1 Integration of Emerging Inspection Devices with Rapid Manufacturing Systems Anath Fischer CAD & Life Cycle Engineering Faculty of Mechanical Engineering Technion - Israel Institute of Technology, Haifa, Israel CIRP-2007-SPCIES-Production System Evolution Dresden, Germany, August, 2007 2 Industry Globalization Challenges Globalization has forced industries to operate in a highly competitive environment. The challenge is to increase the volume and variety of global transactions: products, services, knowledge. Technological developments are need to improve production systems. 1
3 Current Status of Digital Inspection At the end of the manufacturing process, a produced part should be verified to determine whether it fits the CAD model prototype under a given tolerances. The importance of automatic inspection for manufacturing is growing. Current production demands fully automatic, fast and accurate inspection. 4 Scan data problems Physical Object Scanning Scanned model: large scale, noisy, unorganized, missing information on properties of the sampled object 2
5 Traditional Inspection Methods vs. Emerging scanning technologies Existing inspection computational methods do not take diverse scan information into account. Therefore, traditional inspection methods are limited and do not satisfy industry demands. Emerging scanning technologies are capable of capturing diverse data regarding geometrical and physical properties of the sampled object, such as surface normals, color, material. 6 Inspection technologies and computational methods Inspection technologies and computational methods are embedded and strongly interconnected components of a single process. Utilizing the synergy between them should have a significant impact on global inspection by accelerating digital geometric processing and improving reconstructed object accuracy. 3
7 The Goal Developing a new approach for inherently handling diverse data and utilizing scanning technologies for inspection. 8 Utilizing the Diverse Data for Inspection: The Approach Scanning process Cloud of points Analysis of scanned points Calculation of normals Sharp feature detection Segment identification Diverse scan data Surface normals Sharp feature knowledge Segment information Color Material Data verification and HSDM reconstruction Scan data filtering Inspection 4
3D Non-Contact Measurement Technology Recent advances in computer vision and optics have led to development of emerging non-contact scanning devices. Non-contact scanners are capable of sampling a very dense cloud of points from a part s surface in seconds. The following emerging non-contact technologies are described: o Laser scanners o Lasers based on Conoscopic Holography technology o 3D cameras 9 10 3D Non-Contact Technology: Laser Scanners Laser scanner uses stereoscopic technique (Cyberware, Nextech). The distance of a sampled point is computed by means of a directional light source and a video camera. The CCD camera s 2D array captures the image of the surface profile along the laser sampling. The disadvantage is that the camera collects a small percentage of the reflected energy and therefore additional processing is needed. 5
11 NEXTEC- Digital CMM based on laser scanning technology Product Laser 3D scanning for quality control, inspection and reverse engineering applications. Specifications Automatic Alignment Time: 50 sec Total Machine Accuracy: 14 µm (2 sigma) Data capture rate: 50 points per sec Advantages Fully compatible with traditional CMM equipment Highly reliable measurements High flexibility 12 NEXTEC- Digital CMM adapts globalization strategy Nextec- Laser scanner Blades Technology - Pratt & Whitney and Rolls Royce CMMs- Coord3, Brown&Sharp, Mitutoyo 6
13 NEXTEC- Digital CMM Product Laser 3D scanning for quality control and reverse engineering applications. Typical Applications Injection molding Stamping Inspection & RE Dimensional verification 3D Non-Contact Technology: Conoscopic Holography Technology 14 Conoscopic holography is a polarized interference process based on crystal optics. The measurement process retrieves the distance of the light sampled point from a fixed reference plane. Advantages: Accuracy measurement is in microns. CH can handle large measurement angles. CH measures metals with high accuracy. Disadvantage: CH process is slow (minutes) with respect to 3D cameras. 7
15 OPTIMET- Laser Scanner Product High precision, 3-D measurement sensors, based on conoscopic holography technology Specifications Accuracy absolute : 2-35 µm Angle of reading up to 85 degrees Reading speed: 800 points per sec Advantages Very high accuracy Measurement from extremely acute angles Robust to variety of surface properties 16 OPTIMET- Laser Scanner Product High precision, 3-D measurement sensors, based on conoscopic holography technology Typical Applications Quality control Jewelry Laser drillings In-process inspection 8
3D Non-Contact Technology: 3D Cameras 17 3D photography is based on reconstructing 3D data from 2D images, taken from different points of view. Matching problem: for a given sampled point, its projected point should be identified in each image. To this end, a pattern is projected on the part where the same pattern points are identified in each image (Cognitens). The advantage is that a large number of measurements can be produced in one shot. 18 COGNITENS- 3D Camera Product 3D camera for fast measurement of manufactured parts in plant floor environment Specifications Image acquisition: <0.001 sec 3x1.3 mega pixel CCD cameras Overall system accuracy: 100µm (3 sigma) Typical Applications automotive industry aerospace industry Advantages Very high speed of measurement Wide area measurement at one shot 9
19 Distribution of 3D scanning Technologies- Universities & Research Institutes Labs [%] Data acquisition technologies 80.00% 70.00% 60.00% 72.73% 63.64% 50.00% 45.45% 40.00% 30.00% 27.27% 20.00% 18.18% 10.00% 9.09% 0.00% CMM Laser Scanners 2D Cameras 3D Cameras CT/MRI Scanners Other 20 Distribution of 3D Scanning Technologies - Hi-Tech Industry 10
21 Distribution of Diverse Data Utilization- Hi-Tech Industry 22 Traditional Shape Retrieval Methods Most commercial methods that deal with shape retrieval do not utilize emerging technologies. They use only the 3D coordinates of the sampled object, while ignoring other diverse data provided by non-contact technology. As a result the digital geometric process becomes very difficult because it must extract implicit or missing information from the 3D coordinates. Moreover, it retrieves the object shape without preserving sharp features and smoothness, or detecting a functional surface. 11
23 Shape Retrieval Problems Non-detection of sharp features Reconstructing inefficient mesh Ignoring mesh physical-based properties 24 Using Diverse Scan Data Inspection developers Coordinates of sampled points Surface normals associated with sampled points RT users Coordinates of sampled points Information on sharp features of the sampled objects Surface curvature associated with sampled points Potential of emerging scan technologies is not utilized. New methods needed to provide synergy between RT developers and users. 12
25 The Goal Developing new reconstruction methods that utilizes diverse scan data 26 Utilizing the Diverse Data for Inspection: The Approach Scanning process Cloud of points Analysis of scanned points Calculation of normals Sharp feature detection Segment identification Diverse scan data Surface normals Sharp feature knowledge Segment information Color Material Data verification and HSDM reconstruction Scan data filtering Inspection 13
27 Utilizing the Diverse Data for Inspection: The Grid-based Approach (2) Diverse scan data For each voxel set of points Classify scan points to voxels New reduced points Apply 3D GBF filter to calculate representative surface point Inspection process and shape retrieval 28 2D Bilateral Filter J 1 = [ (, )] [ (, )] c f d n c k g δ J J J n Ω n c n c c n 14
Stage 2: Applying 3D Geometric Bilateral Filter (3D GBF) for Data Fusion 29 3D GBF calculates Representative Surface Point (RSP), which replaces the voxel set of points. The proposed 3D GBF is based on 2D bilateral filter technique. 3D GBF belongs to edge-preserving filters. 3D GBF utilizes diverse scan data and surface normal information to prevent smoothing across edges. 3D Geometric Bilateral Filter (3D GBF) 30 1 sj = f d( pij, c ) j g δ( Lpij, Lsj ) pij kj = f d( pij, cj) g δ( Lpij, Lsj ) k j pi Vj pi Vj s j RSP point of voxel V j. f [ ] closeness weight function. g [ ] similarity weight function. d(p ij, c j ) similarity of positions (Euclidian distance) between point p ij and centroid c j. δ(lp ij,ls j ) similarity of property L between points p ij and s j. 15
31 3D Geometric Multilateral Filter (3D GMF) 3D GBF can be extended to use several types of diverse data. Several properties L 1, L 2,, L n can be involved in the filtering process 1 s = f d p c g Lp Ls g Lp Ls g Lp Ls p 1 1 2 2 n n j ( ij, j) ( ij, j) ( ij, j) ( ij, j) ij k δ δ δ j pi Vj k f d p c g Lp Ls g Lp Ls g Lp Ls 1 1 2 2 n n j = ( ij, j) δ( ij, j) δ( ij, j) δ( ij, j) pi Vj δ(l n p ij,l n s j ) similarity of property L n between points p ij and s j. 32 The proposed 3D GMF approach advantages Removes noise Preserving the underlying sampled surface including fine details and sharp features Reduces data, leading to further fast data processing or surface reconstruction Is robust, can handle complex topology Is fast Is simple to implement 16
Diverse data based method Examples 34 HSDM Reconstruction 17
Neural-Network Grid-based Re-meshing for micro-mri Technology 35 Rapid Prototyping models of bone micro-structure: (a) specimen from lumbar spine; (b) specimen from femoral bone Micro-structures re-meshing using Neural Network: (a) Original mesh with 47% failure -angles test; (b) Re-meshed microstructure results with only 22% failure in the same test 36 Complexity analysis Stage Time Space HSDM construction O(dn) O(dn) Scan data filtering O(n) O(n) Connectivity graph construction O(dn) O(n) Mesh generation O(n) O(n) 18
37 Summary & Conclusions The non-contact scanning technology can significantly improve automatic inspection in real time for rapid manufacturing (RM). Therefore, this technology enhances globalization. Hi-tech industries and RM use non-contact laser scanners, but do not utilize diverse data. A complete system that consists of non-contact technology and a robust surface reconstruction method is needed. In this talk we described diverse data based reconstruction methods that can be well integrated, in real time, with the noncontact technology. Thank you 19