Modeling zero-bias drift in a fiber-optic gyroscope using Dynamic Mode Decomposition (DMD)

This study presents a computationally efficient approach to modeling the temperature-induced zero-bias drift of a fiber-optic gyroscope for subsequent use in error-compensation algorithms of inertial navigation systems. The thermal behavior of the fiber coil was investigated using three-dimensional transient numerical simulations that account for anisotropic heat transfer. The numerical model was compared with climate-chamber test data obtained by recording temperatures at several locations within the gyroscope together with the corresponding zero-bias drift. An integral thermal metric was introduced to characterize the accumulated temperature variation over the volume of the optical-fiber coil. Its temporal evolution was shown to reproduce the main increasing and decreasing trends observed in the measured zero-bias drift. Linear, quadratic, and cubic regression models were considered to establish a quantitative relationship between the thermal state and the drift. The quadratic model provided the best compromise between accuracy and model complexity. To reduce computational cost, the full temperature field was replaced by a reduced-order representation whose dynamics were reconstructed using dynamic mode decomposition with external thermal inputs. The resulting reduced-order model closely reproduced the integral thermal response of the full numerical model while retaining sufficient accuracy in zero-bias drift estimation. The coefficient of determination was 0.9903 for reconstruction of the integral thermal metric and 0.9788 for drift estimation using the quadratic regression model. The computation time per iteration was reduced from 13.279 ms for the full numerical model to 0.019 ms for the reduced-order model. These results demonstrate the feasibility of the proposed approach for implementing an onboard temperature-compensation algorithm for fiber-optic gyroscopes.