Simulation Data Management in Aerospace
The aerospace industry has been a large and important user of virtual simulation methods over many decades. Finite element structural and thermal analysis, computational fluid dynamics, electromagnetic simulation and other methods are routinely used in the design and analysis of the airplane, helicopter, rocket, jet engine, or satellite. The aerospace industry was an early adopter of product lifecycle management (PLM) systems to manage the design, manufacturing, assembly, and maintenance of the aerospace system. The industry is now looking for ways to manage their simulation data.
There are some important differences between managing design/manufacturing data and simulation data. For design data, the enterprise needs to keep lots of files that are relatively small in size. For simulation data, the enterprise needs to keep a much smaller set of files but the file sizes can be several orders of magnitude larger. As computers get more powerful, the data size of simulation models is growing rapidly to achieve even more accurate results. Companies are also doing more stochastic or design of experiments analysis with their simulation data to better understand the effect of manufacturing variability on performance and often these studies are done on HPC (High Performance Computing) systems. The need for interfacing to such HPC systems to address the workflow for job submissions and remote visualization becomes important to manage the large amount of simulation data. All of this is driving the data size of simulation to enormous levels and the size of the data for simulation can cause significant problems when you try to manage your simulation data in your PLM system which might not be setup to handle the data explosion from CAD to simulations.
Much of the data in the PLM system is needed for the simulation but it is often not structured in the PLM system in a way that is best utilized for simulations. Simulation often ignores many of the fine details in the geometry, such as fillets, corner radii, holes, and fasteners but it can be difficult to filter this data out of the PLM system so that the simulation engineer just gets what they need. On the other hand, there is information needed for the simulation, such as material properties, that is likely not stored in the PLM system.
For most types of simulation, the geometry is discretized into small elements.The discretization requirements can be different for different types of analysis and the refinement of the discretization can have a significant influence on the accuracy of the simulation. One of the issues that surfaces in addressing simulation data management requirements is deciding what to store. Do you store each of the discretized models and the results files for each simulation, do you just store the discretization parameters, do you store the simplified geometry you used for the analysis system, or do you just store the simulation report? These decisions can have a large influence on the data management system you set up. Storing the input file can document all of the parameters used in the simulation but it doesn’t document why the analyst chose to set up the model in a particular way. Some of this information is likely contained in the analysis report but the simulation data management system should provide the ability for the analyst to document modeling methods, strategies, and reasoning in the model set-up. Companies may also want to incorporate a knowledge management system in conjunction with the simulation data management to document modeling methods so the learning can be better captured and utilized for future simulations.
For certification issues, failure analysis, or accident investigations that may occur years after the analysis was carried out, there may be a need to re-run the analysis. In addressing this requirement, many also need to consider keeping the hardware and software systems around that you used for the original analysis. At a minimum, this information all needs to be recorded in the simulation data management system so that you can at least try to setup virtual machines with the appropriate operating system and necessary simulation software to duplicate the environment after procuring the necessary licenses. For aircraft seating and occupant simulation, the industry is moving towards a heavy reliance on simulation and less reliance on physical tests so the certification agencies will likely have their own set of requirements for storing data which will need to be incorporated into the data management system.
It is common in aerospace to have one simulation dependent on the results of another simulation. For example, structural loads used to analyze a section of the airplane come from an aerodynamic analysis that was performed with computational fluid dynamics software. Another common case is the structural loads on a component coming from a global structural model of the airplane. For the simulation data management system, you also need to address documenting from where you are getting the loading conditions and if it is from another simulation model, where this model is stored, how you extracted key performance indicators and metrics from this model, and which version was utilized to allow traceability and consistency across engineers.
Today, simulation data management is usually carried out by file naming conventions or spreadsheets in common hard drives with some backup/archiving. Although this provides some data management capabilities, especially for small analysis groups, it does not allow the enterprise to take full advantage of all that simulation can provide the organization in the form of traceability, reusability and capturing best practices. In the same way that PLM provides significant advantages to the enterprise, simulation data management can provide similar benefits but it will require systems that meet the unique needs of the simulation teams and their management.