Research Project enlight
Interactive visualization of machining processes by complex geometry models using highly parallelized algorithms.
Visualizations in automation and mechanical engineering for simulation and modeling are very complex. Geometric models with significant data size require powerful methods for calculation and visualization. The aim of the „Enlight“ project is to develop and implement a new process that modifies and visualizes the complex geometry data in real-time, utilizing a heterogeneous framework to support modern hardware architectures such as Multi-/Many-Core Central Processing Units (CPUs) and General Purpose Graphic Processing Units (GPGPUs).
The areas of application are design (CAD) and manufacturing simulation (CAM), as well as in medicine. Today, CAM models of very detailed work pieces consist of hundreds of millions of surface elements. Crucial for technical and scientific applications is an interactive and accurate representation of the geometry.
A prerequisite for the reliable simulation and verification of the entire manufacturing process is a complete and accurate modeling of the geometry of the work piece, which changes in thousands of processing steps.
The method developed for interactive visualization of machining processes basically consists of the following two core components:
Modeling: geometry models are managed according to a space partitioning scheme. When applying subtraction volumes (resulting from the tool paths of processing steps), a geometric Boolean subtraction is performed. An intelligent clearing strategy ensures that non-surface geometry parts are removed.
Visualization: Using a ray casting method that is adapted in terms of geometric Boolean subtraction.
A prototype implementation uses the parallelization of modern hardware architectures and was able – on the basis of a manufacturing simulation of an impeller – to demonstrate the capability of the method. The initial geometry and the 2493 subtraction volumes consist of approximately 21million triangles (total triangles). The visualization was carried out on an Intel Core i7 quad core CPU with a screen resolution of 800 by 600 pixels. The figure below shows the good scalability in terms of runtime (ray casting time) and memory efficiency (total triangles reduced to model triangles by clearing strategy) with respect to the number of subtraction volumes.
Special attention is given to the accuracy of the visualization: Other methods often use approximations, which can lead to problems regarding their visualization accuracy. The developed method is not subject to this restriction (see figure).
This project was funded by the European Regional Development Fund and by the Upper Austria State Government under the Regional Competitiveness Program 2007-2013 Upper Austria.