Título: The robust design of robotic mechanical systems /
Autores: Al-Widyan, Khalid
Fecha: 2004
Publicador: McGill University - MCGILL
Fuente:
Tipo: Electronic Thesis or Dissertation
Tema: Engineering, Mechanical.
Descripción: Task uncertainty and noise---generically referred to as uncertainties in the thesis---in the design and operation of robotic mechanical systems pose a major challenge to the designer. The aim of this thesis is to contribute to the laying down of the foundations on which the design of robotic mechanical systems, in the presence of uncertainties, is to be based. Uncertainties are accounted for by means of robustness. To this end, a theoretical framework as well as a general methodology for model-based robust design are proposed. Within this framework, all quantities involved in a design task are classified into three sets: the design variables (DV), which are to be assigned values as an outcome of the design task; the design-environment parameters (DEP), over which the designer has no control; and the performance functions (PF), representing the functional relations among performance, DV and DEP. A distinction is made between globally robust design and locally robust design, this thesis focusing on the latter. Locally robust design is amenable to mathematical modelling of the performance by means of smooth functions of the DV and DEP. Resorting to the mathematical model available for the object under design, a design performance matrix, mapping the space of relative variations in DEP---referred to as noise in the literature on robust design---into that of relative variations in PF, is derived. Then, the locally robust design problem is formulated as the minimization of a norm of the covariance matrix of the variations in the PF upon variations in the DEP. Moreover, one pertinent concept are introduced: design isotropy. Next, it is shown that an optimally robust design can be secured by means of isotropy, whenever this is achievable under the constraints of the problem at hand. As proven in the thesis, designs obtained via isotropy lead to robustness even in the absence of a priori knowledge of the statistical properties of the variations of
In connection with robotic mechanical systems, the design task is decomposed into three subtasks, namely, kinetostatic, elastostatic, and elastodynamic, in this order. We show that parallel manipulators allow for isotropic designs, but their serial counterparts do not.
The motivation behind the research work reported here being the design of robotic mechanical systems, the simulation of their dynamic response becomes an essential component in their design. In order to validate the proposed design, a robust algorithm for the simulation of conservative linear systems, which model accurately the systems of interest in the presence of "small" environment perturbations, is introduced. The robustness of the algorithm lies in its immunity to roundoff and truncation errors, which could lead to either instability or a dissipative response in the simulation results otherwise.
Idioma: en