基于Levenberg-Marquardt算法的遥测数据建模预测

Modeling and Forecasting of Telemetric Data Based on the Levenberg-Marquardt Algorithm

大型航天电子设备的遥测数据通常呈现不规则周期,对其建模并做出长期预测可以尽早及时发现设备性能异常.文中利用SumSin模型,采用FFT变换分析求解出模型参数初始点;分析指出调整阻尼参数μ是实现Levenberg-Marquardt优化方法快速收敛的关键,给出了其求解模型参数的详细算法;最后通过对两种类型遥测数据快速地求取模型参数和实现长期预测,说明了算法的工程有效性.

The telemetric data of the large-scale spaceflight electronic equipment usually appears the irregular period ,and the performance abnormity would be discovered early through modeling and prediction on these data .The mathematic SumSin model was used in the paper and its initial parameters were acquired by the FFT transform technique .The excellent characters of the Levenberg-Marquardt （L-M ） method were analyzed theoretically ,in w hich the key is to adjust the damping parameter μ.T hen the detail implementation of the L-M arithmetic was introduced .At last it was realized that fleetly calculating the model parameters and long-range forwards predicting on two types telemetric data ,which shows that the L-M arithmetic in the paper has strong ability in engineering practice .