Machine learning based fault tolerant thrust vector control of a small launch vehicle

Date

2025-1-02

Author

Gök, Gökberk

Metadata

Show full item record

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Item Usage Stats

80
views

80
downloads

Cite This

The methodology proposed in this thesis integrates Deep Neural Network-based fault detection and isolation techniques with the Weighted Least Squares control allocation to actively counteract thrust vector control actuator faults such as lock-in-place, hard-over, loss of effectiveness, and float for a launch vehicle. This integration allows the system to maintain accurate trajectory control and prevent excessive loads on the system immediately after the event of a fault. This is achieved through a deep neural network-based fault diagnosis and isolation model that quickly and accurately identifies the fault, combined with the online reconfigurability of the Weighted Least Squares control allocation. Together, these features effectively overcome the challenges posed by faults. The focus here lies on the small launch vehicle Firefly Alpha, which serves as a representative model for testing the proposed active fault-tolerant thrust vector control system’s performance and capability. A high fidelity 6-degree-of-freedom model for the launch vehicle is built and used for both the simulations and analysis for training the deep neural network model to detect and isolate faults. Finally, Monte Carlo analysis is performed to evaluate the performance of the proposed method.

Subject Keywords

Active fault tolerant control, Weighted least square, Control allocation, Machine learning, Neural network, Launch vehicle, Ascent phase

URI

https://hdl.handle.net/11511/113066

Collections

Graduate School of Natural and Applied Sciences, Thesis