结合多体动力学先验知识与核方法的位标器性能预测技术

位标器的性能与特征参数之间存在着非常复杂的非线性映射关系，且由于装调工艺复杂，测量手段昂贵，位标器实际测量数据较少。如何在小数据的情况下实现对位标器性能的精确求解，是准确预测位标器装调性能的关键。提出一种结合陀螺动力学先验知识和核方法的位标器装配性能预测方法(A-LPSVR)。通过构建陀螺Adams动力学模型以缩小解空间范围。使用仿真性能值和真实装调性能值的差值训练支持向量回归模型，得到最终的位标器性能预测模型。实验结果表明该方法所构建的模型预测准确性与泛化性能要优于现有模型，能够在测量数据稀缺情况下实现位标器陀螺仪性能的精确预测。

The Performance Prediction of Coordinator Based on Prior Knowledge of Multibody Dynamics and Kernel Method

There is a very intricate nonlinear mapping relationship between the performance of the coordinator and the characteristic parameters. Because the assembly process is complicated, the measuring method is expensive, and the real measured data of the coordinator is insufficient. How to achieve accurate solution of the performance in the case of small dataset is the key to predict the performance of the coordinator precisely. This paper proposes a method for predicting the assembly performance of the coordinator (A-LPSVR) combining the prior knowledge of gyrodynamics and the kernel method. The gyro Adams dynamics model is constructed to narrow the solution space. The support vector regression model is trained using the difference between the simulated performance value and the real measured performance value to obtain the final coordinator performance prediction model. The experimental results show that the proposed model has better prediction accuracy and generalization performance than the existing algorithms, and can accurately predict the performance of the gyros under the condition of measured data scarcity.