Intelligent Feedrate Optimization using Uncertainty-aware Physics-based and Data-driven Servo Dynamic Models
Kim, Heejin
2023
Abstract
Intelligent manufacturing machines envisioned for the future must be able to autonomously select process parameters that maximize their speed (productivity) while adhering to quality specifications. One of the major sources of quality degradation in feed drives of the manufacturing machines is motion-induced servo error, which can be caused by several aspects such as limited bandwidth of feedback controllers, flexible structures, or nonlinear friction. Another source of servo error is force-induced servo error, which is caused by process force such as cutting force. Given that motion- and force-induced servo errors tend to increase with higher motion speeds, there is a keen interest in maximizing the speed of motion while respecting the tolerances on servo errors. To enable this, numerous works in the feedrate optimization aim to maximize machine speed subject to servo error constraints. However, the vast majority of the available methods only consider constraints on kinematics and/or static models of servo error without any incorporation of the dynamic servo models, which leads to sub-optimal feedrates. Moreover, they do not quantify the uncertainty of the servo error predictions, and hence may not effectively adhere to constraints in the presence of high uncertainty due to the model inaccuracy. To address these shortcomings, this dissertation proposes the framework and a set of methodologies for intelligent feedrate optimization approach enabled by uncertainty-aware physics-based and data-driven servo dynamic models. First, it proposes feedrate optimization with constraints on kinematics and servo error using physics-based servo dynamics. The optimization is formulated with respect to a time-based path parameter, which enables the linear dynamic model to be included in the optimization solved using linear programming. The integration of the servo dynamic model enables dynamic components of the servo error to be incorporated in the feedrate optimization, such as servo error pre-compensation, which allows for faster motions without violating tolerance constraints. Furthermore, the accuracy and computational efficiency of the feedrate optimization is improved using windowed sequential linear programming. Numerical feasibility is guaranteed by imposing smooth switching between the feedrate-optimal trajectory and a conservative backup trajectory. The performance of the feedrate optimization using physics-based servo model is validated using a desktop 3D printer and a precision motion stage to demonstrate reduction in cycle time while achieving similar quality to that of conservative approach used in the status quo. Second, the dissertation augments the physics-based servo models with a data-driven servo model to form an uncertainty-aware digital twin in the feedrate optimization to correct for inaccuracies introduced by dynamic uncertainties that are not modeled in the physics-based models. Known uncertainty is incorporated via a set of physics-based models, while unknown uncertainty is learned online by training the data-driven models via sensor measurements. The uncertainty-aware digital twin predicts the distribution of servo error, which is used in the feedrate optimization to constrain the servo error under desired tolerance and stringency. The proposed intelligent feedrate optimization is validated using a desktop 3D printer and a CNC machine tool prototype to demonstrate cycle time reduction while achieving similar tolerance stringency with conservative approach. The broader impact of this dissertation is to achieve desired quality and higher productivity with less trial-and-error. It is expected to be applicable to any manufacturing machines that use feed drives.Deep Blue DOI
Subjects
Feedrate optimization Dynamics Modeling and control Digital twin Computer numerical control (CNC) 3D printing
Types
Thesis
Metadata
Show full item recordCollections
Remediation of Harmful Language
The University of Michigan Library aims to describe library materials in a way that respects the people and communities who create, use, and are represented in our collections. Report harmful or offensive language in catalog records, finding aids, or elsewhere in our collections anonymously through our metadata feedback form. More information at Remediation of Harmful Language.
Accessibility
If you are unable to use this file in its current format, please select the Contact Us link and we can modify it to make it more accessible to you.