The aim of this assignment consisted on re-designing the vertical actuation mechanism in a wafer handling robot so that the thermal input to the wafer is minimized. The robot is intended to transport silicon wafer in a deep-vacuum and clean environment.
To the left the typical SCARA configuration wafer handling robot can be seen. The robot is intended to transport silicon wafers in a deep-vacuum and clean environment.
In this existent configuration, the vertical actuation is performed at the base of the robot. Being the wafer far away from any actuator, it is safe to say that any hazardous thermal input to the wafer is prevented. However, the vertical actuation needs to overcome the weight of the whole robot. A novel configuration is presented at b (Figure to the right), where the vertical actuation is done at the tip of the robotic arm. This could possibly increase the dynamic performance of the robot, and thus increase the throughput of the machine. However, the vertical actuator is now close to the wafer. Then it is paramount to reduce the power dissipated by it.
In an environment like such, is essential to avoid solutions that involve tribology (e.g. bearings) as the particles released could potentially contaminate the wafer.
High positioning accuracy is required for this application. To prevent damaging the wafer, the robot must be able to position the wafer accurately. An elastic guiding parallelogram seems a convenient choice then, as this solution if free of friction and backlash. Additionally, an elastic parallel guide can easily be manufactured from a monolithic block. Consequently, no particles are released into the environment during operation.
The figure to the right schematically shows how a guiding parallelogram behaves. Two leaf springs constrain the payloads in all DOFs but one. The free end of the parallelogram moves in a quasy-straight motion by deflecting the leaf springs.
However, elastic guiding relies on the deflection of flexible components. Then, a parasitic stiffness in the guiding direction needs to be overcome by the actuators. This generates and undesired thermal input to the wafer.
In order to passively minimize the required actuation force, a weight and stiffness compensator is designed. This mechanism is intended to statically balance the elastic guide for its operational stroke, resulting in a mechanism with effectively zero stiffness. Statically balanced elastic mechanisms have been successfully used as low frequency vibration isolators for instruments [1], or to reduce the thermal input to an instrument [2].
The chosen mechanism is presented in the figure below. An optimization algorithm was set to obtain the optimal dimensioning of the compensation mechanism. The kinematics where computed so that the resulting stiffness was minimal.
A tension spring transmits through the triangular body (ABC) and the strut (BD) a force over the moving end of the parallelogram. As the moving end translates from its nominal position, the strut exerts a force with an increasing vertical component that compensates both for the weight of the moving parts and the residual stiffness of the parallelogram.
The initial off-set of the BD strut (ΔDz) allows to compensate for the weight of the payload.
The fixed point F is adjusted by means by two blocks with elastic hinges. This way manufacturing tolerances are compensated.
As a result of this study, a quasi-zero stiffness compensator was achieved. The thermal output by the actuators was reduced by two orders of magnitude in steady state, consequently proving that the novel configuration is indeed feasible.
Once the design is completed, a proper control strategy must be used to ensure an accurate positioning of the end-effector. The dynamics behavior of the system was modeled and evaluated. The lumped mass dynamic model was taken as shown in the following figure. A PID controller with a low pass filter was designed. The results are follows:
Additionally, considering that this system performs a task that is repeated over time, Iterative learning control (ILC) seems an appropriate control strategy. In ILC the controller is tuned through repetition. The algorithm was set so that once a working prototype is ready, this control strategy can be applied.
[1] D.K. Ferry, Negative-stiffness vibration isolation improves reliability of nano-instrumentation, Laser Focus World 43, no. 10.
[2] C.Werner, A 3D translation stage for metrological Atomic Force Microscope. PhD thesis. Technical University of Eindhoven, 2010.