jball1
Mechanical
- Nov 4, 2014
- 71
I am attempting to run a dynamic analysis of a large composite structure. I am getting an “Exception_Access_Violation” error when I try to datacheck the model.
The error seems like it is related to a specific composite section definition that I am trying to use.
The structure is represented with conventional shell elements. I first ran the model with a homogenous shell section assigned with a defined anisotropic constitutive matrix. This model runs without issue. I then tried to modify the section definition to a composite section with the ply layup specified (4 plies defined on either side of a core). Making this change resulted in the access violation error.
Before changing the composite section representation in my model, I tried out the modified section representation first on a flat plate and then on a simplified model of the structure. I was able to successfully run both test models. So the issue is not that I’ve screwed up the section definition.
I used a winmerge to confirm that the only differences between my input file of my working model and the one that gives me the error is the composite section definition.
I googled the error to see if anyone else has had this error. Some people suggested that it may possibly be related to a segmentation error and may be related to model size. My revised section definition includes 11 integration points through the thickness vs 5 for my original definition. Since my new section definition is more computationally intensive, I tried modifying the input file from the analysis that ran, changing only a small portion of the structure to the new section definition. I was able to complete a datacheck with this input file.
I then continued adding the new section definition to more and more of the structure, running a datacheck each time. When I was about 2/3 of the way through the structure, I got the access violation error again.
This made me conclude that my error is related to the amount of elements that this more computationally intensive section definition is assigned to. To test this idea out, I significantly coarsened the mesh of the structure (reduced number of elements by ~70%) and tried to assign the modified section definition to these elements. Unfortunately I still got the exception access violation error with this model.
I am running Explicitly. I always run with double precision (double=both). On a whim, I tried the datacheck with single precision, and it completed.
I need to run with double precision. My understanding is that double precision results in larger files and therefore requires more memory? Is there some way I can give the program more memory when it is running the datacheck? I tried reading the documentation, and thought setting pre_memory=”16gb” was the answer… but apparently this parameter is not used for a datacheck.
Any help would be much appreciated!
The error seems like it is related to a specific composite section definition that I am trying to use.
The structure is represented with conventional shell elements. I first ran the model with a homogenous shell section assigned with a defined anisotropic constitutive matrix. This model runs without issue. I then tried to modify the section definition to a composite section with the ply layup specified (4 plies defined on either side of a core). Making this change resulted in the access violation error.
Before changing the composite section representation in my model, I tried out the modified section representation first on a flat plate and then on a simplified model of the structure. I was able to successfully run both test models. So the issue is not that I’ve screwed up the section definition.
I used a winmerge to confirm that the only differences between my input file of my working model and the one that gives me the error is the composite section definition.
I googled the error to see if anyone else has had this error. Some people suggested that it may possibly be related to a segmentation error and may be related to model size. My revised section definition includes 11 integration points through the thickness vs 5 for my original definition. Since my new section definition is more computationally intensive, I tried modifying the input file from the analysis that ran, changing only a small portion of the structure to the new section definition. I was able to complete a datacheck with this input file.
I then continued adding the new section definition to more and more of the structure, running a datacheck each time. When I was about 2/3 of the way through the structure, I got the access violation error again.
This made me conclude that my error is related to the amount of elements that this more computationally intensive section definition is assigned to. To test this idea out, I significantly coarsened the mesh of the structure (reduced number of elements by ~70%) and tried to assign the modified section definition to these elements. Unfortunately I still got the exception access violation error with this model.
I am running Explicitly. I always run with double precision (double=both). On a whim, I tried the datacheck with single precision, and it completed.
I need to run with double precision. My understanding is that double precision results in larger files and therefore requires more memory? Is there some way I can give the program more memory when it is running the datacheck? I tried reading the documentation, and thought setting pre_memory=”16gb” was the answer… but apparently this parameter is not used for a datacheck.
Any help would be much appreciated!