Online Parameter Estimation Using SR-UKF in Turning of Slender Bar


Farbod Akhavan Niaki


Laine Mears


Automotive Engineering (College of Engineering and Science)


In this work square root unscented Kalman filter (SR-UKF) for online estimation of process parameters is proposed. UKF may diverge in some nonlinear processes with large difference in order of magnitude for covariance matrices. In this case taking square root of covariance matrix is suggested. SR-UKF can accurately predict the process parameters with 5.92 root mean square error. While it has been suggested by some authors that the feed exponent in nonlinear force feed equation is constant, the results shows change in the mean values of parameters including feed exponent. In addition, the filter estimates a stationary covariance between parameters, which indicates the constant correlation of parameters through the cutting time.

Presented by:

View other submissions to GRADS 2014

Nominate a Student
Judge an Event
Review Submissions