It’s a push-pull problem space engineers know all too well: On the one hand, they need detailed models to obtain accurate results. On the other, they need to move quickly, ever faster than before.
Is it possible to find a balance? Perhaps.
But what happens when even the happy medium of adequately fast and adequately accurate is not good enough?
Spacecraft thermal modeling with Simcenter 3D Space Systems Thermal
Traditional computational methods cannot efficiently handle the complexity of detailed CAD models. The standard workaround to date has been to use abstraction. But this comes at a cost.
Although simplifying model geometry speeds things up, it leads to less accurate results. In other words, you might be moving faster, but you risk being off target.
For New Space, the stakes are too high to compromise. New engineering solutions are in order.
GPU technology: an accurate, high-speed solution
Compared to central processing units (CPUs), graphics processing units (GPUs) can perform parallel computations extraordinarily quickly.
In fact, GPUs are ideal for thermal modeling radiation calculations. GPU-based ray tracing methods make it possible to work directly with CAD to generate your detailed mesh and minimize the simplification needed by using GPU for radiation calculations.
It’s no longer a question of choosing to be either fast or accurate. With Simcenter 3D Space Systems Thermal, GPU technology lets engineers reduce computation time and maintain exceptional accuracy.
GPU vs CPU: Key benefits
- Speed: Perform calculations up to 400 times faster
- Accuracy: Work directly with CAD to generate your detailed mesh and minimize the simplification needed by using GPU for radiation calculations
- Efficiency: Achieve more design iterations and optimizations within the same timeframe
Read on to learn about the real improvements GPU technology can deliver in each of these areas.
What are the real calculation speedups?
Surface-to-surface view factor
The table below, based on real-world customer models, shows the speedup in calculation when using GPU for surface-to-surface view factor calculation, with grey body properties.
| Model | CPU Time (s) | GPU Time (s) | Speedup |
| Hemicube 8 cores | RTX 2080 TI | ||
| Model 1 | 68,050 | 1,451 | 46x |
| Model 2 | 44,436 | 1,307 | 34x |
| Model 3 | 118,406 | 297 | 399x |
| Model 4 | 127,233 | 311 | 409x |
Surface-to-surface view factor calculation, with grey body properties
Orbital fluxes (solar, planet IR and albedo)
The calculation of orbital fluxes (solar, planet IR and albedo) for space applications is another potential bottleneck.
Below are the results of a comparison run on two models.
Camera with orbital heating
- 279,733 elements
- 12 positions
- 2 enclosures
- Specular/transparent effects
| VUFAC-VFRTGPU time | Total elapsed time | |
| 12 CPU Deterministic 2 Subdivision | 9 h 02 min | 9 h 44 min |
| GPU 20k rays | 16 min (962 s) | 58 min 38 s |
| Speedup | 34x | 10x |
Full satellite model with orbital heating
- 134,721 elements
- 9 enclosures
- Rotational articulations for solar panels
- Specular/transparent effects
- Simplifications: temperature constraint on whole model for quick solver run time
| VUFAC-VFRTGPU time | Total elapsed time | |
| 40 CPU Deterministic 3 Subdivision | 9 h 22 min | 13 h 36 min |
| GPU 10k rays | 38 min | 1 h 49 min |
| Speedup | 15x | 8x |
Accuracy gains with the Monte Carlo approach
Approximate uniform illumination
The GPU method does not use the uniform illumination approximation that Deterministic and Hemicube use.
Therefore, it is more accurate as shown below.
Fence problem
The GPU method effectively avoids the “fence problem,” in which radiation leaks into or out of a model if an element is in a T junction between two enclosures, as shown below.
Greater efficiency with ray tracing
When it comes to handling specular or transparent effects, the GPU delivers even more speed.
Below are the results and a summary of the calculation times for a teapot model built using 168,929 elements.
| VUFAC-VFRTGPU Ray Tracing time | |
| CPU based MonteCarlo | 3681 sec |
| GPU (RTX 3080 TI) | 30 sec |
| Speedup | 122x |
Discover the benefits of GPU-based radiation calculation with Simcenter 3D Space Systems Thermal
By bringing GPU-based radiation calculations into the thermal modeling workflow, engineers can create high-resolution models directly from CAD data and significantly reduce setup and computation times.
This new approach improves the accuracy of thermal predictions and accelerates the overall design process so space engineers can develop spacecraft faster and more efficiently.
Learn more about Simcenter 3D Space Systems Thermal
Contact a Maya HTT expert today