Does SMAC underreport the impact duration?

General questions on the SMAC, msmac3D and other Collision Simulation programs
MSI
Site Admin
Posts: 2302
Joined: Thu Jun 18, 2009 12:37 pm

Does SMAC underreport the impact duration?

Post by MSI »

Q: I think it is a fair statement that smac under-reports the impact duration, generally if not always. My understanding is that the duration is shorter because the “unloading” or “restitution” phase is not calculated after max engagement. Instead, the restitution forces are “added” at the end of each time increment during the approach phase. Is this correct?

A: The SMAC model is a dynamic force/deflection model that includes modeling the restitution. Implementation of collision modeling in the SMAC computer program includes identical loading and unloading load-deflection rates. Therefore the unloading (restitution) phase of the collision was at near peak levels. The effect of partial restoration of the energy at near peak levels is that the SMAC program returns the restitution part of the absorbed energy but it does so with a less-than-actual dimensional recovery. This acts to reduce the duration of the restitution phase which therefore reduces the duration of the overall impact.

See our paper on Restitution SAE 970960, Effects of Restitution in the Application of Crush Coefficients, from the first few pages should provide an explanation:
  • The selected analytical approach to restitution at that time (presently retained in both original SMAC and EDSMAC and other clones of SMAC) included the use of extensive simplifying assumptions aimed at reducing the requirements for associated computer memory and logic. In particular, identical load-deflection rates were applied for loading and unloading.The unloading was implemented at force levels close to the peak values The duration of the impact in SMAC is less than measured tests due to the shorter time period for restitution."
We proposed a modification of SMAC in 1997 to better mimic the restitution behavior. The figure below is from the 1997 paper and provides a comparison of the difference in time between original SMAC and the proposed modification of SMAC
  • (and we also, of course, compared the two models with full scale test measurements to demonstrate the correlation of the revised model with the restitution measured in full scale tests)
smac restitution.jpg
smac restitution.jpg (34.44 KiB) Viewed 1850 times
This translates into the following Force v Crush plot (original smac vs revised smac)
force v crush revised smac.png
force v crush revised smac.png (54.04 KiB) Viewed 1765 times
Here are a measurements of a few crash tests of vehicles which demonstrate the correlation of our revised model (these are from our 1986 SAE paper 861894)
torino.png
torino.png (51.78 KiB) Viewed 1765 times
citation.png
citation.png (51.53 KiB) Viewed 1765 times
Due to erroneous issues created by the ill-informed reviewer during the SAE peer review process and the fact that the SAE session organizers ignored the situation, we haven't felt compelled to implement the proposed changes in any distributed version of SMAC (yet!).

Engineering Dynamics in 1999 made a poor choice as part of their modification EDSMAC4 to try to 'improve' restitution behavior of SMAC by overloading the initial loading portion of collisions in EDSMAC4.
From our 1999 SAE TOPTEC Review: SMAC Computer Program presentation (from CURRENT STATUS part of the presentation) you'll find:
  • In 1999, EDSMAC4 claims to include a "more realistic modeling of actual vehicle structural behavior" based on "an A, B stiffness model". Please note that the "A,B stiffness model" is a "virtual" model equating residual crush to dissipated kinetic energy. The "A,B stiffness model" was not intended to be a dynamic model. With the "A,B stiffness model" a very stiff near plastic vehicle can share the same "A,B stiffness" as a very soft near elastic vehicle. The residual crush on a vehicle does not tell you anything about the vehicle stiffness EXCEPT in the case of a plastic vehicle. Motor vehicles crushing in collisions are not plastic. They have restitution.

    Restitution consists of two separate aspects:
    • (1) a partial dimensional recovery and
      (2) a partial restoration of kinetic energy.

    In our 1997 SAE paper on restitution, we pointed out that the current implementation of collision modeling in the SMAC computer program includes identical loading and unloading load-deflection rates and that the unloading (restitution) phase of the collision was at near peak levels. The effect of partial restoration of the energy at near peak levels is that the SMAC program acts to return part of the absorbed energy but it does so with a less-than-actual dimensional recovery.

    EDSMAC4 includes in the "A,B stiffness model" a "threshold force to be applied before deformation begins". What mechanism exists in the real world which provides a force without a deflection? Are they assuming that the bumper is a pre-loaded spring? You have to do work to absorb energy in the vehicle structure. Work equals Force * Displacement. The effect of the changes in EDSMAC4 is that the acceleration peaks are higher and the duration of the impact impulse is shorter.

    EDSMAC4 has chosen to artificially stiffen the vehicle to try to better match the residual crush. In effect they have created a "residual crush" stiffness model which reduces the validity of the modeling of the loading phase of the collision.
    • The most important aspect of a collision is the loading phase.
The following from the EDSMAC paper demonstrates the issue with the 'front end loading' or 'overloading' the loading phase of the collision modeling.
What they don't show is that the crash test measurements during loading are more accurately modeled by the original SMAC model (EDSMAC and other clones, the dashed line).
  • The EDSMAC4 modifications act to further REDUCE the collision duration and increase the maximum acceleration over original SMAC (EDSMAC) and therefore reduce the correlation with actual test measurements

NOTE: generally when improving a simulation model of collisions you compare the results with the full scale tests, not another version of the simulation model. The RICSAC tests were thoroughly documented and measured and we demonstrated the results were accurate in our RICSAC97 SAE paper 97-0961.
The acceleration measurements for RICSAC were at locations other than the CG and easily addressed with the SMAC simulation model. See after the figures from the EDSMAC4 tests for some scans of the RICSAC measurements.
EDSMAC4 front end loading.jpg
EDSMAC4 front end loading.jpg (21.93 KiB) Viewed 1849 times
EDSMAC4 ricsac 12.jpg
EDSMAC4 ricsac 12.jpg (27.11 KiB) Viewed 1319 times
Here are a couple scans from RICSAC CRASH TESTS
ricsac9 firewall accel res.jpg
ricsac9 firewall accel res.jpg (239.55 KiB) Viewed 1319 times
ricsac 12 accel.jpg
ricsac 12 accel.jpg (239.83 KiB) Viewed 1319 times