No abstract

In this work, a full and complete development of the tangent stiffness matrix is presented, suitable for the use in an absolute interface coordinates floating frame of reference formulation. For simulation of flexible multibody systems, the floating frame formulation is used for its advantage to describe local elastic deformation by means of a body's linear finite element model. Consequently, the powerful Craig-Bampton method can be applied for model order reduction. By establishing a coordinate transformation from the absolute floating frame coordinates and local interface coordinates corresponding to the Craig-Bampton modes to absolute interface coordinates, it is possible to satisfy kinematic constraints without the use of Lagrange multipliers. In this way, the floating frame does not need to be located at an interface point and can be positioned close to the body's center of mass, without requiring an interface point at the center of mass. This improves simulation accuracy. In this work, the expression for the new method's tangent stiffness matrix is obtained by taking the variation of the equation of equilibrium. The global tangent stiffness matrix is expressed as a local tangent stiffness matrix, consisting of both material stiffness and geometric stiffness terms, transformed to the global frame by the rotation matrix of the floating frame. Simulations of static and dynamic validation problems are performed. These simulations show the importance of including the tangent stiffness matrix for both convergence and simulation efficiency.

Design optimizations of flexure-based mechanisms take a lot of computation time, in particular when large deformations are involved. In an optimization procedure statically deformed configurations of many designs have to be obtained, while finding the statically deformed configuration itself requires tens to hundreds of load step iterations. The kinematically started deformation method (KSD-method) [1] computes deformed configurations fast by starting the computation from an approximation. This approximation is obtained by allowing the mechanism only to move in the motion-direction, based on kinematic equations, using data of the flexure joints in the mechanism. This is possible as flexure based mechanisms are typically designed to be kinematically determined in the motion directions. In this paper the KSD-method is extended such that it can also be applied without joint-data, such that it is not necessary to maintain a database with joint-data. This paper also shows that the method can be used for mechanisms containing joints that allow full spatial motion. Several variants of the KSD method are presented and evaluated for accuracy and required computation time. One variant, which uses joint-data, is 21 times faster and shows errors in stress and stiffness below 1%, compared to a conventional multibody analysis on the same model. Another variant, which does not use joint-data, reduces the computation time by a factor of 14, keeping errors below 1%. The KSD-method is shown to be helpful in design optimizations of complex flexure mechanisms for large range of motion.

Design and optimization of flexure mechanisms and real time high bandwidth control of flexure based mechanisms require efficient but accurate models. The flexures can be modeled using sophisticated beam elements that are implemented in the generalized strain formulation. However, complex shaped frame parts of the flexure mechanisms could not be modeled in this formulation. The generalized strain formulation for flexible multibody analysis defines the configuration of elements using a combination of absolute nodal coordinates and deformation modes.This paper defines a multinode superelement in this formulation, i.e., an element having its properties derived from a reduced linear finite element model. This is accomplished by defining a local element frame with the coordinates depending on the absolute nodal coordinates. The linear elastic deformation is defined with respect to this frame, where rotational displacements are defined using the off-diagonal terms of local rotation matrices. The element frame can be defined in multiple ways; the most accurate results are obtained if the resulting elastic rotations are as small as possible. The inertia is defined in two different ways: the so-called “full approach” gives more accurate results than the so-called “corotational approach” but requires a special term that is not available from standard finite element models. Simulations show that (flexure based) mechanisms can be modeled accurately using smart combinations of superelements and beam elements.

The use of beam elements with constant cross-section limits the design freedom. This thesis presents a formulation for beam elements with a varying cross-section, to allow for the analysis of non-prismatic flexures. Design optimizations of several flexure joints shows that this extra design freedom can result in significantly better flexure joints.Flexure joints contain parts which connect the flexures to each other. Although these socalled frame parts are intended to be very stiff, their compliance can significantly decrease the overall support stiffness of the flexure joints, reducing the performance. An accurate evaluation of the performance requires the stiffness of the frame parts to be modelled. However, frame parts typically have complex shapes, such that they can barely be modelled using beam elements. This thesis formulates a superelement by which complex shaped parts can be modelled efficiently. Examples show that flexure joints can be modelled efficiently and accurately by using the superelement to model frame parts and using beam elements to model the flexures.Very complex flexure-based mechanisms can be analysed more efficiently and more accurately by using the methods and elements that are introduced in this thesis. This may help the development of new flexure mechanisms, increasing the potential for using flexure mechanisms in practice.