Date of Award
Spring 2019
Document Type
Dissertation
Degree Name
Doctor of Philosophy (PhD)
Department
Mechanical Engineering
First Advisor
Nagurka, Mark L.
Second Advisor
Barth, Eric J.
Third Advisor
Voglewede, Philip A.
Abstract
A fluidic artificial muscle is a type of soft actuator. Soft actuators transmit power with elastic or hyper-elastic bladders that are deformed with a pressurized fluid. In a fluidic artificial muscle a rubber tube is encompassed by a helical fiber braid with caps on both ends. One of the end caps has an orifice, allowing the control of fluid flow in and out of the device. As the actuator is pressurized, the rubber tube expands radially and is constrained by the helical fiber braid. This constraint results in a contractile motion similar to that of biological muscles. Although artificial muscles have been extensively studied, physics-based models do not exist that predict theirmotion.This dissertation presents a new comprehensive lumped-parameter dynamic model for both pneumatic and hydraulic artificial muscles. It includes a tube stiffness model derived from the theory of large deformations, thin wall pressure vessel theory, and a classical artificial muscle force model. Furthermore, it incorporates models for the kinetic friction and braid deformation. The new comprehensive dynamic model is able to accurately predict the displacement of artificial muscles as a function of pressure. On average, the model can predict the quasi-static position of the artificial muscles within 5% error and the dynamic displacement within 10% error with respect to the maximum stroke. Results show the potential utility of the model in mechanical system design and control design. Applications include wearable robots, mobile robots, and systems requiring compact, powerful actuation.The new model was used to derive sliding mode position and impedance control laws. The accuracy of the controllers ranged from ± 6 µm to ± 50 µm, with respect to a 32 mm and 24 mm stroke artificial muscles, respectively. Tracking errors were reduced by 59% or more when using the high-fidelity model sliding mode controller compared to classical methods. The newmodel redefines the state-of-the-art in controller performance for fluidic artificial muscles.