Are you excited about the future of chips and the high-tech motion control industry? Would you like to combine and bring further the latest advances in machine learning and vibration control to improve current mechatronic system designs? Join us in this endeavor where data, machine learning, advanced vibration control, and high-tech mechatronics come together.
Information The semiconductor roadmap is driven by throughput, accuracy, and costs. These drivers put ever stringent requirements on the high-tech mechatronic systems needed to produce chips, e.g., lithography machines like wafer scanners, but also wafer inspection tools and semiconductor assembly tools. All these tools consist of highly sensitive optical and measurement systems on the one hand, and tracking systems that support aggressive motion on the other hand. Needless to say that in these tools the undesired excitation of structural dynamics is to be avoided as much as possible. The resonances that remain being excited are often damped by the use of passive tuned-mass dampers or TMDs. The mass, stiffness and damping of these TMDs are matched with the properties of the structure that needs to be damped, and hence every module or functional component in these complex machines requires its own dedicated design. For the next generation of semiconductor equipment we envision active solutions that use a control loop based on measurements and/or predictions of the resonances with actuators to counteract the resulting vibrations. By exploiting the data available from online measurement through machine learning methods, these active solutions should become self-learning and able to adapt their dynamical properties as to withstand a priori unknown dynamic disturbances that potentially induce repetitive as well as non-repetitive vibrations.
Here is where you come in! Within the EAISI project MASC, which stands for hybrid and modular modelling and control for complex, flexible and interconnected systems in lithography applications, you will join the Dynamics and Control section at the
Mechanical Engineering Department and help design, analyze, and test semi-active TMD systems with data-driven control strategies for active vibration control. That is, you will:
- Conduct a detailed literature study on active TMDs, noise/disturbance cancellation, and machine learning techniques within the scope of the problem.
- Formalize the constrained optimization problem in which energy dissipation in the TMD through control is maximized while satisfying constraints on maximum gains, stability margins, and limitations of the hardware, e.g., actuators, sensors, and its electro-mechanical components.
- Develop machine learning control strategies that through disturbance feedforward control and/or parameter-varying feedback control provide the ability to adapt the TMD dynamical properties on the spot as to deal with the disturbance situation at hand without compromising robust stability properties of the controlled system.
- Design (or modify an existing) semi-active TMD system to validate the developed control strategies and to test the optimization objective in meeting vibration cancellation specifications.