the issues needed to be notice in computer simulation.

2 Principles about simulation system

Generally, the computer simulation system for structural analysis is constituted by the following four parts[3,4]:

(1) model input section;

(2) computation section;

(3) data deliver section;

(4) graphics display section

According to the relationships of these sections, the computer simulation system can be divided into two types: the separated system and the integrated system.

In separated systems, each section is mutually independent, and data are delivered between the sections with data files. Because of the independence of each section, different programming languages and software platforms for different sections are acceptable. By this way it is favorable to take the advantages of various languages and software. And it is relatively simple to deliver data with data files. Furthermore, many current structure analysis software, such as common FEA software like ANSYS or SAP, can be used directly as a part of whole system. For example, we may use these FEA software to build a numerical model of structures under special loads and process the simulation. Then develop an independent result-display program by reading the result files from the common FEA software. Similarly, we may also write a computational program firstly and output the computational results with the data format of result files in current FEA software, and use the post-process function of these software to display the computational result. Because of the above agility of the separated system, this type is widely used in computer simulation systems for structure analysis. But the data transmission is very time consumable and it is unable to provide real time simulation.

However, in integrated systems, all sections are combined together. Data deliver is carried out by building public data module, global variable or dynamic-link-library (DLL). The program design for integrated system is comparatively more complex than the separated system especially when various computer program languages are mixed. The integrated system is always applied when real-time data display is demanded.

In the separated systems, programming languages that have strong ability in computation, such as Fortran or C, are used for the computational section of the system, while some other languages such as Visual Basic, Delphi or Visual C++ are used for the display section because of their outstanding graphic interface. In the integrated systems, programming language of C is the mostly widely used language for its balance performance on computation, graphic and friendly interface between some graphic library such as OpenGL or Direct3D[5-7].

In the four examples introduced in this paper, the former three ones use common FEA software to analyze the structures under disaster load. And by using their advanced function of graphic display, the whole damage process can be explicitly simulated. The last example is an integrate system to complete the real time simulation of structure.

3 Safety analyses for blast-resist-doors

3.1 Background information

A series of blast-resist-doors were designed 40 years ago. In order to know the damage process and the actual safety margin, the doors needs to be studied in-depth. If using the experimental method, it will be very expensive and the test is not safe. Hence, dynamic FEA simulation is chosen to analyze the doors.

3.2 Numerical models and results

The construction of the doors is shown in Figure 1 . The numerical models of the doors are shown in Figure 2. In order to know the real load conditions and damage process of the doors, contact surfaces[8,9] are define between the door leaf and doorframe, as well as between door hinges and door hinge bearings. So the interaction relationship of the door leaf, doorframe and door hinge bearings is simulated precisely. At the same time, considering the indeterminacy of the blast load, with the convenience of numerical simulation, 8 load groups, all together 24 load cases, are analyzed[10,11], almost covering all the load conditions the doors may meet. The 8 groups of load cases include 4 groups of conventional blast whose peak load densities are the design one, 2 groups of nuclear blast and other 2 groups of conventional blast whose peak load densities larger than the design ones.

Fig. 2 Element meshes and contact surfaces  

Fig. 2 Element meshes and contact surfaces

The plastic strain distribution on the door leaf at the time of maximal positive displacement is shown in Figure 3. The damage process of the doors is shown in Figure 4. From the numerical results, the internal stress, strain of the doors in the whole load process are obtained, together with the interaction between the door leaf and doorframe, as well as the one between door hinges and hinge bearings, which are difficult to determine in test.

3.3 Issues need to be notice

Fig. 4 Damage process  

Fig. 4 Damage process

Another research group (using ¡°Group B¡± to present) analyzed the doors with computer, too, paralleling with our work. But the results we obtained were much different from theirs. The main reason for the difference is the technical gap on simulating the real condition. Firstly, Group B used static FEA for the doors, which can not consider the strength enhancement of steel under high strain speed. Secondly, Group B simplified the doorframe to simple support edges, which can not obtain the damage process of the doors. So they had to use the allowable stress design method[12] to estimate the safety of the doors. However, the doors are designed that just need to resist blast only once. And even though there are some relatively large plastic strains in these doors, the doors will still be safe if they do not slide out of the frame. Hence, according to our research, the doors still have a lot of safety margin when the maximal plastic strain in the door leaf reaches 3%.

By comparing these differences, we can see that the kernel of simulation is to represent the real conditions on the computer as similar as possible. Especially to those issues which may influence the results seriously, it is important to simulate them carefully. Otherwise, maybe the whole simulation results are incorrect.

4 Study on the effect of new earth-penetration-bomb to RC tunnel

4.1 Background information

A type of earth-penetration-bomb showed striking damage ability to underground structures on the battle field of Kosovo and Afghanistan. In order to know the penetration ability of the bomb to the RC tunnel, dynamic FEA was carried out to simulate the penetration process.

4.2 numerical model and results

The numerical model and mesh for the RC tunnel are shown in Figure 5. In order to simulate the damage effect of the bomb precisely, the mesh near to the contact point is refined carefully. For the convenience of parameter discussion, the whole penetration process is divided into two stages artificially. The first stage is kinetic energy penetration. The second stage is explosion penetration. The initial speed and the depth where the bomb blast are discussed as parameters to estimate the damage effect[13]. All together 32 load cases are analyzed and lots of important information is obtained. The typical kinetic energy penetration and explosion penetration are shown in Figure 6 and Figure 7.

4.3 Issues need to be notice

The damage of bomb to structure is much more complex than the doors. So it is very important to simplify the models to grasp the key issues. Dividing the damage process into two isolated stages is the critical technical point. Then we can use difference numerical models to analyze the problem, avoiding the difficulties in coupling computational structure dynamics (CSD) and computational fluid dynamics (CFD).

Fig. 6 Kinetic penetration process

(a)            (b)              (c)               (d)

Fig. 6 Kinetic penetration process

Fig. 7 Explosion penetration process

(a)            (b)            (c)               (d)

Fig. 7 Explosion penetration process

5 Simulation for the collapse of World Trade Center (WTC) in New York

5.1 Background information

The twin towers of WTC impacted by airplane on Sep. 11, 2001 and destroyed entirely by chain collapse[14]. In order to know the reason for the chain collapse, Dynamic FEA is carried out to simulate the collapse process[15].

5.2 Numerical models and results

From the qualitative analysis, the possible reasons for the chain collapse are impact of upper floors to the lower ones, heap load effects and soften of steel under fire. So the numerical model is buildup with the following principle.

(1)     The dense-column-deep-beam tube-in-tube structure system of WTC is simplified to tube-in-tube structure made up of shell elements. In this approximate model, the global bending deformation and axial deformation is close to the real conditions, while the degree of freedom (DOF) is reduced a lot.

(2)     Introduce damage and fracture analysis into the numerical model to permit the structure elements breakdown under certain conditions.

(3)     Using the automatic contact search function in the FEA software to simulate influence of upper structures to lower ones.

(4)     Simulating the fire effect by changing the stiffness and strength of element material property[16,17].

The collapse process of the numerical results is shown in Figure 8.1 and Figure 8.2. From the simulation results, the following conclusions are obtained:

(1)   The reasons for the entire collapse of the towers are the structure elements¡¯ soften of fire and impact of the upper layers¡¯ collapse. From the numerical results, the towers does not collapse immediately after the impact. The north tower can go on standing. Likely, the south one dose not collapse, too, though there are some large deformations in it, which are caused by the asymmetric damage. This is consisted with the real situations.

(2)  Improving the structure fire resistance ability or control the fire influence area will avoid or delay the structure collapse, efficiently. We simulate the fire influence by adjusting the material property of elements. From the numerical results, even though the structure has been damaged seriously by the impact, if the influence area is smaller than 20%~25% of the survival section in the tower, the collapse still can be avoided. When more than 30~50% of the survival section near the impact zone fails, the collapse will start.

(3)   When the towers go into the collapse stage, the reason for the chain failure of un-impact layers is the impulse of upper collapsing floors. The impact force of upper floors is much larger than the heap load. And because there are a lot of bump and eject on the contact surface of collapsing floors and lower floors, the fragment of structure falls consecutively so that there is no chance to form a lot of heap load. So the heap load is not the critical reason for the collapse.

(4)   Improving the ductility of structure elements is an efficient way to avoid the chain collapse happens. In the simulation above, if the fracture plastic strain of steel structure is 0.5%, the chain collapse will take place entirely. However, if the fracture strain is improved to 1%, the impact energy of upper floors will be absorbed by the lower structures and the chain collapse will be stopped at about 100m under the airplane impact zone. When the fracture strain is improved to 5%, only part of the structure near the airplane impacting zone will be damaged, and no chain collapse will take place. Hence, if the structure has enough ductility to absorb the energy of upper floors¡¯ collapse, the chain damage will be controlled. Even though consider the influence of heap load, the towers still have much larger chance to escape from the entire collapse.

5.3 Issues need to be notice

Because this analysis is pure dynamic simulation, the structure will vibrate vertically once the gravity load is applied. It is not consisted with the real conditions. Hence, the loads are applied to the model by two steps. The gravity load is applied first, and a relatively large damp is given to the structure. After the vertical vibration stops, reduce the damp to ordinary value and apply the impact damage and fire influence.

How to build the numerical model is a critical difficulty in this analysis. Too simple model may not be able to reflect the interaction of structure fragments, while too complex model may bring too large computation workload. At the same time, the damage of airplane impact, the fire influence and the material property of structure in the collapse are difficult to determine. All these are the difficulties we met when we buildup the numerical model. Then, firstly, we consider the collapse of south tower was inclined. It is spatial asymmetry. Hence, the 3-D model should be adopted. Secondly, there are too many structure elements in the towers. They need not be buildup in the numerical one by one. So simplifying the tower with shell elements and using more elements to simulate the fragments behavior are suitable in this problem. To those parameters which are difficult to determine when buildup the model, such as impact damage zone, fire influence and fracture of material, are left as undetermined parameters and settle them by back analysis.

From this example we can see that, to some very complex simulation problems, the user should tried to use every possible functions that applied in the FEA software to simulate the various real problems, such as pure dynamic analysis, contact analysis, element active/inactive, modify material property, elastic-plastic analysis and fracture analysis. The user should be very familiar with the computational software and invest the target problem carefully. To those parameters which are difficult to determine, we should settle them with back analysis.

Fig. 8.1a Collapse of north tower (perspective view)  

Fig. 8.1a Collapse of north tower (perspective view)

Fig. 8.1b Collapse of north tower (vertical view)  

Fig. 8.1b Collapse of north tower (vertical view)

Fig. 8.2a Collapse of north tower (perspective view)  

Fig. 8.2a Collapse of north tower (perspective view)

Fig. 8.2b Collapse of south tower (vertical)  

Fig. 8.2b Collapse of south tower (vertical)

6 Displacement real-time display system for bridge under Typhoon

6.1 Background information

In order to know the safeties of large-span bridges, monitoring their real-time displacement and recording their fatigue history are very important. With the development of Global Position System (GPS) and computer science, the GPS of navigation satellites now can be used for the real-time monitoring of large-span bridge[18-20]. The data about the displacement of large span bridge under typhoon are real-time transmitted to the central control room and the displacement shape of the bridge should be able to displayed immediately to give a clear and direct conception about the situation of the bridge. Hence, a 3-dimensional displacement real-time display system is added into the bridge monitoring system[21].

6.2 Technical requirement

(1)   It should be able to communicated with the monitoring system and analyze the data in the monitoring system directly.

(2)   It should be able to share data with other software used on the bridge.

(3)   It should be able to obtain the deformation of total bridge with the survey point displacement.

(4)   It should be able to display the deformation real-time.

(5)   The deformation can be displayed in 2-dimension or 3-dimension, respectively. The display scale can be modified by the user.

(6)   It can display the time-displacement curves of survey points.

6.3 Technical details and display examples

With Object-Oriented programming technology and ODBC open database interface, the display system is connected to the total monitoring system smoothly, while it also can execute independently as an external function. With the displacements of survey points which are placed on the different span position, fast-fourier-transform (FFT) algorithm method is adopted to do the frequency spectrum analysis to obtain the amplitudes of former 3 vibration models. Thus, the horizontal and vertical displacement of each point on the bridge are computed by interpolation, and displayed on the screen. With advanced graphic engine, the display of 2-D and 3-D deformation graph is processed parallel with the data processing and spectrum analysis. The deformation can be displayed real-time, and the users can ¡°wondering¡± on the bridge to see the deformation details[22]. The displacements of survey curves are shown together with the 2-D and 3-D graph. The working interface of this system is shown in Fig. 9.


Fig. 9 The working interface of display system

6.4 Issues need to be notice

Not like the former examples, this simulation system is an integrated system because real time results display is required. Hence, the functional sections are integrated together and data are transmitted directly between the data collection system and display system with database interface. The programming and modification of this system is much more difficult because various data constructions, various operation systems and various programming languages are used while the display speed is required strictly. But all these work is deserved for the real time safety monitory of large span bridge.

7 Conclusions

With the development of computer science and computational mechanics, many problems, which can not be computed or only can be computed by simplified method, can be simulated on the computer nowadays, and the cost and workload are much lower than experimental research. However, for some problems, especially for those under disaster load condition, it is difficult to research by test. So they are lack of test data and the numerical results may be the only data the user can find. Then, the user should be very carefully with the numerical model and computational parameters. Otherwise, if the model is not correct, the results are often wrong, and the user often can not find out because there are no test data to compare.


1.         Zienkiewicz, O. C. The finite element method, Oxford ; Boston: Butterworth-Heinemann, 2000

2.         Graybeal, Wayne T, Simulation, principles and methods, Cambridge : Winthrop Publishers, 1980.

3.         JIANG Jianjing, HE xiaogang. Computer Simulation Analysis for Structures. Beijing: Tsinghua University Press, 1996

4.         LU Xinzheng, JIANG Jianjing. Application of Computer Simulation Technology on Special Load Problems. Computer Engineering and Applications. 2002.(7),vol.38, 12~13

5.         Stevens, Roger, The C graphics handbook, Boston : Academic Press, 1992

6.         Stein, Michael (Michael L.), Direct3D professional reference, Indianapolis, Ind.: New Riders Pub., 1997

7.         Fosner, Ron, OpenGL programming for Windows 95 and Windows NT, Reading, Mass.: Addison-Wesley Developers Press, 1997

8.         ANSYS User¡¯s manual. ANSYS Agency in Beijing, 1999

9.         LS-DYNA Structural User¡¯s Manual, Livemore Software Technology Corp. 1999

10.     LI Yiqi, Blast mechanics, Beijing: Science Press, 1992

11.     CHEN Zhaoyuan etc. Performance of RC structure under impact load, Beijing: Tsinghua University Press, 1986

12.     Wang Guozhou, Qu luqian. Steel structure----principles and design. Beijing: Tsinghua University Press, 1993

13.     Smith, P. D. Blast and ballistic loading of structures, Oxford; Boston: Butterworth -Heinemann, 1994

14.     Z. P. Bazant and Y. Zhou, Why Did the World Trade Center Collapse?, Society for Industrial and Applied Mathematics News, vol. 34, No. 8 (October, 2001).

15.     LU Xinzheng, JIANG Jianjing, Simulation for the collapse of WTC after aeroplane impact , Chinese Journal of Civil Engineering, 2001, 34(6): 8~10

16.     Jiang Jianjing, Gong Xiaonan, Wang Yuanqing, Cui Jinghao. Analysis and dispose for building accidence. Beijing: China Building Industry Press, 1998

17.     GUO Zhenhai, Principle of reinforce concrete, Beijing: Tsinghua University Press, 1999

18.     Guo Jinjun, Jiang Jianjing, Shang Ruibin, Li Qiguang, Tan Guohuan. Test research on the Tsing-ma Bridge of Hong Kong with GPS. Research Report of Civil Engineering Department, Beijing: Tsinghua University, 1999

19.     Institute of Bridge Science, Bridge Bureau of Railway Department, P. R. China. Suspend Bridge. Beijing: China Science and Technology Press, 1996

20.     Zhu Guixin etc. 3-D Displacement Real-time Kinematic Monitoring System of Humen Bridge. Research Report of Civil Engineering Department, Beijing: Tsinghua University, 2001

21.     Jiang Jianjing, Lu Xinzheng, GUO Jingjun, Study for Real-time Monitoring of Large-Span Bridge Using GPS, proceedings of ISSST 2002 ¡°Progress in Safety Science and Technology¡±, Beijing/New York: Science Press, 2002, 308¡«312

22.     Sun Jiaguang, Yang Changgui. Computer Graphics. Beijing: Tsinghua University Press, 1995



Our Lab

Collapse Prevention Committee