You are here
How do I change loading rate to speed up quasi-static simulation in Abaqus/Explicit?
Big Picture: I am trying to perform a quasi-static, displacement controlled, 3-point bend simulation on a 2D notched beam. My simulation size is about 7100 variables. Ultimately, I am trying to run this simulation using a VUMAT implementing a damage accumulation material model, but for right now I am using the built in Elastic model while I figure out how to manipulate loading rate. I want to run this model for at least 100 cycles which will be done using steps. I am trying to get the simulation to run in a reasonable amount of time so that I can get information quickly. I have been able to use Mass Scaling to speed up the model, which seems to work, but does distort my results. The only information I want from the simulation is reaction force and displacement, this is because I want to calculate stiffness of the structure (i.e. force divided by displacement). A lot of the documentation online talking about doing Quasi-Static simulations using Abaqus Explicit says that the two ways of speeding up the simulation are mass scaling and controlling loading rate.
Problem: I want to compare the results using the two methods and see if using a faster loading rate will help with the distortion of my results. I have tried to figure out how to control loading rate but my results are not making sense. I have stepped back from my simulation and am working with a 2D, one by one, square plate while I try to figure out how to manipulate loading rate. Basically, my question is, how do I control loading rate using Abaqus Explicit?
What I have tried: Under the amplitude keyword there are 4 numbers in a row separated by commas. I have tried adjusting the third number both greater than 1 and less than 1. From what I understand in both the Abaqus documentation and what I have read on line these numbers come in pairs such that the amplitude follows this: time_0, amplitude_0, time_1, amplitude_1, time_2, amplitude_2. I have been using this such that: time_0, amplitude_0, time_1, amplitude_1 where 0 represents the beginning of the step and 1 represents the end of the step. Therefore, I have manipulated the number representing time_1.
From what I can understand it seems like reducing this number to a value less than 1 would increase the rate at which loading is applied and therefore reduce the time the simulation runs. I say this because a rate is an event divided by time. So, if time is decreased the rate should increase.
My results: It is hard to see in the two graphs but essentially all graphs look like the light green one. In that the data points are not evenly spread out throughout the x-axis. The simulation goes from 0,0 to the second point and then all other points are in a vertical line. It seems like the graphs are converging, but the vertical lines seem odd. Even if I apply this to my notched beam simulation with 7100 variables, it doesn't speed up at all.
Extra Information:
System: Abaqus 6.11 run on a Unix cluster
Attached are:
Input file for the simple Square plate I have been practicing with
Input file for the Notched beam I am using as my actual model
2 Graphs showing the results from my square trial
| Attachment | Size |
|---|---|
 FullGraph.png | 69.3 KB |
| ZoomedIn.png | 82.54 KB |
 NotchedBeam-FL.inp_.txt | 136.32 KB |
| Square-FL.inp_.txt | 1.97 KB |
1) You're using smooth step
1) You're using smooth step amplitude definition (automatically, it's variable rate), where as mass scaling works as linear change in rate. In your case, it'll be 30*t^2 (1-t)^2. So there is no way you can have apple to apple comprasion here.
2) In both of your plots, you see (atleast) on case deviating, and a vertical line in the force deflection curve. This means either of these two things, (i) that particular solution has quite decent dynamics, or (ii) the solution has broken :P. I don't know how you can judge a quasi static analysis without looking at (kinetic energy, velocity and accelerations). Read the point 3 below.
3) Refer to any good dynamics book (and also abaqus users manual) to understand what exactly you are doing, and what exactly you want to do. What is the meaning of mass scaling, and how it's related toquasi static analysis, what is quasi static analysis, etc.
------
The world started with 0, is progressing with 0, but doesn't want 0.