Mass Scaling

MASS SCALING

Objective: The purpose of this project is to increase the timestep thereby reducing the runtime of the simulation by invoking the Keyword CONTROL_TIMESTEP, where a nonphysical mass is added using DT2MS and maintaing the stability by TSSFAC Timestep Scale Factor. The limit for the mass scaling is restricted to 8%.

Procedure: The file can be opened in either text editor( say notepad++ ) or in the LS-Prepost. But, it is less time consuming and easy to make changes in the text editor.

Trial 1: The model has the TSSFAC and DT2MS values as 0.9 and -3.5E-5 respectively. It is run without any changes and the clock time it displayed is 32 hrs and 20 mins.

Trial 2: The DT2MS is changed to -4.5E-5 while keeping the TSSFAC value same. The estimated clock time is 31hrs and 59mins, with no added mass.

Trial 3: Now the DT2MS is changed to -5.5E-5 and the TSSFAC as it is. It is observed that a mass of 2.86E+3 has been added and the percentage of increase is 1E-2. The runtime has reduced to 28hrs 27mins.

Trial 4: Now the DT2MS is changed to -7.5E-5 and the TSSFAC as it is. It is observed that a mass of 1.0608E+5 has been added and the percentage of increase is 3.73E-1. The runtime has reduced to 21hrs 5mins.

Trial 5: DT2MS has been changed to -9.5E-5 and the TSSFAC is untouched. It is observed that a mass of 1.227E+6 has been added and the percentage of increase has raised to 4.313. The runtime has reduced to 16hrs 39mins. Which is reduction of 3hrs and 34 mins.

Trial 6: In the same way DT2MS is changed (increased) to -1.5E-4 and the TSSFAC is kept the same. It is observed that a mass of 3.253E+7 has been added and the percentage of increase is 114.32. The runtime has reduced to 10hrs 32mins. The percentage increase is 114.32 which is 14.29 times 8%. So, the DT2MS (addition of mass) is to reduced.

Trial 7: Now the DT2MS has been changed (reduced) to -1.1E-4 and keeping the TSSFAC . It is observed that a mass of 5.803E+6 has been added and the percentage of increase is 20.393. The runtime has increased to 11hrs 56mins. The percentage increase is 20.3 which is 2.5375 times more than 8%. So, the DT2MS (addition of mass) is to be reduced further.

Trial 8: DT2MS is changed (reduced) to -1.05E-4 and the TSSFAC is maintained the same. It is observed that a mass of 3.0583E+6 has been added and the percentage of increase is 10.748. The runtime has increased to 15hrs 4mins.

Trial 9: DT2MS is increased to -1.02E-4 and the TSSFAC is constant. It is observed that a mass of 2.127E+6 has been added and the percentage of increase becomes 7.477 which is well below 8%. The runtime has increased to 15hrs 30mins compared to the earlier trial, but reduced when compared with the default run.

Trial 10: DT2MS is changed to -1.021E-4 and the TSSFAC as it is. It is observed that a mass of 2.1438E+6 has been added and the percentage of increase is 7.534. The runtime has reduced to 15hrs 29mins.

Trial 11: Now the DT2MS is changed to -1.022E-4 and the TSSFAC as it is. It is observed that a mass of 2.16E+6 has been added and the percentage of increase is 7.5915. The runtime has reduced to 15hrs 29mins.

Trial 12: Now the DT2MS is changed to -1.023E-4 and the TSSFAC as it is. It is observed that a mass of 2.1766E+6 has been added and the percentage of increase is 7.6493. The runtime has reduced to 15hrs 28mins.

Trial 13: Now the DT2MS is changed to -1.024E-4 and the TSSFAC is the same. It is observed that a mass of 2.193E+6 has been added and the percentage of increase is 7.7075. The runtime has reduced to 15hrs 27mins.

Trial 14: Now the DT2MS is changed to -1.025E-4 and the TSSFAC as it is. It is observed that a mass of 2.2098E+6 has been added and the percentage of increase is 7.7660. The runtime has reduced to 15hrs 26mins.

Trial 15: Now the DT2MS is changed to -1.0265E-4 and the TSSFAC as it is. It is observed that a mass of 2.235E+6 has been added and the percentage of increase is 7.8548. The runtime has reduced to 15hrs 24mins.

Trial 16: Now the DT2MS is changed to -1.0275E-4 and the TSSFAC as it is. It is observed that a mass of 2.252E+6 has been added and the percentage of increase is 7.9145. The runtime has reduced to 15hrs 24mins.

Trial 17: Now the DT2MS is changed to -1.0285E-4 and the TSSFAC as it is. It is observed that a mass of 2.2692E+6 has been added and the percentage of increase is 7.9747. The runtime has reduced to 15hrs 23mins.

Trial 18: Now the DT2MS is changed to -1.0289E-4 and the TSSFAC as it is. It is observed that a mass of 2.2761E+6 has been added and the percentage of increase is 7.999. The runtime has reduced to 15hrs 22mins.

The trial and error for adding the mass have been performed till the percentage increase of mass is upto 8 and found that the computational time or the runtime is 15hrs and 22mins.

The trials which have the same runtime have been omitted due to the repetitiveness. But, the total no of trials have been plotted in the Histogram.

Histogram :

RUNTIME

There are 11 trials that are very much nearer to the runtime that is stable at 8% of mass scaling out of which one trial is exceeding the limit of 8%.

DT2MS

The plot for DT2MS for the trials conducted to achieve the runtime for 8% of added mass. Most of the trials fall between -1.1E-4 and -1.0E-4 more number of trials have been conducted because, for a slight change in mass i.e., with the fourth decimal change of the value the percentage of mass increases drastically. That is why more number of trials are being conducted.

IMPLICIT ANALYSIS:

The model is now to be run with implicit method and the runtime is compared to that of the explicit method.

Material: The plate is rigid and the pins are given MAT24 PIECEWISE_LINEAR_PLASTIC material.

Section: A solid section with an ELFORM -1 has been defined.

Define Curve: Two curves, one with velocity and the other stress strain curve has been defined.

Velocity vs time

Stress vs strain

To shift from explicit to implicit analysis CONTROL_IMPLICIT_GENERAL card has to be used and the IMFLAG is set to 1 to activate implicit analysis.

CONTROL_IMPLICIT_AUTO card has been defined with IAUTO set to 1 which automatically adjusts the timestep size. ITEOPT optimizes the operation count step. ITEWIN is the iteration window. DTMIN is DT/1000 and DTMAX is DT*10.

CONTROL_IMPLICIT_SOLUTION: This card is used to choose between linear and nonlinear analysis in implicit method. NSOLVR is set to 12 by default and LSMTD which is a Line Search Convergence Method is by default set to 4.

CONTROL_IMPLICIT_SOLVER: The stiffness matrix inversion is being done through this keyword. LSOLVR is by default set to 4 which is a single precision solver, but it is advisable to use 5 which is a double precision solver for better results in case of a complex implicit analysis.

The model is then run and ended at 40mins 5sec as a Normal Termination.

Conclusion:

The time step in explicit method is usually small and the stablity limits the time step. In order to reduce the element processing time and improve the performance, mass scaling is done, which is to increase the time step and reduce the runtime of the model.

Positive DT2MS was not applied as it would reduce the elements with higher timestep to DT2MS value which has no meaning in applying Mass scaling.

The mass scaling has been done and at -1.0289E-4 DT2MS, 8% of mass has been added and the total runtime of the model has been reduced from 32hrs 20mins to 15hrs 22mins, which is 52.7% reduction in the total time taken without mass scaling.

Whereas in Implicit method the timestep size is large and the accuracy dictates the timestep.

The model is run with the cards mentioned in the procedure for implicit method.

The model is a Nonlinear static implicit analysis.

Nonlinear: The ELFORM-1 in the section assignment tells that it is used in Nonlinear analysis. In addition to this, the stress strain curve is clear statement that the model considered is Nonlinear.

Static: The velocity curve is static throughout the termination time.

Implicit Analysis: The difference between the implicit and explicit analysis can be decided based on velocity or strain rates. If the strain rate is less than 10m/s then the model is taken as implicit method. These are some of reasons that the model should be an implicit method.

The termination time is also an other point that conveys that this model should be an implicit one but it does mean explicit method should not be performed.

Here ITEOPT and ITEWIN are the options to like reduce or increase the iterations depending on the timestep. The values are kept default as the model was a Normal Termination.

In the same way CONTROL_IMPLICIT_DYNAMICS card can also be used even though it is used for dynamic analysis. For static problems like this case, defining the GAMMA and BETA values as 0.6 and 0.4 respectively the implicit analysis can be performed and this is also done only when there is a convergence issue

In the same way DTMIN and DTMAX could also be played with, but there is no specific clock time mentioned when it comes to Implicit Method while solving as that of mentioned in Explicit method.

Implicit method takes less time to solve than explicit method. But invoking the mass scale card reduces the time taken when the model is run in explicit method.