Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/332275578

Updating a finite element model by using MFEM

Conference Paper · October 2018

CITATIONS READS
0 4

5 authors, including:

Saikat Bagchi Ardalan Sabamehr


Concordia University Montreal Concordia University Montreal
5 PUBLICATIONS 1 CITATION 16 PUBLICATIONS 3 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

IC-IMPACTS Project View project

SHM of buildings; Seismic Evaluation of Confined Masonry; Vulnerability Assessment View project

All content following this page was uploaded by Saikat Bagchi on 08 April 2019.

The user has requested enhancement of the downloaded file.


International Symposium on Structural Health Monitoring and Nondestructive Testing
4-5 October 2018, Saarbruecken, Germany
More info about this article: http://www.ndt.net/?id=23558

UPDATING A FINITE ELEMENT MODEL BY USING MFEM

F. A. Sakib*, S. Bagchi, A. Sabamehr, A. Bagchi, A. Bhowmick

Concordia University, Montreal, Quebec, Canada

*
Corresponding Author: z_dalia@encs.concordia.ca

Abstract
Updating Finite Element (FE) model of the structural system using data obtained from the real
structure is an essential step to validate the structure for its behavior study. In this study, the FE
model of a 4 story one bay 3D frame is updated using Matrix Update Method with MFEM. The
steel framed structure with a story height of 10 ft and floor area of 20 × 10 sqft is modeled in
SAP2000 v17.3.0. Modal analysis of the structure is performed to obtain its first 4 frequencies
and mode shapes. The structure is then excited applying White Gaussian Noise on the FE model
and the acceleration data noted. The acceleration data are then used to perform Frequency
Domain Decomposition in order to find the first 4 frequencies and mode shapes. The simulated
modal values obtained from the SAP2000 and those of the random excitation are found to be
significantly different. MFEM, a MATLAB based FE program is then used to adjust the
frequencies by suitably modifying the structural parameters. MFEM provides with the stiffness
adjustment factors, which is used to adjust the moment of inertias of the Model. After a number
of iterations the model is updated to match the measured frequencies with the frequencies
obtained from the modal analysis. In absence of the experimental data from the real structure this
process provides with a robust method for model updating, which is facilitates to have a
representative model of the real structure.

1. INTRODUCTION

Structural Health Monitoring (SHM) has brought a new dimension in structural property
assessment. It has become an essential tool to ensure structural safety by monitoring in-situ
behavior of structure [1] [2], reducing maintenance cost and real time monitoring for seismic
threat[4]. In SHM, structural properties (static and dynamic) are assessed on the basis of analysis
performed on the data collected from various sensors. In traditional wired system, sensors are
connected with Data Acquisition System [3] by wire. This process increases the installation time
and the overall cost of the SHM system. Wireless Sensing Network [7] is a very effective
solution to this problem [5]. As it doesn’t need any wire to connect the sensors with data
acquisition system, the cost of wires, labor and time of installation are significantly reduced. But
there are still some shortcomings like energy supply, unexpected data or noise in this system, etc.
Data collected through wireless network are always combined with small or large noise. That
International Symposium on Structural Health Monitoring and Nondestructive Testing 2018,
Saarbruecken, Germany

means the acceleration we get from the sensors are not only due to the acceleration of the
structure. We get the frequency domain from the time domain of the acquired data and then will
apply different algorithms to detect modal properties of the structure. Recently some new
technique like wavelet transform (WT), Hilbert-Huang transform (HHT) [8] have been
developed for the purpose. But SHM is not confined in monitoring a structure only, the data
collected through SHM system can be used to detect damage [6] and correlate Finite Element
model with the real structure. FE modelling software are very strong tools to predict the
behaviors of the structures and for designing structures. For many years engineers and
researchers are depending on them. But there are some shortcomings like potential construction
deviation, differences in actual material properties from that assumed in the model, complicated
connections that is different from the idealized ones used in the model etc. Model
Updating method is used to correlate an FE model to the real structure using its measured
response. In this paper, we will model a structure in SAP2000 and find its frequencies through
modal analysis. Then we will use White Gaussian Noise and keep the acceleration record for
each floor and find frequency with Frequency Domain Decomposition (FDD) [9] method. Then
we will use frequency from FDD as measured frequency and update the FE model with Matrix
Update method using MFEM

2. WORKING METHODOLOGY AND NUMERICAL MODEL

In this study first, a hypothetical framed structure is modeled using SAP2000 (Fig. 1). It is a 4
story frame having a floor area of 20 ft × 10 ft. The story height is 10 ft and the slab thickness is
5 in. The beam and columns are ‘I’ sections with outside height of 1 in, top flange width 0.4167
in, top flange thickness 0.0317 in, web thickness 0.0208 in, bottom flange width 0.4167 in and
bottom flange thickness 0.0317 in. Material is A992Fy50 (from SAP2000V17). Material for slab
was concrete having a strength of 4000 psi. The cross sectional area of 6.6096 square inches,
moment of inertia around strong and weak axis are respectably 157.9859 in4 and 7.9475 in4.
Polar moment of inertia is 0.2321 in4. Modal-analysis of the structure is performed using
SAP2000 were to determine the natural frequencies of the structure. Only XZ plane is considered
or the modal analysis and first four frequencies are obtained. After that, as shown in the Fig. 2, a
White Gaussian Noise (WGN) is produced using MATLAB. The WGN is applied on the FE
model using a time history function named WGN. It can be compared as a simulation of shaking
table test for ambient vibration. Accelerations for each floor due to WGN are found from
SAP2000. The sampling rate was 100 Hz. These accelerations can be considered as the data from
accelerometer or data from wireless sensor network.
International Symposium on Structural Health Monitoring and Nondestructive Testing 2018,
Saarbruecken, Germany

Fig. 1 3D Model of 4 story steel frame in Sap2000 Fig. 2 White Gaussian Noise Time History Function

Using these data, first four natural frequencies and mode shapes are found through Frequency
Domain Decomposition (FDD). These frequencies can be considered as the measured
frequencies through SHM system. There are some differences expected between the measured
frequencies and the analytical frequencies obtained by modal analysis. So the Finite Element
model needs to be updated to match the analytical frequency and measured frequency. Matrix
Update Method were used to update the FE model. To apply the Matrix Update method, MFEM
were used. There are many method to co relate a finite element model with real structure [11]
[12]. Among these method, matrix update method [1][10] has been used in this study. In Matrix
update method, mass matrix or stiffness matrix is adjusted to give the measured frequency.
MFEM is a MATLAB based program which take an input file about the geometry and properties
of materials. Then it tries to find some correlation factor through which the stiffness matrix can
be adjusted to match with the measured frequency [11] MFEM gives correlation factor after each
trial. Stiffness depends on modulus of elasticity and moment of inertia. Modulus of elasticity is a
material property, but moment of inertia is a sectional property. So the elements’ moment of
inertia were multiplied by the correlation factors found from MFEM. It can be noted that, cross
section were kept same every time as cross section of a model can’t be changed. After several
trials, the analytical and measured frequencies were very close and it could be said that mode
were co-related. The initial MFEM input file had 156 nodes, 180 elements, 2 sets of elements
(Beam and Shell), 2 Types of materials(Concrete and Steel), number of iteration per trial was
200,161 sets of real constant. For beam element, real constant means cross section and moment
of inertias. For shell elements only thickness is included. Only XZ plane were Considered and
FDD frequencies were used as measured frequencies.
International Symposium on Structural Health Monitoring and Nondestructive Testing 2018,
Saarbruecken, Germany

Fig. 3 Graphical plot of the Model from MFEM

3. RESULTS

Modal Analysis was done in SAP2000 for the first four modes and only XZ plane was
considered (fig 4). WGN was applied through a time history function and from SAP2000,
accelerations for each floor were found for every 0.01 second. These data were then transferred
to a MS Excel sheet and FDD was performed for first four peaks. Corresponding frequencies,
mode shapes and modal Assurance Criterion (MAC) were obtained (fig 5and 6). From the MAC
value, it can be observed that the selection of peaks were correct. These frequencies were
considered as measured frequencies through SHM procedure. There were differences between
the frequencies (table 1). So the Finite Element Model of the structure needed to be updated so
that there will be no difference between the analytical and the measured frequencies. Matrix
update Method through MFEM was used to correlate the model.

Fig. 4 From left 1st mode (1.854 Hz), 2nd mode (5.978 Hz), 3rd mode (10.914 Hz), 4th mode (15.488 Hz)
International Symposium on Structural Health Monitoring and Nondestructive Testing 2018,
Saarbruecken, Germany

Fig. 5 PSD of the acceleration data showing the selected four peaks and their MAC values

(a) (b)

(c) (d)

Fig. 6 First 4 modes from FDD,respectively (a),(b),(c) and (d)


International Symposium on Structural Health Monitoring and Nondestructive Testing 2018,
Saarbruecken, Germany

Analytical Measured
Difference
Mode Frequency Frequency
(%)
(Hz) (Hz)
First 1.854 1.9531 -5.074
Second 5.978 6.25 -4.352
Third 10.914 10.9375 -0.2149
Fourth 15.488 15.625 -0.8768
Table 1 Analytical (From SAP2000) and measured frequencies

After the first trial, the following mode shapes (fig 7) and correlation factor for each element were
obtained (8a)

(a) (b)

(c) (d)
Fig. 7 First 4 mode shape from MFEM respectively (a),(b),(c) and (d)
International Symposium on Structural Health Monitoring and Nondestructive Testing 2018,
Saarbruecken, Germany

The difference between the analytical and measured frequencies for the first trial is:

Analytical (MFEM) Measured


Mode Difference (%)
frequency (Hz) Frequency (Hz)
1 1.837 1.9531 -5.96
2 6.007 6.25 -3.89
3 11.23 10.9375 2.65
4 16.4 15.625 4.94

Table 2 Analytical frequencies from MFEM and measured frequencies (for 1st trial)

After 52 more trials the correlation factors were almost 1 (one) (fig 8b), which means the
differences between analytical and measured frequencies were not significant.

(a) (b)

Fig. 8 First Correlation Model from MFEM; (a) After 1st trial, (b) After 52nd trial

After correlation was completed, the analytical and measured frequencies are as follow:

Analytical Measured
Mode Difference (%)
Frequency (Hz) Frequency (Hz)
1 1.947 1.9531 -0.29
2 6.243 6.25 -0.11
3 10.93 10.9375 -0.07
4 15.62 15.625 -0.02

Table 3 Analytical frequencies from MFEM and measured frequencies (for final trial)
International Symposium on Structural Health Monitoring and Nondestructive Testing 2018,
Saarbruecken, Germany

From Table 3, it can easily be observed that the analytical and measured frequencies have very
insignificant differences. So it can be decided that the Finite Element model of 4 story steel
frame can behave like the actual structure now.

4. CONCLUSION AND DISCUSSIONS


Difference between Finite Element model and real structure mainly occurs due to the change in
bonding nature of beams and columns, change of material properties in the field condition etc.
SHM systems provide the data for field situation. So using these data, Finite Element model can
be correlate with real structure and the finite element model will be more useful than previous
one. Model can be updated by Matrix update method using MFEM. Number of element is very
important while using MFEM. The more number of element, the more efficient MFEM can be.
In this study, number of element were 180, if this number is increased, than it would take less
trial to get the perfect correlation factors. While updating model with MFEM, only moment of
inertias are multiplied by the correlation factor, not the cross-sections or thickness or modulus of
elasticity. So for the shell elements, nothing was done. The correlation factors for shell
eventually get near one after the adjustment of beam elements. Only XZ plane was considered in
the study. Further study can be made for other planes including torsional modes. In that case, the
boundary conditions of the nodes will be changed

References
1. Celebi, Ahmet Sanli, Mark Sinclair, Sharon Gallant and Dan Radulescu, (2003). REAL-
TIME SEISMIC MONITORING NEEDS OF A BUILDING OWNER AND THE
SOLUTION. 11th FIG Symposium on Deformation Measurements, Santorini, Greece.
2. Lynch J.P. and Loh K.J. A Summary Review of Wireless Sensors and Sensor Networks for
Structural Health Monitoring
3. Agre, J. R., Clare, L. P., Pottie, G. J., and Romanov, N. P., 1999, “Development Platform for
Self-organizing Wireless Sensor Networks,” in Proceedings of the Conference on Unattended
Ground Sensor Technologies and Applications, Orlando, FL, April 8–9, Proceedings of the
SPIE, Vol. 3713, 257–268.
4. Roland Priemer (1991). Introductory Signal Processing. World Scientific.
p. 1. ISBN 9971509199. Archived from the original on 2013-06-02.
5. Brincker R, Zhang L and Andersen P. Modal identification of output-only systems using
frequency domain decomposition, Smart Mater. Struct. 10 (2001) 441–445
6. Le Thai Hoa, Yukio Tamura, Akihito Yoshida, And Nguyen Dong Anh, Output-only System
Identification Using Wavelet Transform
7. B. A. D. Piombo, A. Fasana, S.Marchesiello And M. Ruzzene , Modelling And Identification
Of The Dynamic Response Of A Supported Bridge
8. Filipe Manuel Rodrigues Leite De Magalhães, Operational modal analysis for testing and
monitoring of bridges and special structures
9. Morteza PAYAB, Farbod AHMADIFAR. A Comparative Review on Operational Modal
Analysis Methods
International Symposium on Structural Health Monitoring and Nondestructive Testing 2018,
Saarbruecken, Germany

10. Sabamehr,A. ; Bagchi,A. ; Tirca,L. ; Panigrahi,S.K. ; Chourasia,A. : System


Identification and Model updating of a steel cantiliver beam using vibration test
11. Lei, Y. (Dept. of Civil Eng., Xiamen Univ., Xiamen, China); Wang, H.F.; Shen,
W.A. Source: Smart Structures and Systems, v 10, n 4-5, p471-83, Oct.-Nov. 2012: Update
the finite element model of Canton Tower based on direct matrix updating with incomplete
modal data.
12.Duan, Zhongdong (School of Civil Engineering, Harbin Institute of Technology, Harbin
150090, China); Spencer,B.F.; Yan,Guirong; Ou, Jinping Source: Earthquake Engineering
and Engineering Vibration, v 3, n 1, p 67-74, June 2004: An improved optimal elemental
method for updating finite element models

View publication stats

You might also like