A Nonlinear Friction Identification Method Combining Separable Least Squares Approach and Kinematic Orthogonal Property

This paper proposes a nonlinear friction identification method for friction compensation in servo-driven mechanisms. The proposed identification method captures the nonlinear Stribeck effect at low velocity which exacerbates the degradation on accuracy performance. Without the requirement of closed loop scheme, the proposed method approximates the friction along a specific stroll with Stribeck model in least squares sense, utilizes a separable least squares (SLS) approach for nonlinear optimization, uses the kinematic orthogonal property (KOP) to avoid inertial force estimation, and takes the features: (1) Computational cost is considerably reduced compared with conventional least squares approaches. (2) It identifies the friction model parameters without any inertial information, and noise introduced by acceleration estimation is avoided. (3) It can be implemented under open loop scheme which does not require controller tuning like most of other identifications do. Component-wise sensitivity analysis is also applied to evaluate the derivation of the model parameters. Additionally this identification method of combining SLS and KOP can also be generalized to other nonlinear mechanical systems while keeping these favorable features. The proposed method is implemented on the friction compensation in the feed drive control of a machine tool. The accuracy of the identified Stribeck model is evaluated and the compensation performance with implementation feasibility considered is validated by experimental comparisons.