Calibration method and uncertainty propagation methodology for a portable parallel kinematic machine tool

The manufacture of all kind of goods relies mainly on industrial articulated robots and machine tools. Off-the-self robots have evolved from performing repetitive tasks to working in non-structured spaces and even near human-beings, by integrating powerful sensors and adopting advanced information processing algorithms. Machine tools, on the other hand, have evolved towards high accuracy, high throughput, and versatility. On the other hand, the in-situ maintenance and repair of goods that are in service, such as industrial installations or critical installations, rely on large sets of specialised tools that most of the times are of single use. The manufacture of features in large parts during setup also suffers the same lack of versatile tooling. The portable machine concept aims to fill this gap. However, the feasibility of performing in-situ machining-like operations poses several challenges such as accessing the place of intervention, fixating of the machine to the part, referencing the coordinate systems or the variety of operations that are required to perform. In this research, a self-portable parallel kinematic machine is proposed to perform in-situ machining interventions that can walk to the place of intervention. The parallel kinematics were extensively explored for machine tool applications between 1995 and 2005 but were almost abandoned as all-purpose machine topology due to difficulties related to design and limited workspace and accuracy. However, the high stiffness to mass ratio inherent to parallel kinematics machines is a crucial factor in building in-situ machining solutions on top of them. On the other hand, the relatively low workspace, 100x100x60 mm and rotations of up to 20 degrees, and limited accuracy, 0.1 mm, of the proposed self-portable machine tool still provide valuable support to many in-situ interventions. The thesis covers the modelling that supports the analysis of the workspace, the estimation of the actuator forces and velocities and the stiffness at the tool centre point. Moreover, key technologies and implementation details that enable the self-portable concept are described: (i) disengageable spherical joint actuator that allows switching from walking to machining configuration; (ii) communication and power cable propulsion system; (iii) active feet. The preliminary results of walking and machining trials that demonstrate the feasibility of the concept are presented. The main contribution of the thesis consists of (i) the calibration design methodology, supported by propagation of uncertainties in the design phase that enable to predict the positioning performance of the machine, and (ii) the solution prepared to calibrate the machine autonomously once it reaches the place of intervention. Such calibration is necessary because the machine kinematics depends on the relative positions where the feet are attached to the part. It relies on three onboard cameras and simple features of the feet of the machine. Without information on the position of the features on the feet, from the imaging of the features at many poses, it is possible to estimate the machine kinematics and the positions of the features simultaneously. Uncertainty propagation calculations extensively support the kinematic calibration solution design methodology. The methodology, adopted from the regression theory, has demonstrated to be a powerful design tool to predict the positioning performance of the machine. The uncertainty estimations have been validated experimentally by carrying out 36 calibrations in different configurations and analysing the variance of those parameters that should be constant. The agreement between uncertainty estimation and experimental variance have been good. Finally, the machine positioning performance has been assessed by studying the positioning error in 6 DOF with the aid of a Laser Tracker in four different feet configurations obtaining a general figure of 0.1 mm in translation and 0.02 degrees in orientation. These results agree with the uncertainty propagation estimations and are within the requirements of the machine. Therefore, it can be stated that the work done on this thesis represents an advance in the methodology for the design of calibration solutions, having been successfully applied to a novel concept of portable machine