Modelling Ground Vibrations in Abaqus Using Python
(2025) In TVSM-5000 VSMM01 20251Structural Mechanics
Department of Construction Sciences
- Abstract
- The need for accurate and computationally efficient methods for analysing the effects of ground vibrations on a structure is increasing. This is partly due to continued urbanization leading to structures being placed in closer proximity to vibrational sources, as well as increased environmental awareness leading to more lightweight structures, such as timber structures, being built. The purpose of this study was to create accurate and computationally efficient FE ground models for the analyses of ground vibrations in Abaqus.
This objective was achieved by creating tailored FE ground models based on the properties of the ground layers and the frequency interval that was to be studied. These parameters determined the longest and shortest... (More) - The need for accurate and computationally efficient methods for analysing the effects of ground vibrations on a structure is increasing. This is partly due to continued urbanization leading to structures being placed in closer proximity to vibrational sources, as well as increased environmental awareness leading to more lightweight structures, such as timber structures, being built. The purpose of this study was to create accurate and computationally efficient FE ground models for the analyses of ground vibrations in Abaqus.
This objective was achieved by creating tailored FE ground models based on the properties of the ground layers and the frequency interval that was to be studied. These parameters determined the longest and shortest wavelength of the vibrations, which in turn were used to determine the model size and element size respectively. P-waves have the longest wavelength and were used along with the lower frequency in a given range for determining model size, while Rayleigh wavelengths are the shortest and were used along with the upper frequency to determine element size.
A complete program was developed using Python and consists of the creation of axisymmetric and 3D FE models, creation of input files compatible with Abaqus, submission of jobs for analyses in Abaqus and extraction of results. Additionally, a user interface was created to increase user-friendliness. Parameter studies were conducted with the purpose of determining the appropriate number of P-wavelengths, the number of elements per Rayleigh wavelength, the size of the frequency increments and the size of the frequency bands that each model was to be tailored to. The resulting models were validated by comparing the results to a well-established semi-analytical model used for analyses of traffic-induced vibrations. The time saved through the use of the tailored models was determined by comparing analyses times for the tailored models with less tailored models.
The parameter study resulted in the tailored models consisting of 1.5 P-wavelengths used to determine model size, 5 elements per Rayleigh wavelength used to determine element size, 1 Hz sized increments being used in the analyses and each model being tailored to a 5 Hz frequency band. When compared to results from the semi-analytical model, the chosen values from the parameter studies resulted in sufficiently accurate results up to about 80 Hz, after which the results appeared to oscillate more. The accuracy above 80 Hz was deliberately compromised in favour of reasonable analyses times.
The 3D models were, even when tailored, too large for analyses on a regular computer and had to be submitted to a supercomputer for analyses. This was mostly due to the models' extensive RAM memory usage, often requiring 256 GB RAM memory and in some cases even 512 GB. This was an unexpected complication which resulted in the models, neither 3D nor axisymmetric, being able to be analysed in the 1-100 Hz range when not divided into smaller models, even on the supercomputer available for this project. In order to determine the computational efficiency of the tailored models, they were therefore compared to less tailored models that used the full computational capacity available for this project.
For sequential analyses, the tailored axisymmetric models were in the 1-100 Hz interval 4.4 times faster, taking 8 minutes instead of 36.5 minutes. The tailored 3D models were for sequential analyses in the 36-100 Hz interval 3.8 times faster, taking 2 days and 10 hours instead of 9 days and 4 hours for its less tailored counterparts. When applying parallel analyses, even more significant time savings were achieved. The tailored axisymmetric models were for the 1-100 Hz interval 12.2 times faster, taking 2.5 minutes instead of 30.5 minutes, while the tailored 3D models in the 75-100 Hz interval were 21 times faster, taking 4.5 hours instead of 4 days. (Less) - Popular Abstract
- The need for fast and accurate methods to analyse the effects of traffic-induced vibrations on a structure is increasing. This is partly due to continued urbanization causing structures to be built in closer proximity to roads, as well as an increased sustainability awareness leading to more environmentally friendly materials being used in constructions, such as timber which is often considered to be sensitive to traffic-induced vibrations. Analysis of vibrations on, for example, an underground basement can without effective methods take days or even weeks to process, whereas it was found in this project that the same analyses using efficient models can be achieved in mere hours instead.
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/9209222
- author
- Jupolli, Argeta LU
- supervisor
-
- Linus Andersson LU
- Jonas Lindemann LU
- Khuong An Ung LU
- organization
- alternative title
- Modellering av markvibrationer i Abaqus med Python
- course
- VSMM01 20251
- year
- 2025
- type
- H3 - Professional qualifications (4 Years - )
- subject
- keywords
- ground vibrations, tailored modelling, Abaqus, Python
- publication/series
- TVSM-5000
- report number
- TVSM-5278
- ISSN
- 0281-6679
- language
- English
- id
- 9209222
- alternative location
- http://www.byggmek.lth.se/english/publications/tvsm-5000-present-2014/
- date added to LUP
- 2025-08-19 09:09:09
- date last changed
- 2025-08-19 09:09:09
@misc{9209222,
abstract = {{The need for accurate and computationally efficient methods for analysing the effects of ground vibrations on a structure is increasing. This is partly due to continued urbanization leading to structures being placed in closer proximity to vibrational sources, as well as increased environmental awareness leading to more lightweight structures, such as timber structures, being built. The purpose of this study was to create accurate and computationally efficient FE ground models for the analyses of ground vibrations in Abaqus.
This objective was achieved by creating tailored FE ground models based on the properties of the ground layers and the frequency interval that was to be studied. These parameters determined the longest and shortest wavelength of the vibrations, which in turn were used to determine the model size and element size respectively. P-waves have the longest wavelength and were used along with the lower frequency in a given range for determining model size, while Rayleigh wavelengths are the shortest and were used along with the upper frequency to determine element size.
A complete program was developed using Python and consists of the creation of axisymmetric and 3D FE models, creation of input files compatible with Abaqus, submission of jobs for analyses in Abaqus and extraction of results. Additionally, a user interface was created to increase user-friendliness. Parameter studies were conducted with the purpose of determining the appropriate number of P-wavelengths, the number of elements per Rayleigh wavelength, the size of the frequency increments and the size of the frequency bands that each model was to be tailored to. The resulting models were validated by comparing the results to a well-established semi-analytical model used for analyses of traffic-induced vibrations. The time saved through the use of the tailored models was determined by comparing analyses times for the tailored models with less tailored models.
The parameter study resulted in the tailored models consisting of 1.5 P-wavelengths used to determine model size, 5 elements per Rayleigh wavelength used to determine element size, 1 Hz sized increments being used in the analyses and each model being tailored to a 5 Hz frequency band. When compared to results from the semi-analytical model, the chosen values from the parameter studies resulted in sufficiently accurate results up to about 80 Hz, after which the results appeared to oscillate more. The accuracy above 80 Hz was deliberately compromised in favour of reasonable analyses times.
The 3D models were, even when tailored, too large for analyses on a regular computer and had to be submitted to a supercomputer for analyses. This was mostly due to the models' extensive RAM memory usage, often requiring 256 GB RAM memory and in some cases even 512 GB. This was an unexpected complication which resulted in the models, neither 3D nor axisymmetric, being able to be analysed in the 1-100 Hz range when not divided into smaller models, even on the supercomputer available for this project. In order to determine the computational efficiency of the tailored models, they were therefore compared to less tailored models that used the full computational capacity available for this project.
For sequential analyses, the tailored axisymmetric models were in the 1-100 Hz interval 4.4 times faster, taking 8 minutes instead of 36.5 minutes. The tailored 3D models were for sequential analyses in the 36-100 Hz interval 3.8 times faster, taking 2 days and 10 hours instead of 9 days and 4 hours for its less tailored counterparts. When applying parallel analyses, even more significant time savings were achieved. The tailored axisymmetric models were for the 1-100 Hz interval 12.2 times faster, taking 2.5 minutes instead of 30.5 minutes, while the tailored 3D models in the 75-100 Hz interval were 21 times faster, taking 4.5 hours instead of 4 days.}},
author = {{Jupolli, Argeta}},
issn = {{0281-6679}},
language = {{eng}},
note = {{Student Paper}},
series = {{TVSM-5000}},
title = {{Modelling Ground Vibrations in Abaqus Using Python}},
year = {{2025}},
}