Optimal Configuration Selection for Accuracy Enhancement of Programmable Machines
Funded by the NSF



In this project, we propose to simulate optimal configurations for two typical machines: Puma 560 robot and a Stewart-platform-type machine tool. The Puma 560 robot is an open-chain robot manipulator that has been extensively used in laboratories and industries. Determination of optimal configurations for this robot has not only theoretical but also practical significance. It will provide guidelines for practitioners to use optimal robot poses for its accuracy compensation.

The accuracy of programmable machines, including computer numerically controlled (CNC) machine tools and robot manipulators, is of paramount importance as it affects various aspects of product quality. The ever-increasing computational power of digital controllers possesses the potential of allowing the user to enhance the machine accuracy by periodic execution of special calibration software routines, using appropriate sensing devices. Enhancing machine accuracy through software compensation usually follows four steps:

1. modeling of the system structure,

2. measuring of position and orientation errors of the machine,

3. identification of parameter errors of the system model,

4. modification of control commends to improve machine accuracy.

In order to effectively determine and hence compensate machine error sources, it is crucial to perform off-line selection of measurement configurations. Optimal selection of a set of measurement configurations can greatly improve the efficiency of parameter identification, and hence, the accuracy of the machine. However, since the dimension of the parameter space is very large and the cost function is highly nonlinear, this selection process could be beyond the computation power of today’s PCs if a global optimal solution is sought by an exhaustive search. On the other hand, gradient-based algorithms are often trapped into a local minimum.

There exist a number of nonlinear search algorithms that can effectively escape local minima, such as simulated annealing and genetic algorithm. On the other hand, computational intensiveness is a common drawback of these algorithms. Fortunately, the optimal configuration selection problem can be partitioned in a number of ways that parallel computation can be used to speed up the numerical computation process.

The geometry of CNC machine tools is typically an open chain. A new type of machine tool, developed at Ingersoll Milling Machine, Gidding and Lewis, and Hexel, is based on parallel mechanisms analogous to an inverted Stewart platform. Unlike conventional machine tools, the Ingersoll machine, for instance, has a fixed table that carries work pieces and an octahedral hexapod carrying the spindle hung above the table. While the new machine tool is more rigid, its workspace is small and the structure of the hexapod is large. Therefore, it is more important to carefully plan its trajectory before a pose measurement task is conducted. In this research, we will also use the Beowulf cluster to determine the optimal configurations for this type of parallel CNC machine tool. After the optimal configurations of the machines are obtained, we will use these configurations to calibrate the machines in our laboratory. The result of the accuracy enhancement, including the pros and cons, will be reported through conference presentations and journal publications.