基于滑动多项式回归分析的区间估计和飞行数据异常值处理

针对飞行数据中的异常值给出了一种处理算法，对其进行识别、剔除与补正。根据飞行数据的特点建立了滑动二次多项式回归模型，并变换为正交多项式计算回归系数。在给定的置信概率下求下一个点的置信区间，判断实测数据是否为异常值。将已判定的异常值剔除并用拉格朗日法补正。在某型共轴式直升机的动力学特性研究工作中，利用这一方法对多次外场试飞数据进行了异常值处理，得到了比较满意的结果。

The Interval Estimate and the Process on Abnormal Value of Flight Data Based on Moving Poly-nomial Regression Analysis

The algorithm to process the abnormal value in the flight data is given,which can iden-tify,eliminate and correct the abnormal value.According the characteristics of the flight data,the moving quad-ric polynomial regression model is established and is transformed to the orthogonal polynomial to calculate the regression coefficient.Then,the confident interval of the next point in the given probability was calculated to judge if the test data is abnormal value.Finally,the conformed abnormal value was corrected with the Lagrange algorithm.In the research on some coaxial helicopter dynamics,the abnormal value of the dozens of outfield flight data is processed with the algorithm and satisfactory results are acquired.