How to get the correct acceleration which can compare with experiment result?
Altair EmployeeHi all,
Based on a drop test simulation, I think we can get one node acceleration by using three ways.
First, get acceleration result directly by using Build Plot about Acceleration.
Second, get velocity result directly by using Build Plot about Velocity, and then do one derivative to get acceleration.
Third, get displacement result directly by using Build Plot about Displacement, and then do double derivative to get acceleration.
From my simple test example, I found the acceleration results with the same node by using the above three methods were very different (the z direction acceleration result of node 7085 listed below method 1: -66~56, method 2: -27~27, method 3: -16~19). The result was figured below.
So my doubt is why so different acceleration results were gotten by using these three methods? Which method result is correct and can be compared with experiment testing acceleration result by using accelerator in experiment?
<?xml version="1.0" encoding="UTF-8"?>
Roy
Answers
-
Hi Roy,
The results in the above cases are plotted from the animation file instead of the time history file T01 file and so they are only getting 500 data points.
In RADIOSS, we calculate the acceleration using acceleration= nodal force / nodal mass and then integrate to get the velocity and the displacement. Therefore I would recommend to plot the acceleration from the T01 file. Also filter the data like would in a test. And make sure a small enough value in /TFILE is used so they get enough data points.
0 -
Altair Forum User said:
Hi Roy,
The results in the above cases are plotted from the animation file instead of the time history file T01 file and so they are only getting 500 data points.
In RADIOSS, we calculate the acceleration using acceleration= nodal force / nodal mass and then integrate to get the velocity and the displacement. Therefore I would recommend to plot the acceleration from the T01 file. Also filter the data like would in a test. And make sure a small enough value in /TFILE is used so they get enough data points.
Hi George,
I have tried your idea, however three different results were still gotten by using three different method I mentioned before even though I have output T01 file with very small increment value (1e-9). I think the node information was output every time step since 1e-9 was used for T01 output. I have uploaded my model. So I think the reason about output increment may be not the reason. what's the reason about this question?
<?xml version="1.0" encoding="UTF-8"?>
0 -
Hi Roy,
The values are not matching , but seems closer. Taking the derivative introduces some aliasing in the data, so the peaks are missed.
We always recommend to use accelerometer to capture acceleration (/TH/ACCEL) and please continue to use this method.
0 -
Altair Forum User said:
Hi Roy,
The values are not matching , but seems closer. Taking the derivative introduces some aliasing in the data, so the peaks are missed.
We always recommend to use accelerometer to capture acceleration (/TH/ACCEL) and please continue to use this method.
Hi George,
1. I have tried with /th/accel to get node acceleration. From the results comparison, the same results were gotten by using /th/accel and /th/node. Check the model and figure below.
<?xml version="1.0" encoding="UTF-8"?>
2. From the radioss theory reference manual, it seems that the displacement is the first solution variable, and then velocity and acceleration are to be derived. This is a little bit different from what you said (you mentioned we calculate the acceleration using acceleration= nodal force / nodal mass and then integrate to get the velocity and the displacement). Which procedure is used in radioss? could you show some validation documents?
<?xml version="1.0" encoding="UTF-8"?>
3. I don't think I have understood your opinion about the not matching results. how to understand the meaning of 'aliasing'?
Roy
0 -
Hi Roy,
Taking derivatives for finding acceleration is not a recommended method, as this introduces some misidentification of a signal frequency and so the peaks are missed.
I consulted regarding this with some experts and they recommended /TH/ACCEL method only for extracting acceleration results.
Regarding this, I don't have any valid documents but I'll search for some and I'll update you if I get some which throw more light to this topic.
0 -
Thank you George. From my test, I got the same acceleration results by using /TH/ACCEL and /TH/NODE. I'm looking forward for your good news about this topic.
0
