In-Space Servicing Mission Simulations Using Mathematical Surrogate Models

On-orbit space vehicle servicing mission operations are complex. These sophisticated maneuvers in the space (0-g, vacuum) environment where mechanisms and structures can couple and exhibit unexpected adverse dynamic characteristics. The development of a satellite servicing system is challenging as the system operations cannot be fully verified and validated through ground testing in a 1-g environment. It can be difficult to replicate the space environment through ground testing to correlate the space operations. Hence, high fidelity analytical simulations are important to accurately predict space vehicle servicing operations and develop a validated and verified mission. On-orbit space vehicle servicing mission simulations use design reference scenarios defined by the mission operations systems team and attempt to develop mission data that would allow the team to not only design the servicing mission vehicles, but also contribute to the verification of the mission operations. Utilizing high fidelity simulations, we can generate data that would allow us to perform a comprehensive statistical analysis which verifies reference mission operations and considerably reduces mission risk. Predicting high risk mission operations as well as quantifying how sensitive the system operations may be to key mission parameters is invaluable. In this study, the space vehicle servicing reference mission simulation is performed using a multi-body dynamics (MBD) simulation and a combination of rigid and flexible bodies. Complex simulations are time consuming, but the results are important for designing successful mission operations. This effort focused on reducing computation time using Reduced Order Models (ROMs), while retaining key characteristics of the model. These mission operation simulations require thousands of runs, which generate enormous amount of data that needs to be post-processed to be able to verify and validate design reference missions. Reduced Order Models (ROMs) based on response surface models and neural network methods is recommended and is formulated from a series of randomized MBD simulations. Monte Carlo simulation is performed using the mathematical surrogate models to accelerate simulation time without losing fidelity of the simulation results. The main focus of the study is to define process steps and assess the accuracy of the mathematical surrogate models to demonstrate benefits and shortfalls. It is important to understand the value of the proposed techniques and how they affect the analytical conclusions of the space vehicle servicing operations.