Tom Oomen


Motion systems

Motion systems are a central component in many manufacturing and consumer applications. On the one hand, such motion systems are found in inexpensive consumer electronics, including optical disc drives, harddrives. Also, printing systems are a typical application, both for consumer use and industrial use. On the other hand, manufacturing machines are often equipped with motion systems, e.g., one of the most accurate and expensive motion systems can be found in lithography.

Traditional control of motion systems

Traditionally, the control design for positioning manufacturing machines has been significantly simplified by relying on certain model assumptions. Essentially, traditional positioning machines are assumed rigid and thus internal deformations are neglected. The result of this modelling assumption is that the control design process effectively breaks down into separate decentralized control designs, see Figure 1. Essentially, for each of the six motion degrees-of-freedom (three translations and three rotations), the control design amounts to driving six independent force actuators based on six independent measurements and six independent models.

Thus, in this traditional rigid-body motion control approach, the system mainly behaves as a rigid-body, i.e., a -2 slope in the frequency domain within the control bandwidth. Flexible dynamics are typically seen as high-frequent parasitic effects (and sometimes locally the gain of the controller is reduced through a notch-filter). The loop-gain (i.e., plant times controller) basically looks like

Developments in motion systems: lightweight system designs

Accuracy and speed requirements on motion systems are ever increasing. Increasing accuracy enables the production of smaller products, e.g., smaller transistors in lithography to keep up with Moore's law. Increasing speed enables a higher throughput, which is essential for the market position of the resulting overall system.

Increasing accuracy requirements require a larger control bandwidth, i.e., control has to be effective over a larger frequency range. This leads to the following loop-gain in a Bode-diagram

The increasing speed/throughput requirements affect the system on at least two aspects. These can best be understood from Newton's law
F = m a
First, increasing accelerations are obtained by increasing the force F. This inevitably leads to a larger excitation of flexible dynamics. Second, we foresee lightweight system designs, i.e., decreasing m, to enable throughput requirements. This implies that the flexible dynamics will typically occur at lower frequencies (this is not an immediate consequence, see section 1.3 of Oomen, 2010). This leads to

Our next-generation motion control design approach

Increasing accuracy and speed requirements in next-generation motion systems leads to a situation where flexible dynamics appear in the controller cross-over region. In contrast to traditional motion control design approaches, the internal deformations caused by flexible dynamical behavior cannot be neglected anymore. In our research team, we have addressed several aspects that we believe are of key importance in next-generation motion control:

  • Multivariable control to deal with interaction. Traditional motion system can be directly rigid-body decoupled. However, flexible dynamics are not necessarily aligned with the rigid-body degrees of freedom, introducing coupling. We have developed multivariable motion control design approaches, see, e.g., Oomen et al., 2014 and Boeren et al., 2015 to systematically design controllers for this situation.

  • Overactuation and oversensing to actively compensate flexibilities. Here, we have already successfully compensated flexible dynamical behavior by adding additional stiffness and damping.

  • Inferential control to deal with unmeasurable performance variables. Due to flexible dynamical behavior, a dynamic relation will exist between the point where performance variables are defined and the point of measurement. In contrast, in traditional systems, these relations are considered static due to rigid-body assumptions. To address this, we have developed an inferential control design framework, including appropriate controller structures, identification aspects, and observer design.

  • Position-dependent control due to changing measurement positions, changing performance locations, and/or changing dynamics due to moving masses, etc.. Motion systems inherently move and thus typically lead to position-dependent dynamics. Essentially, this leads to nonlinear system behavior. Luckily, it can typically be captured in the so-called Linear Parameter-Varying (LPV), which from the perspective from position-dependent dynamics may be the next best thing to linear models and controllers. We are developing various aspects of LPV identification and control in view of position-dependent mechanical systems.

Due to flexible dynamical behavior, a dynamic relation will exist between the point where performance variables are defined and the point of measurement. In contrast, in traditional systems, these relations are considered static due to rigid-body assumptions. To address this, we have developed an inferential control design framework, including appropriate controller structures, identification aspects, and observer design.

The basic idea looks like this:

Challenges for identification and control

Model-based control is a key aspect in systematically dealing with the above aspects in our next-generation motion control design framework. This puts stringent requirements on the modeling framework. We typically use system identification, i.e., experimental modeling, since it is fast, inexpensive, and accurate. The increasing requirements necessitate and justify more complex models, i.e., models including high-order lightly damped flexible dynamical behavior. In addition, overactuation/oversensing and inferential control lead to a situation where many more measured variables, manipulated variables, and performance variables have to be distinguised. This requires advanced system identification tools that we are currently developing, see the identification for control page and the numerically reliable identification page.

Further reading

Our vision on next-generation motion control and important contributions in this direction are reported in:

  • Model-based control for high-tech mechatronic systems
    Tom Oomen and Maarten Steinbuch
    In: The Handbook on Electrical Engineering Technology and Systems: Volume 5 Factory and Industrial Automated Systems, CRC Press/Taylor & Francis, To appear

Further details on various topics can be found on the following pages and references there:

Acknowledgement

The above results are in collaboration with many co-workers, both from TU/e-ME and industry.

Note that all figures shown on this page can be found in the mentioned papers. Please follow the guidelines regarding copyright and references when citing these.