A novel differentiator: A compromise between super twisting and linear algorithms
Résumé
Based on the frequency argument, a novel secondorder sliding mode differentiator with a variable exponent is proposed in this article. The super twisting differentiator(exponent = 0, 5) is not sensible to perturbation but its accuracy isdegraded when the signal is affected by the noise. The linearobserver (exponent= 1) has better property in the presence of noisebut is less robust to perturbations. The goal of this paper isto propose a trade-off between the exact differentiator andlinear observer. To reach this objective, the exponent parameter ismade variable. In the absence of noise exponent goes to 0; 5 andtends to 1 when the noise increases. In free-noise case andwith or without perturbation, the novel differentiator behavesas a super twisting differentiator (exact differentiation). Whenthe signal is affected by noise, only a practical stability of thedifferentiator is ensured. Finally simulation results are givento show that the novel differentiator has better performancescompared to differentiators having exponent fixed.
Domaines
Sciences de l'ingénieur [physics]Origine | Fichiers produits par l'(les) auteur(s) |
---|