Model identification, updating, and validation of an active magnetic bearing high-speed machining spindle for precision machining operation

High-Speed Machining (HSM) spindles equipped with Active Magnetic Bearings (AMBs) are envisioned to be capable of autonomous self-identification and performance self-optimization for stable high-speed and high quality machining operation. High-speed machining requires carefully selected parameters for reliable and optimal machining performance. For this reason, the accuracy of the spindle model in terms of physical and dynamic properties is essential to substantiate confidence in its predictive aptitude for subsequent analyses.;This dissertation addresses system identification, open-loop model development and updating, and closed-loop model validation. System identification was performed in situ utilizing the existing AMB hardware. A simplified, nominal open-loop rotor model was developed based on available geometrical and material information. The nominal rotor model demonstrated poor correlation when compared with open-loop system identification data. Since considerable model error was realized, the nominal rotor model was corrected by employing optimization methodology to minimize the error of resonance and antiresonance frequencies between the modeled and experimental data.;Validity of the updated open-loop model was demonstrated through successful implementation of a MIMO micro-controller. Since the micro-controller is generated based on the spindle model, robust levitation of the real machining spindle is achieved only when the model is of high fidelity. Spindle performance characterization was carried out at the tool location through evaluations of the dynamic stiffness as well as orbits at various rotational speeds. Updated model simulations exhibited high fidelity correspondence to experimental data confirming the predictive aptitude of the updated model. Further, a case study is presented which illustrates the improved performance of the micro-controller when designed with lower uncertainty of the model's accuracy.