A CAD MODELLING SYSTEM AUTOMATION FOR REVERSE ENGINEERING APPLICATIONS

A CAD MODELLING SYSTEM AUTOMATION FOR REVERSE ENGINEERING APPLICATIONS Jafar Jamshidi, Antony R. Mileham and Geraint W. Owen The University of Bath, UK J.Jamshidi@bath.ac.uk, A.R.Mileham@bath.ac.uk, G.W.Owen@bath.ac.uk In the Reverse Engineering process, scanned data digitised by various scanning systems from the component geometry is used to regenerate the original design information. CAD model creation from such data requires numerous stages of modelling processes. Several algorithms for the modelling processes were developed and verified in the earlier stages of this research. In this paper the automation of these algorithms into a unique CAD modelling system is introduced. The system user can create a complete CAD model by integration and minimum manipulation of the high resolution scanned data, obtained using a high accuracy Coordinate Measurement Machine with touch probe, and low resolution scanned data obtained using a high speed Laser Scanner, rapidly and accurately. 1. INTRODUCTION Reverse Engineering (RE) is the process of extracting design information from an existing part, for which such information is unavailable of mislaid. In order to digitise part geometry, various scanning systems are available in industry including the Coordinate Measurement Machine (CMM) with touch probe and the 3D Laser Scanner (LS). CMMs can create high accuracy data-points from the part surface, but when the number of the required data-points is large they are considered to be inefficient. On the other hand LS can rapidly create thousands of points from the part surface but in lower resolution compared to CMMs. In a typical component both high and low accuracy features can be seen. Therefore to have a fast and accurate scanning system, CMMs manufacturers have now equipped them with LS probe heads. However, the data integration of the two scanning systems is still in its early stages of development. For instance, several modelling processes are required for CAD model creation and this can be tedious, time consuming, and error prone. In this paper the automation of the modelling algorithms, which were developed and manually verified in the earlier stages of this research, is introduced. Although the system is not fully automated, however, the user interaction with the system is minimised to a great extent. The system performs modelling processes based on the user s selection within the available options in each stage. Alternative options are

2 Digital Enterprise Technology created for features to enhance the flexibility where possible, i.e. the user can decide to perform some of the modelling stages with or without interaction with the system. 2. RELATED WORK CMMs integrated with LS probe are now available in industry (Rooks, 2004). However, they are mostly used for individual probe usage or for basic applications such as part position finding (Chan et al, 2001) or for CMM probe guidance (Li and Liu 2003). In a different topology the LS is used as a separate device and the scanned data are used for CMM automated probe path planning (Shen et al, 2000). It is proposed to integrate the scanned data between the two systems of LS and CMM with touch-probe. Bradley and Chan (2001) used the CMM touch probe to scan the feature boundary of complex surfaces for surface extraction from data cloud scanned by LS. Their method reduces the surface extraction time in the modelling process; however it is unable to create a complete geometry of inward sloping faces and also the precise touch probe data-points are treated in a similar way to the points in data cloud, not as high accuracy data-points. If multiple probes or multiple sources of digitised data are used, the calibration of all the probing systems into a unique coordinate is necessary. Current practice is that different probes are calibrated prior to scanning to the same datum (Bradley and Chan, 2001). This method is rather inflexible due to the datum selection that is set to the centre of a single and fixed calibration ball attached to the CMM table. In another method, Shen and Menq (2001) have used the tip of the probe for 3D LS camera calibration with touch probe. Although this scheme results in accuracy of greater than 0.1mm, it is inflexible if the features on the part are far away from the initial calibration position. Other research has attempted to create an automated measurement planning and digitisation system. Spyridi and Requicha (1994) have automated the CMM programming using the CAD model to generate the CMM setup, probe selection, path and direction. Despite the dependency on CAD model, which makes it unsuitable for RE applications, their method can be used for automation of post RE measurements on similar components. In an ideal RE system, the complex geometric surfaces can be scanned with a non-contact LS in a very short time and leaving the accurate features for a CMM with touch probes. The scanning results then are required to be integrated together to create the CAD model. This requires faster and more accurate CAD modelling methods, as the existing modelling and optimisation processes are time consuming and/or not accurate enough for RE of components with precision features. 3. AUTOMATED CAD MODELLING SYSTEM A general overview of the system is given in Figure 1. In this method (Jamshidi et al [7], 2006) the sensitive features of the desired component are scanned using the conventional CMM with high precision (at 35m) touch probe in order to create CMM-model from the part geometry.

A CAD Modelling System Automation for Reverse Engineering Applications 3 Sensitive features are the features that require more accurate dimensions and have tighter tolerances. However, for the RE problem, where the tolerance information is unavailable, these features can be identified visually through their surface texture and machining used (Jamshidi et al [5], 2006). If a feature on the component has for instance grinding surface finish, one can assume that such feature has precision dimensions and tolerance. Therefore such feature can be accounted for as sensitive. Input Component Modelling Process Output CMM-model IGES User entries and data manipulation Automated CAD Modelling System CAD Result STL LS-model Return to digitisation in case of excessive missing data Figure 1 Overview of the CAD modelling process in the system Any other features of the component are considered as non-sensitive, thus can be digitised rapidly by LS. If complete coverage scanning is required the part should be digitised from different scanning view angles. Individual scanned sections of the component from different view angles should be registered to a unique coordinate to create the LS-model. Such a model is then registered to the same coordinates system of the CMM-model for further modelling stages. Here the registration of the scanned models is an automated operation, which utilises 3-Coordinate Balls (3CB) (Jamshidi et al, 2005) and can be followed to the fine registration process (Jamshidi et al [6], 2006) if necessary. Therefore the inputs of the automated system are the CMM-model, LS-model and the user inputs that are entered through selection options and user interaction (Figure 1) during the CAD modelling processes. The level of user inputs and interaction varies depending on the geometric complexity of the component. This can be a minimum or no user manipulation, for components with simple and standard geometry.

4 Digital Enterprise Technology 4. DATA INPUT TO THE SYSTEM The complete modelling processes take place in three software environments. For the work presented in this paper the Minolta Polygon Editing Tool, as the LS software, creates the triangulated mesh from the scanned data-cloud and exports the scanned section models into Stereolithography (STL) file format after the initial view registration and data-cloud reduction. The CMM-model is created using PC- DMIS which is standard CMM software. It is possible to create volumetric features using the measured data-points within PC-DMIS if necessary. Nevertheless, it is preferable to perform model manipulation in the CAD software environment as it has greater flexibility. The scanned CMM-model is then exported into IGES (Initial Graphics Exchange Standard) file format for importing to the CAD software. The LS and the CMM stage models, as the main inputs of the automated CAD modelling system are imported into the desired layers of the CAD software. For this a consistent naming convention, as shown in Table 1, is used to automate the process. For instance the 3CB data-cloud in the LS-model of the first scanned view can be saved under LS_3CB_V1.stl name. This approach can assure that such data-cloud will be loaded to its assigned layer of the CAD environment, in the file import process. Table 1 Target layer assignment for imported and newly created entities Source Data-cloud / Feature / Function CAD Object/Feature View # Import File Name New Entity Target Layer Part - CMM-model.igs - 20. CMM-model LS-model_V1.stl to 30. LS-model to Part V1 to V20 - LS-model_V20.stl 49. LS-model LS_3CB_V1.stl to 50. 3CB to 3CB V1 to V20 - LS_3CB_V20.stl 69. 3CB CMM-model - - Large Scale 70. Selection LS-model V1 to V20 - Selected Points 80. Redundant The highlighted selected points in Table 1 are in fact sections of the existing data-cloud that are automatically selected as redundant points, therefore eliminated from the data-cloud. These points include the unreal data-points, which are the result of noise and reflection in the scanned LS-model (Jamshidi et al [4], 2006) and also the points on and around the sensitive features. The resulting empty space of missing information in the data-cloud will be filled by accurate features generated from the features in the CMM-model. The remaining data input in to the system are the user entries that can include surface finish and tolerance information, and any other user selections that are available via options in user menus. 5. INFORMATION FLOW AND OUTPUT The information flow in the automated CAD modelling system is shown in Figure 2. After the data input and registration, in Stage-3, the volumetric features are formed by using the high accuracy measured data gained by the CMM touch probe. In this stage the user selects the required measured features from the CMM-model in order

A CAD Modelling System Automation for Reverse Engineering Applications 5 to create the real feature. For example two circles and two planes are required for a straight cylindrical feature, or a set of three points can be used to create a particular plane. Import models to desired layers 1. Register LS stage models 2. Register LS-model to CMM-model 3. Create volumetric feature 4. Create closure envelope 5. Select redundant points 6. Replace points by volumetric features 7. Delete redundant points 8. Stitch boundaries 9. Fill gaps 10. Further beatification (Reduce and smooth data-points) 11. Unification of CAD model elements 12. Create dimensions 13. Add tolerance to dimensions CAD model ready Figure 2 13 stages of automated modelling process in CAD environment The closure envelopes in stage-4 are the larger scale of the features from the previous stage. These closure envelopes are required for data-cloud selection by restraining the points in the neighbourhood of the sensitive features (Jamshidi et al [7], 2006). The selected points are considered as redundant and can be replaced by the high accuracy volumetric features, in stage-6. Mesh boundary Virtual mesh Feature edge Real mesh Figure 3 Filling gaps between digitised models by virtual mesh creation In stage-8, the boundaries of the data-cloud are stitched, in a zigzag approach, to the boundaries of the appropriate features by a virtual triangulated mesh (Figure 3). This process requires the user to rotate the model in the graphic view window so that the desired feature can be visible in a normal angle to the screen, and then the appropriate edge of the feature is selected by the user. Then the feature edge is divided by new points on the proposed edges in an automated mode. The stitching process starts by creating a line between the closest pair of points from the triangulated mesh boundary in LS-model and the feature edge. The next line begins from the end of the previous line to the closest point on the opposite model, i.e. if the line starts from LS-model the end of such line has to be located on the feature

6 Digital Enterprise Technology created or existed in the CMM-model. A virtual mesh surface is also added to each one of the newly created triangle sets. This can be a triangle with three straight lines or two straight lines and one curved line. Stage-9 is an essential step for complete surface model creation, which is necessary for solid CAD model construction in stage-11. When the CAD model is completed the desired drawings can be created in which the dimensions on sensitive features are obtained from the existing measured data in the CMM-model. Initially the measured data are considered as nominal. However, this may not be optimal for components produced near to the acceptable boundary of their dimensions, i.e. close to upper or lower limit tolerance. Nevertheless, for the purpose of this paper the measured data are assumed as nominal. The final stage for CAD modelling includes adding tolerance to the dimensional features. This can be done automatically by the embedded tolerance approximation method (Jamshidi et al [5], 2006), or can be accurately determined, by tolerance accommodation analysis, and be added to the desired dimensions by the user. 6. EXPERIMENTAL RESULTS For the validation of the automated modelling system, several components were reverse engineered, using the hardware and software described in Table 2. For the problem of model registration between the two scanned model types, an auxiliary 3CB, from which the coordinates information is extracted, is used. The 3CB was fixed, by conventional clamping, in the scanning view of the LS in order to digitise its visible sections throughout the scanning process. Table 2 Systems used for the method experiments Accuracy (mm) Hardware X Y Z Software Minolta 900 Laser Scanner with Minolta tele-lenz Brown & Sharpe Global CMM with TP20 Renishaw Probe ±0.22 ±0.16 ±0.19 ±0.035 ±0.035 ±0.036 Minolta editing tool Version 1.1 PC-DMIS Version 3.5 MR1 Beta Auxiliary 3CB set Ball Diameters = 28.63 Unigraphics NX3 Figure 4 illustrates a simple cubic component, which is fully digitised using the CAD modelling system. In this example it is assumed that only the diagonal holes are sensitive to the functionality of the part. The measured data in the CMM-model are used to create the hole features, which have the accuracy of micron level. Other sections of the part are left as triangulated mesh, as scanned by LS within an accuracy of ±0.22mm (the accuracy level of LS). The registration process between the scanned models is done automatically. Each hole is created as an internal cylinder from one circle and two planes representing the two ends of the cylinder. The closure envelopes, are created by increasing the hole diameters by 1.5mm and adding 0.75mm on either side of the cylinder. The orientation of the closure envelopes are intact with the same spatial location of the corresponding true hole features in the CAD model. The stitching

Y Z X A CAD Modelling System Automation for Reverse Engineering Applications 7 process is then taken place after the deletion of the selected redundant points by closure envelopes. Finally the drawings are created including the dimension values. Component Setup & CMM scanning Laser Scanning CAD result (section view) Sensitive Features 3CB set Figure 4 RE of simple cubic part using automated CAD modelling system The second experiment is the RE of a sample component with more features, details of which are shown in Figure 5. The overall geometry of the component is digitised by the LS from different view angles and registered to a unique coordinate system (Figure 5, A). The complex geometry features of the component are exported to CAD software for data integration with model scanned by CMM (Figure 5, B). All the other features are assumed as sensitive and/or standard, therefore scanned by CMM. Volumetric features are created using the high accuracy features in CMMmodel (Figure 5, C). The final result with the CAD drawing is given in Figure 5, D, which is a solid model combined of triangulated mesh and standard geometry feature. This model can be improved by beautification of the triangulated mesh then recreation of the solid model, and adding tolerance to the dimensional features. A View Registration B Scanned views Complete LS-model Complex geometry, but non-sensitive features C D CAD drawing with dimensional information Figure 5 RE of sample component using automated CAD modelling system 7. CONCLUSION AND FURTHER WORK In reverse engineering, data gained from different sources that have different resolution and accuracy level require to go through several modelling stages. In this paper, the automation of the previously developed modelling processes, which can

8 Digital Enterprise Technology be time consuming, tedious and expensive is introduced. By using this automated system, the user does not require extensive expertise for CAD modelling. This system is most suitable for parts which have sensitive machined features and also complex but non-sensitive geometry on their casting surfaces. The results are reliable in accuracy due to the use of the actual measurement data in the sensitive features creation. No approximation is in place for sensitive and accurate features. Moreover, the resulting CAD model can include dimensional and approximated or actual tolerance information, which makes it more suitable for CAD CAM use. Further development of the automated data-fusion method will focus on better accuracy and faster modelling processes, by employing more automation in the system. The system can also be integrated with automated system for machining process planning and measurement systems. Tolerance analysis can also be integrated into the system by parametric feature creation. 8. ACKNOWLEDGEMENT The work has been carried out as part of the EPSRC E-IMRC at University of Bath under grant # GR/R67597/01. The authors wish to thank Phil Williams of Renishaw and Peter Smith from Konica Minolta Photo Imaging UK for their technical inputs. 9. REFERENCES 1. Bradley C and Chan V, A Complementary Sensor Approach to Reverse Engineering, Dept. of Mech. Eng., Uni. of Victoria, Canada, Trans. of ASME, Vol. 123, pp74-82, 2001 2. Chan VH, Bradley C and Vickers GW, A multi-sensor approach to automating co-ordinate measuring machine-based reverse engineering, Dept. of Mech. Eng., Uni. of Canada, Comp. in Ind., Vol. 44, pp105-115, 2001 3. Jamshidi J, Mileham AR, Owen GW, A Laser Scanning Registration Technique for Reverse Engineering Applications, IMRC, Uni. of Bath, Int. Conf. on Manu. Research, Cranfield Uni., UK, 2005 4. Jamshidi J, Mileham AR, Owen GW, A Proposed Data-cloud Reduction Method for Laser Scanned and CMM Data Integration, IMRC, Uni. of Bath, Int. Conf. on Manu. Research, John Moores Uni., Liverpool, UK, 2006, to appear 5. Jamshidi J, Mileham AR Owen GW, Dimensional Tolerance approximation for Reverse Engineering applications, IMRC, Uni. of Bath, Int. Design Conf., Dubrovnik Croatia, May 15-18, 2006 6. Jamshidi J, Mileham AR, Owen GW, High Accuracy Laser Scanned View Registration Method for Reverse Engineering using a CMM generated CAD Model, IMRC, Uni. of Bath, Int. Design Eng. Tech. Conf., DETC, 32nd Design Auto. Con., Philadelphia, Pennsylvania, September 10-13, 2006 7. Jamshidi J, Owen GW, Mileham AR, A New Data Fusion Method for Scanned Models, IMRC, The University of Bath, JCISE Special Issue on 3D Computation Metrology, 2006, to appear 8. Li YF and Liu ZG, Method for determining the probing points for efficient measurement and reconstruction of freeform surfaces, City Uni., Hong Kong, Measuring Sci. Tech. Vol. 14, pp1280-1288, 2003 9. Rooks B, A Vision of the Feature at TEAM, Sensor Review, Emerald Group Pub. Ltd., Vol. 1, No. 2, pp 137-143, 2004 10. Shen TS, Huang J and Menq CH, Multiple-Sensor Integration for Rapid and High-Precision Coordinate Metrology, Ohio State Uni., Trans. on Mechatronics, Vol. 5, No. 2, pp110-121, 2000 11. Shen TS and Menq CH, Automated Camera Calibration for a Multiple-Sensor Integrated Coordinate Measurement System, Ohio State Uni., IEEE Trans. on Robotics and Auto., Vol. 17, No. 4, pp 502-507, 2001 12. Spyridi AJ and Requicha AAG, Automated Programming of Coordinate Measuring Machines, Comp. Sci.Dept. and Inst. for Robotics and Int. Sys., Uni. Of Southern California, IEEE, 1994