Slide 1

Automatic Generation of 3D Machining Surfaces With Tool Compensation From Graylevel Image Models Theodor BORANGIU, borangiu@cimr.pub.ro Anamaria DOGAR, dogar@cimr.pub.ro Alexandru DUMITRACHE, alexdu@easynet.ro Slide 2 Summary Height Map images Modelling the machining surface and tool shape Performing tool compensation Generating roughing and finishing toolpaths Error analysis Future plans Slide 3 Height Map Images Slide 4 Obtaining Height Map Images 3D Model in POV-Ray Height Map Model Slide 5 Obtaining Height Map Images 1)Remove light sources 2)Remove material textures 3)Use an ortographic camera 4)Apply a pigment with brightness proportional to the distance from the camera plane: farthest point:pure black (brightness 0) closest point: pure white (brightness 1) Slide 6 2.5D Surface Modelling Pixel graylevel at (i,j) encodes surface height at (x,y) Pixel-to-millimeter ratio: x = R i y = R j Minimum Z of the surface: black pixel Maximum Z of the surface: white pixel Graylevel value: 8-bit integer: low precision, low storage space 16-bit integer: good precision Floating point: best precision, high storage space Slide 7 Simplest Case: 2 Dimensions Part model: Tool model: Tool compensation: Slide 8 Simplest Case: 2 Dimensions Offsetting is done by image dilation For efficiency, only contour pixels need to be processed Tool path is generated by extracting the contour By image erosion we obtain the machined shape Slide 9 Tool Shape Modelling Conic Mill Round End Mill Bull End Mill Slide 10 Tool Compensation Objective: Generating gouge-free tool paths Idea: For each (x,y) position, find the maximum depth at which the end mill can go down without cutting extra material Algorithm: Graysale image dilation Image: Surface model Structural element: Tool model Slide 11 Tool Compensation Result A surface on which tool's end point can move safely Slide 12 Advantages Gouge-free tool paths for many tool shapes Immediate generation of basic roughing and finishing tool paths Simple implementation, no need for complex 3D geometry computations Slide 13 Disadvantages High computation time (Example: image 1000x1000 pixels, tool 50x50 pixels, running time: 11 seconds on a Pentium 4-M 2.00 GHz) High storage space for the image model Compromise between precision and speed! Slide 14 Increasing Speed It is not always necessary to compute the entire surface The algorithm can be parallelized Slide 15 The Software Slide 16 Software Features Grayscale model support Tool shape editor Roughing toolpath generation Finishing toolpath generation ISO CNC (G-Code) output Slide 17 Tool Shape Editor Slide 18 Predefined tool shapes: spherical end mill conical end mill flat end mill bull end mill User defined tool shapes Slide 19 Roughing Each roughing stage is performed at constant Z level At a given Z level, selecting the region where the cutter should clean up is an image thresholding operation For flat endmill cutters we use 2D offset compensation Slide 20 Roughing Slide 21 Finishing First Method In XY plane, the tool moves parallel with either one axis or an arbitrary direction The tool moves on the safe surface There is no need to compute the whole safe surface Slide 22 Finishing Second Method Tool paths are at constant Z levels Because of the tool shape, we cannot use 2D compensation any more The whole surface needs to be computed! Slide 23 Finishing - Combined Slide 24 Semi-Finishing Slide 25 Finishing Slide 26 Error Analysis A tool can be too big to machine the fine details At first, we can use a bigger tool to machine surfaces without fine details, and then a smaller tool to machine only the small details We can simulate one cutting operation and see what could not be machined Slide 27 Final Touch Slide 28 ISO CNC Output Toolpaths are made of linear segments and circular arcs Succesive segments may be approximated with circular arcs Toolpath optimization: reduce the time for moving the head without cutting M03 G0 X80 Y7.75 G1 Z-40 F100 G1 X65.5 G0 Z0 G0 X71.75 Y10.75 G1 Z-40 F100 G1 X80 G1 Y13.75 G1 X73 G1 Y16.75 G1 X80 G1 Y19.75 G1 X73.25 G0 Z0 M05 Slide 29 Sample Workpiece Slide 30 Technical Data Dimensions: Material: Cutters used: Image size: Roughing: Semifinishing: Finishing: Final Touch: 100 x 50 x 20 mm Wood: Beech (Fagus) Flat 5mm, Round 6mm and 3mm 1000 x 500 379 instructions, 22 minutes, 200 mm/min 2034 instructions, 36 minutes, 200 mm/min 7131 instructions, 53 minutes, 400 mm/min 88 instructions, 1 minute, 300 mm/min Slide 31 Other Example: Semifinished Workpiece Slide 32 Other Algorithms Used Contour Detection: Moore Neighbourhood Search Simplifying Polylines: Douglas Peucker Generating Discrete Line Segments: Bresenham Slide 33 Future Plans: Collision Detection The whole tool shape can be modelled, including the tool holder, to check if a tool path will cause a collision with the workpiece At every moment the amount of material can be computed; this is useful to check if the maximum allowed cutting depth of a tool is not exceeded. Slide 34 Future Plans: Better Milling Strategies Roughing example Slide 35 Future Plans: CNC Simulator Input: a file with RS274 G-Code Realtime 3D Simulation for CNC Mill Engine based on height map images Collision detection Possibility of exporting animations Slide 36 Thank You!