Parallel-in-time Disturbance Region Update for Flight Vehicle Dynamics
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for predicting the dynamic characteristics of flight vehicles are inefficient due to the sequential advancement of solutions in time, limiting the potential for parallel processing and leading to significant computational burdens.
Innovation Solution
A parallel-in-time disturbance region update method is introduced, which utilizes dynamic computational domains and the Chebyshev time pseudo-spectral method to enable simultaneous solution of multiple time layers, thereby achieving time parallelism and reducing unnecessary computational efforts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the dual time-stepping method based on second-order backward difference is used for predicting dynamic characteristics, then the prediction accuracy is improved, but the computational efficiency deteriorates due to sequential time advancement
Solution Approach 1:
The patent transforms the computational approach from sequential time-marching (1D time dimension) to parallel time-spectrum computation by introducing Chebyshev polynomial expansion in the time dimension. This allows multiple time layers to be computed simultaneously rather than sequentially, resolving the contradiction between accuracy and efficiency.
Solution Approach 2:
The patent replaces the mechanical sequential time-stepping mechanism with a spectral method based on Chebyshev polynomials. Instead of advancing solutions step-by-step through time, the method uses polynomial interpolation and transformation to compute multiple time layers in parallel, eliminating the sequential bottleneck while maintaining second-order accuracy.
2Productivity
If Chebyshev time pseudo-spectral method is used to enable time parallelism, then the computational speed is improved, but the method fails to exploit full efficiency due to static computational domain and lack of disturbance region update
Solution Approach 1:
The patent applies local quality by identifying and updating only the disturbance regions where flow changes occur, rather than computing the entire static domain. The computational domain is dynamically adjusted based on local flow characteristics, concentrating computational resources in regions that require updates while reducing effort in converged regions.
Solution Approach 2:
The patent introduces dynamic computational domains that adapt during the simulation. The disturbance region update mechanism dynamically identifies areas requiring computation based on flow gradients and changes, making the computational domain flexible rather than static. This allows the method to exploit full efficiency by adjusting the computational workload to match the actual physics of the problem.
3Measurement precision
If implicit time integration is applied to achieve second-order accuracy, then the prediction precision is improved, but the device complexity increases due to multiple acceleration techniques
Solution Approach 1:
The patent merges the Chebyshev time pseudo-spectral method with the dual time-stepping method and disturbance region update technique into a unified computational framework. By combining these methods, the patent achieves second-order accuracy through implicit time integration while using the Chebyshev-based time parallelism to manage the complexity, ultimately reducing the overall computational burden despite the increased algorithmic sophistication.
Data Source
AI summary
A parallel-in-time disturbance region update method for dynamic characteristics of flight vehicles includes reading-in the data; obtaining the components of Chebyshev transformation matrix; initializing the flow field; establishing a dynamic computational domain; solving the flow-governing equations of all time layers simultaneously in the dynamic computational domains; updating the dynamic computational domain; judging whether the parallel computation in the current time period converges and whether the computation completes; if so, outputting the results; if the computation converges but not completes, jumping to the step that initializing the flow field, and performing the computation in next time period; if not, jumping to the step that solving the flow-governing equations of all time layers simultaneously in the dynamic computational domains, and entering the next iterative step.