Nonlinear Adaptive Filtering Disturbance Observer for Real-Time PMSM Drives
8th Global Power, Energy and Communication Conference, GPECOM 2026, Naples, İtalya, 3 - 05 Haziran 2026, ss.362-367, (Tam Metin Bildiri)
- Yayın Türü: Bildiri / Tam Metin Bildiri
- Doi Numarası: 10.1109/gpecom70462.2026.11578929
- Basıldığı Şehir: Naples
- Basıldığı Ülke: İtalya
- Sayfa Sayıları: ss.362-367
- Anahtar Kelimeler: Adaptive filtering, disturbance estimation, disturbance observer, embedded motion control, nonlinear innovation shaping, servo drive systems
- Yıldız Teknik Üniversitesi Adresli: Evet
Özet
Disturbance observers are widely used in industrial servo systems to compensate for unmodeled dynamics, load variations, and mechanical uncertainties. However, practical innovation signals contain not only actual disturbance variations but also measurement imperfections, encoder quantization effects, and high-frequency torque ripple components. In conventional fixed-gain disturbance observers, these components are injected into the correction dynamics with constant sensitivity. Thus, increasing the observer bandwidth improves disturbance-tracking speed but also increases the transmission of parasitic high-frequency components into the disturbance estimate. To address this limitation, this paper proposes an adaptive filtering-based disturbance observer (AFDO) with nonlinear innovation shaping. Rather than relying on explicit adaptive gain laws, the proposed method realizes adaptation through an innovation-amplitude-dependent effective correction mechanism. Accordingly, small-amplitude ripple-dominated residuals are corrected with reduced sensitivity, whereas larger disturbance-related variations retain strong bounded correction authority. A rate-limited disturbance update is further incorporated to constrain abrupt sample-tosample variations under high-bandwidth operation. Experimental results on a real-time permanent magnet synchronous motor (PMSM) drive show that, compared with a non-adaptive baseline at the same nominal bandwidth, the proposed observer provides a smoother and more consistent disturbance estimate without compromising speed-tracking performance. The smoother iq-axis current response further indicates a better-conditioned control effort under varying mechanical loads. Overall, the proposed strategy improves disturbance-tracking robustness, reduces ripple-induced correction activity, and provides a practical observer structure for embedded servo-drive applications.