An optimized estimation of AR model parameters with inhibiting spectrum deviation

Abstract

To solve the problem of spectrum deviation brought about by existing AR parameter estimation methods in engineering application, this paper proposed an improved and optimized estimation method for parameters of the AR model with constant parameters. The method was figured out as followed. Firstly, the initial estimation of AR parameters was derived according to LS criteria; Secondly, through the parametric spectrum estimation formula and the consecutive function's extreme theorem, the mathematical constraint formula was derived with which a combined objective function was constructed under the guidance of punishment function idea; Lastly, the Genetic Algorithm was adopted to optimize the LS estimation of AR parameters. The method proposed was used to estimate spectrum of the vertical vibration acceleration voltage data in turning condition. The results demonstrated that the method proposed overcome such problem as the spectral spectrum deviation, characteristics extraction was more precise and inhibited side-lobes well.