As LS-DYNA model sizes continue to grow, there is a greater need to reduce the file sizes or rather the model footprint. If good modeling practices are maintained, there can be tremendous file size reductions even for large models and this post will review some of the best practices aimed to reduce the model input/output sizes.
1. Set Definitions
Several options exists in LS-DYNA to create sets of nodes, elements, etc that are referenced by other entities. It is always a good practice to refrain from defining keywords that refer entities at node or element level. Avoiding the direct reference to nodes and elements not only helps to reduce the file size but they also help us to in making modeleling changes. Consider a case of a rigidwall definition which requires a node set to be associated as slave nodes. By picking individual nodes in the model , unless they are sequentially numbered, the resulting SET_NODE_LIST will consist of a large number of input lines. Any future updates to the model, will rely heavily on the pre-processing software to auto update the updated node from the sets. An alternative and concise solution to address both the process and the size would be to define a SET_GENERAL or a just a simple 3-dimensional box using DEFINE_BOX that the rigidwall can refer to determine the slave nodes at runtime during the initialization process. The same guideline can be used to define node or element entities that are referenced by keywords such as *CONTACT, *AIRBAG, *INTIAL , etc. to help achieve a more elegant definition of entities.
2. Reducing D3HSP file size.
The default output into D3HSP files though desiged to aid improved debugging during early model development, the excess data can become an overhead for production models. The following three parametes can help to reduce the overall size of the D3HSP output from LS-DYNA:
NPOPT (default =0 ). The default option allows to output the initialized nodal coordinates, element connectivity, rigidwalls and initial velocity informations. This output can dominate the overall size of D3HSP and can be set to 1 to suppress the output.
NEECHO (default = 0). As in NPOPT, NEECHO controls the output of similar information but to the standard output such as screen. If the output from LS-DYNA is stored to a file, then it could be large in size. It is recommended to set this option to 3 to suppress both nodal and the element output.
IPNINT (default = 0). This parameter controls the option to ouptut the initial timestep computed by LS-DYNA. By default, only the first 100 smallest timestep controlling elements are included. However, in some cases, it may be helpful to ouptut for all of elements in the model by setting it to 1. If not in debugging phase, it is recommeneded to leave the value to 0.
3. Storing Re-usable Component
Redundant data is a big contributor for increasing the simulation footprint as more design studies are made. A simple example is a FMVSS 201 regulation which requires the valiation of a vehicle against multiple impact locations. The analyses is performed on the structure using a validated and calibrated free-form head model which is then impacted at more than 40+ impact locations on the vehicle. Embedding the head model in all of these impact locations causes duplication of data that in result increases the database size. A better approach would be to store re-usable components in a central location and use *INCLUDE file to embed in the required model. To facilidate renumbering and/or repositioning of validated components, INCLUDE_TRANSFORM can be used. More recent improvements in LS-DYNA to embed reusable components is the specification of include paths using *INCLUDE_PATH option where a list of active directories can be specified which eliminates the need to include the absolute path of the included file. This is shown in the following figure
4. Load curve based D3PLOT output
6. Component based D3PART output
7. Use of LSPOST Database or other compression software such as FEM-Zip