Thesis Research

Digital Twin Model-Free Manufacturing in Voxel-based CAM with Applications to Machining Castings

Unique topics researched/implemented:

  • Quaternions / Quaternion Rotation 
  • Dual Quaternions / Dual Quaternion Transformations
  • Maximization of Mutual Information (Shannon Entropy)
  • Genetic Algorithm Optimization 
  • Particle Swarm Optimization 
  • Voxels
  • Parallel Computing (GPU)

Overview of application:

My Master's thesis research is focused on the automation and solution to a very particular problem: how does one process the gcode needed to machine a part when CAD models either do not exist or are not practical to construct?

This question is pertinent to areas of reverse engineering and remanufacturing. Despite the vast majority of industrial machining operations having all desired part dimensions readily available, engineers still require methods to rapidly manufacture components when such data is not readily available or is too costly to determine directly. 

This study is further expanded to consider machining from irregularly shaped starting stock geometries and investment castings. Again, most parts encountered in a modern machine shop originate from a relatively well defined stock material. However, applications where castings must be machined in post processing (e.g. turbine blades) require that the starting stock material be unique to a given part. These processes employ the use of datum points or fiducial references to orient the casting in a fixture to ensure proper positioning relative to the CNC coordinate space. Such castings, however, are not dimensionally precise. Comparing the part model with a digital scan of the casting would ensure that the part be maximally centered inside the casting before the toolpath generation begins. This also ensures that the casting is within tolerance, thus introducing an element of quality control that saves time.  

Finally, the need to accurately gather CAD/CAM data in real time during the manufacturing process will likely be necessary with advances in additive-subtractive machining. Intermediate point cloud scans combined with dynamically generated gcode can define an important type of feedback required in precision hybrid manufacturing. 

 
CollinsResearchObjectiveSlide.gif
 
PointCloudScanning.jpg

Overview of Process: 

This research topic is a new area for Georgia Tech's Precision Machining Research Consortium so will begin as an exploratory measure to introduce and demonstrate a procedure whereby parts can be scanned, registered and machined from a voxel structure. To do this, two types of machining will be done. First, a part will be scanned and processed in a voxel-based CAM to be machined. Second, two scans of a final geometry and a simulated casting volume will be properly aligned in an image registration program. Based on this relative positioning, the toolpaths and gcode will be generated to machine away excess material from the simulated casting to create the desired part.