Non-reflecting boundary setting in air blast simulation
yh p
l...@schwer.net
Non-reflecting boundaries are intended to work with Elastic wave, not blast waves; elastic (acoustic) waves in air are a fraction of atmospheric pressure.
The only solution is to move the Eulerian boundaries far from the region of interest; you can try element geometric ratioing to make larger elements farther away,
Also, have a look at the parameter BNDFLX on *CONTROL_ALE
--len
From: ls-d...@googlegroups.com <ls-d...@googlegroups.com> On Behalf Of yh p
Sent: Tuesday, October 31, 2023 1:30 AM
To: LS-DYNA2 <ls-d...@googlegroups.com>
Subject: [LS-DYNA2] Non-reflecting boundary setting in air blast simulation
Hi all,
I am using the ALE method to simulate a 1/4 air blast model. I defined the non-reflecting boundary, but it still reflected. As the pressure-time curve shows, there is another large peak after the first one. Is there something wrong with my boundary settings?
Best regards,
Yahao
--
You received this message because you are subscribed to the Google Groups "LS-DYNA2" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ls-dyna2+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/ls-dyna2/2d99b70f-bcf0-462f-9784-779ab6f974d1n%40googlegroups.com.
Kağan GENÇ
I have been working on this problem for over 2 years and as Leonard says the cleanest solution is to define a bigger air domain. However, I am applying another method by myself and it worked for me. I would like to share some notes from my thesis.
- In LS-DYNA, numerous contact algorithms are already introduced for different cases. The contact between air domains and structural parts was defined with a penalty-based contact algorithm, CLIS (Constrained_Lagrange_In_Solid), widely utilized in the literature. Blastwave influxes at boundary conditions have always been problematic for the models considering fluid-structure interaction because of the high computational cost of defining big air domains that do not cause influxes. Therefore, the contact algorithm between air domains and structural parts was stopped after a specific time to prevent blastwaves from causing additional pressure on structural parts due to the blastwaves reflecting into the air domain from boundaries. The time to stop the contact was decided according to the summation of arrival time and positive phase duration of explosions, calculated by the Kingery-Bulmash spherical air burst model [14]. Hence, this timing ensures the observation of all effects of explosions. Even though the contact algorithm was canceled after the end of the positive phase duration, blastwave influxes were again observed around boundary conditions, which resulted in a sudden and massive time step decrease and subsequent convergence problems. To dispose of this drawback, a keyword, Boundary_Non_Reflecting, has been widely adopted in the literature. However, this keyword specifically addresses geo-mechanical applications and is appropriate to damp elastic waves, not inelastic waves such as fluid flow [15]. The option, namely BNDFLX, was already introduced in Control_Ale keyword in LS-DYNA to prevent influxes from getting into back to air domain. However, this option also resulted in a sudden time step decrease and convergence problems since it is not a physical solution. It is suggested to use BDFLX option with MINMAS option, which allows elements to have more mass to increase time step size. Even though both options were assigned to numerical models, energy fluctuations, i.e., the cause for time step decreases, could not be eliminated. Instead of all these applications in the literature, the author proposed a quick solution for the problem. The ambient pressure on the free boundary conditions was decreased to zero after a specific time, i.e., the summation of arrival time and positive phase duration. This application helped exit air particles after the interaction between structural parts and air particles would be assumed negligible. This newly proposed method solved time step decreases and unwanted energy fluctuations in air domains.
- Numerical modeling of explosions considering fluid-structure interaction is challenging due to possible wave reflections from finite boundary conditions and related sudden time step decrease in the elements of the air domain. Therefore, using a keyword, i.e., Boundary_Non_Reflecting and MINMAS option accompanied by BDFLX option in Control_Ale keyword have been employed in the literature. However, defining non-reflective boundaries may only dampen elastic waves and is mainly used for geomechanical applications, while the MINMAS option is not a physical solution but a numerical trick that may be turned into a computational misuse. As a result, these two widely used options to eliminate blast wave reflections from finite boundary conditions are not a solid and physical solution to the problem.
- Explosive modeling considering fluid-structure interaction is always problematic, and some proposed solutions in the literature are not sensible, as stated in the conclusion. Even though the safest solution is to increase air domain size, this solution might result in a massive computational cost, especially for full-scale models. Therefore, the author proposed another solution as a quick fix for the problem of reflected blast waves that cause instabilities. It is suggested to decrease ambient pressure to zero after the interaction between fluid and structure becomes negligible. This method enables fluid material to flow out of its domain and prevents inward flows, which would result in extremely small time steps. It must be noted that the ambient pressure needs to be decreased to zero after the positive phase of explosions is complete, and this estimation needs fine-tuning.
Kind regards,
Oğuz Kağan GENÇ
l...@schwer.net
Thanks Kagan for sharing your thesis with numerous concrete targets subject to blast.
Your suggested method of decreasing the ambient pressure at the air domain boundaries is interesting. However, your thesis does not seem to mention how this was done in LS-DYNA? Typically one sets the *CONTROL_ALE parameter PREF=1 atm and that pressure condition is constant for the duration of the simulation. I assume you defined *LOAD_SEGMENT keywords for all the air domain external boundaries as mentioned in Remark 8 for the *CONTROL_ALE keyword?
Did you perform any air domain size studies using your proposed ambient pressure reduction technique, i.e. could one also reduce the size of the air domain with this technique?
I too have noted that the *CONTROL_ALE parameter in BNDFLX does not work as desired. However, the parameter MINMAS has helped me on occasion.
I look forward to an opportunity to try your suggested air domain boundary technique. --len
From: ls-d...@googlegroups.com <ls-d...@googlegroups.com> On Behalf Of Kagan GENÇ
Sent: Wednesday, November 1, 2023 7:24 AM
To: yh p <pyhx...@gmail.com>
Cc: LS-DYNA2 <ls-d...@googlegroups.com>
Subject: Re: [LS-DYNA2] Non-reflecting boundary setting in air blast simulation
Dear Yahao,
I have been working on this problem for over 2 years and as Leonard says the cleanest solution is to define a bigger air domain. However, I am applying another method by myself and it worked for me. I would like to share some notes from my thesis.
- In LS-DYNA, numerous contact algorithms are already introduced for different cases. The contact between air domains and structural parts was defined with a penalty-based contact algorithm, CLIS (Constrained_Lagrange_In_Solid), widely utilized in the literature. Blastwave influxes at boundary conditions have always been problematic for the models considering fluid-structure interaction because of the high computational cost of defining big air domains that do not cause influxes. Therefore, the contact algorithm between air domains and structural parts was stopped after a specific time to prevent blastwaves from causing additional pressure on structural parts due to the blastwaves reflecting into the air domain from boundaries. The time to stop the contact was decided according to the summation of arrival time and positive phase duration of explosions, calculated by the Kingery-Bulmash spherical air burst model [14]. Hence, this timing ensures the observation of all effects of explosions. Even though the contact algorithm was canceled after the end of the positive phase duration, blastwave influxes were again observed around boundary conditions, which resulted in a sudden and massive time step decrease and subsequent convergence problems. To dispose of this drawback, a keyword, Boundary_Non_Reflecting, has been widely adopted in the literature. However, this keyword specifically addresses geo-mechanical applications and is appropriate to damp elastic waves, not inelastic waves such as fluid flow [15]. The option, namely BNDFLX, was already introduced in Control_Ale keyword in LS-DYNA to prevent influxes from getting into back to air domain. However, this option also resulted in a sudden time step decrease and convergence problems since it is not a physical solution. It is suggested to use BDFLX option with MINMAS option, which allows elements to have more mass to increase time step size. Even though both options were assigned to numerical models, energy fluctuations, i.e., the cause for time step decreases, could not be eliminated. Instead of all these applications in the literature, the author proposed a quick solution for the problem. The ambient pressure on the free boundary conditions was decreased to zero after a specific time, i.e., the summation of arrival time and positive phase duration. This application helped exit air particles after the interaction between structural parts and air particles would be assumed negligible. This newly proposed method solved time step decreases and unwanted energy fluctuations in air domains.
- Numerical modeling of explosions considering fluid-structure interaction is challenging due to possible wave reflections from finite boundary conditions and related sudden time step decrease in the elements of the air domain. Therefore, using a keyword, i.e., Boundary_Non_Reflecting and MINMAS option accompanied by BDFLX option in Control_Ale keyword have been employed in the literature. However, defining non-reflective boundaries may only dampen elastic waves and is mainly used for geomechanical applications, while the MINMAS option is not a physical solution but a numerical trick that may be turned into a computational misuse. As a result, these two widely used options to eliminate blast wave reflections from finite boundary conditions are not a solid and physical solution to the problem.
- Explosive modeling considering fluid-structure interaction is always problematic, and some proposed solutions in the literature are not sensible, as stated in the conclusion. Even though the safest solution is to increase air domain size, this solution might result in a massive computational cost, especially for full-scale models. Therefore, the author proposed another solution as a quick fix for the problem of reflected blast waves that cause instabilities. It is suggested to decrease ambient pressure to zero after the interaction between fluid and structure becomes negligible. This method enables fluid material to flow out of its domain and prevents inward flows, which would result in extremely small time steps. It must be noted that the ambient pressure needs to be decreased to zero after the positive phase of explosions is complete, and this estimation needs fine-tuning.
Please see my attached thesis if needed.
Kind regards,
Oğuz Kağan GENÇ
.
Kağan GENÇ
In my case, I tried to use BNDFLX and MINMAS together when we last
talked here in this group. However, I needed to increase the MINMAS
value a lot and it distorted the results. Therefore, it might be
dangerous in my opinion. Also, I could not find to what degree MINMAS
can be increased. For example, a maximum 10% of total mass can be
added for mass scaling. However, I could not find any relevant
information like that. Therefore, I did not prefer to use it.
Yes you are right I applied pressure using LOAD_SEGMENT. As I know,
there is no other way to apply pressure as a function of time. I
should have mentioned it in the thesis.
I have carried out a lot of simulations that employed decreased
ambient pressure. However, I did not write any report about the
outcomes. I can write what I have observed on the effects of air
domain size and the timing to start decreasing ambient pressure.
- The main parameter affecting the air domain size is scaled
distance since the summation of arrival time and positive phase
duration is quite important in this case ( I assume negative phase is
negligible). The earlier decrease of ambient pressure may obviously
result in reduced impulse.
- If there are multiple targets that waves can reflect from one
another, it may be more complicated. and result in a bigger air domain
since all reflections need to be taken into consideration.
- According to my observation from a 1/4 model, I arrange the 2
dimensions of the air domain a little bit longer than the stand-off
distance at the first run. (the other dimension is the stand-off
distance in a quarter model). Then, I estimate the summation of
arrival time and positive phase duration. If the wave starts to
reflect from boundaries in the estimated time, I prefer to lengthen
the 2 dimensions of air domain because it may result in
additional/fictitious effects on structural parts. Thus, the wave may
exit before it starts to reflect when ambient pressure is decreased.
This is the simple way I apply to find out the size of the air domain.
Actually, the idea behind this is quite simple. I think we can agree
that if the interaction between air and structural parts becomes
negligible, we can stop the interaction. As I have seen a lot of
researchers use contact stop command to eliminate the effects of
reflected waves. Most of the time stopping contact algorithms works
fine. Whenever reflected waves somehow cause influxes at boundaries,
it results in extremely small time steps, pressure increase, and
subsequent convergence problems. I was wondering if it is possible to
stop calculating ALE elements after a certain time, however, the stop
command in Control_Ale does not work properly. Consequently, I
prefered to guide waves out of the air domain by decreasing ambient
pressure after a certain time. Only thing is we need to decide about
the timing where the interaction between air and structural parts
becomes negligible. I figured this out by estimating the summation of
up arrival time and positive phase duration.
If you have any suggestions on this way, I would appreciate it.
Kind regards,
Oğuz Kağan GENÇ
tarihinde şunu yazdı:
Ameen Topa
Thanks for the insightful discussion. Indeed, applying pressure vs time curve on the boundary elements seems like an interesting approach. With regards to the stopping of ALE and FSI calculation, I have tried the following and it works:
- Using ALE STRUCTURED MESH, I used TDEATH to set the time at which all ALE elements would be removed from the simulation. This has greatly reduced the run time.
- In ALE STRUCTURED FSI and CONSTRAINED LAGRANGE IN SOLID, we have START and END to control the period at which the coupling is active. This also helps in reducing the run time. (but not as much as the first point above)
- Lastly, I used DEFORMABLE TO RIGID to set all the structural parts to be RIGID until the blast overpressure occurs. If you have multiple targets, this might help a lot.
Sincerely,
Ameen
--
You received this message because you are subscribed to the Google Groups "LS-DYNA2" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ls-dyna2+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/ls-dyna2/CAKF7tsW2j6%3DpNHkcYe_faKPJV7va_pacEBWcdVz%2BWqYNTYbp0Q%40mail.gmail.com.
l...@schwer.net
I plan to perform some numerical experiments.
I believe you have not mentioned the manner in which you decrease the atmospheric pressure at the boundary. Do you suddenly remove the pressure or gradually, and if the later over what time period?
--len
You received this message because you are subscribed to the Google Groups "LS-DYNA2" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ls-dyna2+u...@googlegroups.com.
Kağan GENÇ
Nice to hear that you will try too. I would love to get informed about
your observations.
As I remember, when I tried to assign 10 atm and 0 atm at the same
time, i.e., sudden decrease, I got some problems. Therefore, when I
check my keyword, I see a decrease in 0.1ms, as seen below..
I think we can call this a gradual decrease since 0.1ms is not a too
short time period for explosions. Also, I would like to mention that I
stopped the interaction between fluid and structural parts around
0.0025ms in this study. Therefore, I preferred to decrease the ambient
pressure between 0.0025ms and 0.0026ms.
Kind regards,
Oğuz Kağan GENÇ
tarihinde şunu yazdı:
l...@schwer.net
Hello Kagen --
Sorry for the long pause in responding.
It does appear your idea of reducing the external free air boundary domain pressure from one atmosphere to (near) zero is an improvement over doing nothing to prevent unwanted reflections traveling back into the air domain.
I created a 1D spherical model using beam elements with 1.5 kg of TNT at the center and a free boundary at 2.5 meters. I measure the pressure history at 1.5 meters for three cases:
1\ free boundary using PREF=0.1 MPa -- consider the default configuration
2\ PREF=0 but added a nodal force on the last beam node equal to one atmosphere. A force (0.1) (2*pi*R^2) is required for beam elements rather than pressure. This was a check that prescribing the external pressure was equivalent to PREF=0.1 MPa
3\ Changed the above prescribe constant force history to one that recognizes the arrival at the boundary, t=2.75ms, and decays to zero pressure after an additional 2ms approximately equal to the positive phase duration of the incoming blast wave.
The plot below shows the pressure and impulse histories at 1.5m (Tracer T7 in my model) for the three cases describe above.
The blue lines are Case 1 (PREF=0.1 MPa)
The red lines are Case 2 (PREF=0 but with constant Nodal Force)
The green lines are Case 3 (PREF=0 with decaying Prescribed Nodal Force)
Anyone that would like my spherical model may request it from me. --len
To view this discussion on the web visit https://groups.google.com/d/msgid/ls-dyna2/CAKF7tsUx10rWaxQYm%2B1omu%3D5GgP574Gq19DDCyk2wNCm1QeNgw%40mail.gmail.com.
Kağan GENÇ
Hello Kagen --
Sorry for the long pause in responding.
It does appear your idea of reducing the external free air boundary domain pressure from one atmosphere to (near) zero is an improvement over doing nothing to prevent unwanted reflections traveling back into the air domain.
I created a 1D spherical model using beam elements with 1.5 kg of TNT at the center and a free boundary at 2.5 meters. I measure the pressure history at 1.5 meters for three cases:
1\ free boundary using PREF=0.1 MPa -- consider the default configuration
2\ PREF=0 but added a nodal force on the last beam node equal to one atmosphere. A force (0.1) (2*pi*R^2) is required for beam elements rather than pressure. This was a check that prescribing the external pressure was equivalent to PREF=0.1 MPa
3\ Changed the above prescribe constant force history to one that recognizes the arrival at the boundary, t=2.75ms, and decays to zero pressure after an additional 2ms approximately equal to the positive phase duration of the incoming blast wave.
The plot below shows the pressure and impulse histories at 1.5m (Tracer T7 in my model) for the three cases describe above.
The blue lines are Case 1 (PREF=0.1 MPa)
The red lines are Case 2 (PREF=0 but with constant Nodal Force)
The green lines are Case 3 (PREF=0 with decaying Prescribed Nodal Force)
Anyone that would like my spherical model may request it from me. --len
<image001.png>
Kağan GENÇ
On Nov 15, 2023, at 12:38 AM, l...@schwer.net wrote:
Sorry I should have pointed out that in my model the reflected wave from the free boundary arrives at about 7.25ms as indicted in the image below.
<image001.png>