Development of inversion-based feedforward-feedback control techniques for advanced manufacturing.

摘要

Advanced manufacturing aims to make high-quality products at low cost with high efficiency and highly integrated/controlled processes, thereby, promoting the process integration and providing accommodation of customized and cost-effective miniaturized products. With the increasing demands on product precision and cost efficiency in micro- and nano-scale manufacturing, the development and implementation of control technologies have become an indispensable part of advanced manufacturing. However, challenges exist in the process control of micro- and nano-scale manufacturing. The system dynamics, in general, is complicated and can be excited when the micro- and nano-manufacturing are conducted at high speeds, and other adverse effects including the hysteresis and creep effects of the actuators further complicates the precision control of the manufacturing system. Additional challenges also arise from the variation/uncertainty and environmental disturbances. It has been demonstrated that micromanufacturing could benefit from the augment of ultrasonic vibration in achieving lower power consumptions and elongated tool life. However, the fundamental mechanism of ultrasonic vibration effect on micromanufacturing has not yet been understood. Similar challenges also exist in probe-based nanomanufacturing as the patterning throughput is ultimately limited by the patterning speed, which, in turn, is limited by the vibrational dynamics and hysteresis effect, as well as the cross-axis dynamics coupling effect of the actuation system. Further challenge arises when patterning directly on hard materials using probe-based approach --- even with stiff probe of hardest material, the pattern obtained on hard sample such as tungsten is hardly of any practical usage (feature depth < 0.5 nm). These challenges in micro- and nano-manufacturing motivate the research in this dissertation.;In this dissertation, the dynamics and hysteresis effect are studied and addressed for the magnetostrictive actuator-based ultrasonic-vibration-assisted microforming process and probe-based nanofabrication with an atomic force microscope (AFM). In particular, a magnetostrictive actuator-based mechatronic system is developed for the ultrasonic-vibration-assisted microforming process. The modeling-free inversion-based iterative learning control method (MIIC) is utilized to compensate for the dynamics and hysteresis effect on the ultrasonic vibration generation across a large range of working frequency. The Fibonacci method is utilized to rapidly identify the resonant frequency online for more pronounced ultrasonic vibration effect. To address the backlash and relatively low resolution of the DC-motor, a bulk motion actuation system is designed and fabricated with a mechanical amplification around a magnetostrictive actuator. Such a design allows the bulk motion for large output force and motion stroke with high resolution (< 1 um). The entire microforming process is divided into pre-welding and welding phases. During the pre-welding phase, the data-driven, modeling-free differential-inversion-based iterative control (MFDIIC) approach is developed to address the dynamics and the hysteresis effect of the magnetostrictive actuator for high efficiency. The inversion-based optimal output tracking-transition method is employed to realize the accuracy transition from the pre-welding to the welding phase, and thus improves the product quality. In the study of the probe-based nanofabrication, the MFDIIC method is also utilized and integrated to address the adverse dynamics effect and the hysteresis behaviour of the piezoactuators. An ultrasonic vibration is also augmented in the driving of the piezoactuator in z- axis to increase the impact of the probe and enables the patterning on hard materials.;The MFDIIC technique is further analyzed and theoretically proved of its efficiency in compensating for both of the dynamics and nonlinear hysteresis effects with no needs for modeling hysteresis and/or dynamics, and achieve both precision tracking and good robustness against hysteresis/dynamics changes. The convergence of the MFDIIC algorithm is analyzed with random output disturbance/noise considered. It is shown that precision tracking can be achieved with the tracking error close to the noise level in the statistical sense.