This dissertation describes the kineto-elastic analysis and component structural dynamic model updating of serial modular reconfigurable robots (MRRs). In general, kineto-elastic analysis is concerned with the structural vibrations, elastic deflections, and torque transmissions of robots which undergo motion from one pose (position and orientation) to another. This work focuses on the kineto-elastic analysis of MRRs undergoing low-speed quasi-static motion. When determining an MRR's payload capacity, or designing MRR modules, the main difficulty is the large number of module configurations and the infinite number of poses within each configuration. Also, the kineto-elastic models of MRRs can become quite large with an increasing number of modules, thereby increasing the numerical complexity. Furthermore, the analytical models of individual MRR components may contain uncertainties, such as unknown stiffness and material parameters, which may lead to large errors for assembled MRR models.
To alleviate these issues, a new framework was developed for the kineto-elastic analysis of MRR modules with an emphasis on assessing the worst-case poses. First, a combinatory search method was presented to reduce the computational burdens associated with determining the maximum payload capacity, and performing the module stiffness designs. This involved identifying the worst-case configuration and pose amongst a large number of configurations and infinite number of poses. Afterwards, it was demonstrated that the determination of an MRR's payload capacity, as well as the module stiffness designs, can be performed at the worst-case pose and configuration to satisfy a global set of kineto-elastic performance requirements for all remaining configurations. Next, a new component mode synthesis (CMS) model with fixed-free component boundaries was developed to reduce the sizes of kineto-elastic models, mimic natural link-joint connectivity, and allow experimental tests of joint modules in multiple poses to enable test-analysis model correlation. Finally, a novel method was created to update the uncertain model parameters of joint and link modules using frequency response data from component vibration tests in multiple poses (including the worst cases), with boundary conditions matching those from the CMS models. This procedure can completely avoid testing an entire assembly to perform model updating, and can provide accurate updated model results in any assembly pose.
Robust model updating methodology for estimating worst-case load capacity of existing bridges
