Converting Multi-Axis Machine Tools into Subtractive3D Printers by using Intelligent Discrete Geometry Data Structures
Abstract:
Cyber Physical Systems have revolutionized manufacturing to the point that next generation component manufacturing is now referred to as “digital manufacturing.” Parts are made based on digital models that are represented by discrete points or meshes in standard file formats such as STL. The difficulty with these representations is that they can become extremely large when high resolution is desired. As the next step towards complete 3D digital manufacturing, this research targets the formulation of a data model, and trajectory planning platform and methodology to execute a fully digital 3D 5-axis machining capability. This will be formulated to utilize specially designed and readily available, parallel digital processing hardware to transform a CAD model into a hybridized voxel model enabling rapid digital analysis. Initial work in this area has generated promising results where a 3D model is digitized, a parallelized 5-axis tool path planning and machining simulation are run on a multiple Graphical Processor Units (GPUs) and the subsequent machine tool trajectory generated during path planning process in the simulation is downloaded to a machine tool to generate the physical part. Such capabilities will bring classical subtractive manufacturing back into the arsenal of rapid prototyping providing users of typical CNC machine tools with the ability to rapidly determine if a part is viable on their machine, and it is viable, the ability to rapidly machine the part. The intellectual merit of the proposed work has two connected components. The first one is the formulation of a generalized methodology that is capable of employing the native support of modern parallel processing platforms via iterative processing of a fundamentally new geometric representation. From a mathematical perspective the proposed methodology can be described as a generalized numerical approach for geometric problems. The resulting platform provides a hybrid analytical/digital format that is capable of storing high resolution, three dimensional digital information, enabling digital analysis tools to be used in a parallel fashion. It provides a means by which current analytic models can be efficiently and effectively digitized, in real-time, to arbitrary accuracy and resolution levels for rapid and parallel digital processing. The broader impact of the proposed research is the creation of a revolutionary, real-time variable resolution geometric platform for CAD and CAM systems that is formulated to utilize the strengths of high performance computing / parallel architectures such as the GPU. The proposed approach can be easily scaled to processors with hundreds of cores as well as systems possessing multiple processors. The proposed approach fundamentally enables a possibility of splitting an entire problem into a set of completely independent rudimentary problems permitting a complex problem to be transparently solved in parallel by a set of independent computer systems. Thus, any system employing the proposed platform will be capable of natively performing calculations on any number of available computers with virtually none of the limitations due to the lack of complex specially developed algorithms. The proposed platform eliminates performance limitations by providing tens to hundreds times of performance improvement which is available with a GPU. Furthermore, the process improvements are extended several more orders of magnitude (tens or hundreds of thousands times) with the currently available processing clouds and GPU system clusters. As result it enables new options in algorithm development that are currently inconceivable due to performance limitations. From a commercial perspective, the proposed platform provides opportunities for new business models and environmental possibilities. There are some known projects such as Folding@home or Rosetta@home which employ the resources of home computers for the calculation of protein folding problems. Development of CAD/CAM software which can natively use the performance of distributed clusters may create a new market where users will be able to sell computational time of their business or personal computers to companies that are interested in computational performance. Finally, the proposed platform and methodology can be used to better optimize machine trajectories based on a variety of enhanced requirements such as waste generation, reduction of coolant used and minimal energy consumption. Such options are not functionally viable using current techniques as simulation times for iterative / optimal solutions are not realistic in the typical manufacturing arena. This hybrid representation will enable high performance computing to be readily and effectively used in manufacturing. Thus, this work will provide the foundation for a portal significantly enhancing manufacturing capability not only for major manufacturers, but for SME’s that already have access to high performance computing capabilities such as GPU’s.
Submitted by Thomas Kurfess
on
Abstract:
Cyber Physical Systems have revolutionized manufacturing to the point that next generation component manufacturing is now referred to as “digital manufacturing.” Parts are made based on digital models that are represented by discrete points or meshes in standard file formats such as STL. The difficulty with these representations is that they can become extremely large when high resolution is desired. As the next step towards complete 3D digital manufacturing, this research targets the formulation of a data model, and trajectory planning platform and methodology to execute a fully digital 3D 5-axis machining capability. This will be formulated to utilize specially designed and readily available, parallel digital processing hardware to transform a CAD model into a hybridized voxel model enabling rapid digital analysis. Initial work in this area has generated promising results where a 3D model is digitized, a parallelized 5-axis tool path planning and machining simulation are run on a multiple Graphical Processor Units (GPUs) and the subsequent machine tool trajectory generated during path planning process in the simulation is downloaded to a machine tool to generate the physical part. Such capabilities will bring classical subtractive manufacturing back into the arsenal of rapid prototyping providing users of typical CNC machine tools with the ability to rapidly determine if a part is viable on their machine, and it is viable, the ability to rapidly machine the part. The intellectual merit of the proposed work has two connected components. The first one is the formulation of a generalized methodology that is capable of employing the native support of modern parallel processing platforms via iterative processing of a fundamentally new geometric representation. From a mathematical perspective the proposed methodology can be described as a generalized numerical approach for geometric problems. The resulting platform provides a hybrid analytical/digital format that is capable of storing high resolution, three dimensional digital information, enabling digital analysis tools to be used in a parallel fashion. It provides a means by which current analytic models can be efficiently and effectively digitized, in real-time, to arbitrary accuracy and resolution levels for rapid and parallel digital processing. The broader impact of the proposed research is the creation of a revolutionary, real-time variable resolution geometric platform for CAD and CAM systems that is formulated to utilize the strengths of high performance computing / parallel architectures such as the GPU. The proposed approach can be easily scaled to processors with hundreds of cores as well as systems possessing multiple processors. The proposed approach fundamentally enables a possibility of splitting an entire problem into a set of completely independent rudimentary problems permitting a complex problem to be transparently solved in parallel by a set of independent computer systems. Thus, any system employing the proposed platform will be capable of natively performing calculations on any number of available computers with virtually none of the limitations due to the lack of complex specially developed algorithms. The proposed platform eliminates performance limitations by providing tens to hundreds times of performance improvement which is available with a GPU. Furthermore, the process improvements are extended several more orders of magnitude (tens or hundreds of thousands times) with the currently available processing clouds and GPU system clusters. As result it enables new options in algorithm development that are currently inconceivable due to performance limitations. From a commercial perspective, the proposed platform provides opportunities for new business models and environmental possibilities. There are some known projects such as Folding@home or Rosetta@home which employ the resources of home computers for the calculation of protein folding problems. Development of CAD/CAM software which can natively use the performance of distributed clusters may create a new market where users will be able to sell computational time of their business or personal computers to companies that are interested in computational performance. Finally, the proposed platform and methodology can be used to better optimize machine trajectories based on a variety of enhanced requirements such as waste generation, reduction of coolant used and minimal energy consumption. Such options are not functionally viable using current techniques as simulation times for iterative / optimal solutions are not realistic in the typical manufacturing arena. This hybrid representation will enable high performance computing to be readily and effectively used in manufacturing. Thus, this work will provide the foundation for a portal significantly enhancing manufacturing capability not only for major manufacturers, but for SME’s that already have access to high performance computing capabilities such as GPU’s.