Optical Tool Setting and Control for Precision Lathe BY Charles
Contents
1. Name default From State Start File Path C Program FllesiNational Instruments Vision Bullder 3 5 Vision Tool Setter Completeplusaveraging vbal Creation Date Wednesday April 09 2008 1 00 30 PM Last Modification Date Thursday April 24 2008 2 59 08 Print Date Tuesday May 20 2008 10 58 00 AM 43 pur hets EME 0 0 any ZUR RN 1201 Name default From State Apply Cal To State Inspect Transition always true ETERNI pope Intensity From State Check ii Apply From State Results rme E Transition active if Results OK Step Status psmu Transition 12 Name default From State Results OK To State Results OK Pricetty 3 VN te Name Fail From State Acquire from GigE To State Quit Transition active if Acquire Image Step Status is FALSE pg m eik Name Loop Completed From State Loop Count State Update Display Transition active if Calculator 1 is TRUE Name default From State Loop Count To State InspectStart med Inspection Name Completeplusaveraging vbai C Program FllesiNational Instruments Vision Bullder 3 5 Vision Tool Setter Completeplusaveraging vbal Creation Date Wednesday April 09 2008 1 00 30 PM Last Modification Date Thursday April 24 2008 2 59 08 Print2. General Description T bed lathes are type of 2 axis lathe where the configuration 15 such that the X axis and Z axis are not physically attached and the Z axis carry the spindle system An example T bed lathe is illustrated in Figure 7 Figure 7 T Bed Lathe T bed lathes are most often used in machining accurate profiles such as optics or spline based profiles For this reason many contain hydrostatic or air bearing 12 spindles to reduce vibrations and thermal growth concerns Correct tool and nose radius offsets are necessary to achieve the desired accuracies Tool offsets and effect on cutting path The ability to utilize tool offsets is a tool available on many CNC controllers Tool offsets are used to correct for differences in individual tools By utilizing this feature part programs can be generic to the tool being used Tool nose radius compensation goes one step further by also compensating the programmed path based on the tool radius value Appendix D includes figures of offset errors as well as a diagram of the corresponding error transferred to the part 2 7 Lathe Tool Inserts Lathe tool inserts are the component performing material removal on a lathe Because of costs related to custom manufacturing high tolerance tool inserts a generic set are purchased through a commercial manufacturer This lowers operational costs at the expense of relatively low tolerances for any given tool The manufacturer guarantees th
3. Combined Error Note similarity of Red and Blue lines from different sources of error Green Nominal Red X shift Blue TNR Small AND 2 shift 54
4. Threshold Eius Threshold Mean 02 4 Stier 2 3522 192 White Threshold Mean BOE Ster 0 7555 1 Threshold i Blue Threshold nite Threshold Meen 202 2 StDev 1 811 ic hie Threshold Heen BOL 1 505 im Figure 13 Results from Light Threshold Experiment A visual comparison of the blue light and white light images Figure 14 clearly shows the Airy disk in the white light image as a series of concentric radiuses going toward the right from the edge of the tool This was further assurance that the blue light would provide better measurement results 30 White Lig Figure 14 Blue Light vs White Light Comparison Depth of Focus To check the vision system for error induced from the tool out of focus truly an indication that the tool height was set improperly a set of data was recorded and deviations and standard deviation calculated with the tool in 5 different positions 0 0508 mm 0 0254 mm 0 000 mm 0 0254 mm 0 0508 mm These positions were chosen after determining the system had trouble detecting the radius at 0 0508 mm The results are presented in Table 3 Table 3 Depth of Focus Data Deviation Std Dev um 0 0000 0 0000 0 0000 0 0000 0 1457 0 0572 0 4130 0 0254 0 0706 0 0056 0 4645 0 2755 0 0534 0 4074 0 0508 1 4625 0 5115 1 4689 0 6809 0 2097 0 9245 31 The maximum deviation of the cal
5. Step 7 NN 2 StpiS Radus Tolerance Bands E jeg re a RndCrcMartdge2 FndOrwartde 991 Set vartabie wen 19 OVerlay Results Custom Overlay Data Logging 1 Data Logging Inspect Calbrate Image Set Inspection Status Inspection Completeplusaveraging vbai File Path C Program Files National Instruments Vision Bullder AI 3 5 Vision Tool Setter Completeplugaveraging vbal Creation Date Wednesday April 09 2008 1 00 30 PM Last Modification Date Thursday April 24 2008 2 59 08 PM Print Date Tuesday May 20 2008 10 58 00 AM 42 T Lud Set Varinble 1 Intesity Check i inspection step 3 inspection steps User Input 1 Calculator 1 Calculator Inspection Status ustom Overlay lu di Quit E d il i 1 Inspection step Acquire trom GigE 5 inspection steps Update inepecton UI Image Logging Set Inspection Status 22 22 22 Inspection Status InspectStart Inspection step Update Display 4 inspection steps Inspection Steps Kame Type Step 1 Calc Average Results Copy of Set Variable 1 Delay 1 Custom Overlay 1 1 i ili i t States L p d L KEES lt Lo
6. edle ede eee ev ee 7 24 LIGHTING so E c cR 9 2 5 VISION SOFTWARE NI LABVIEW W VISION TOOLKIT ccccccccessceceessceceeseeeecsssccceesseeecessseceenseeeceeues 11 2 0 T BED TZATHE ceto e c E ditte a b 12 2 7 a oot eeen ee d a d tren RT Sits BEER 13 2 8 OPEN ARCHITECTURE CONTROL 13 3 1 TOOL ves eta eee tree oet keit ee v eee E e Pe 15 3 2 TOOL WEAR MONITORING 17 3 3 TOOE SETTING SYSTEM 18 4 1 SOFTWARE E E ATE EAE EEE 21 42 CALIBRATION 23 A ADI qUNGe BINE 26 WHITE EIGHT AND BLEUE LIGHT SOURCE vec re e 28 DEPTH eR tte tu tn d 31 CALIBRATION CHECKS EH ER Torte e aes Gabe seas Eee oe erp 32 TOOL RADIUS AND OFFSET 221 2 0 1000000000000000000000000000010 000000 eet nenas 34 iii TIMESTUDY eoa
7. optical comparator setting the comparator up estimating the tool nose radius and returning to the machine The tool insert had to 35 be installed into tool holder tool height had to be set and then insert had to be touched off of the part to set the offset values The total time averaged 11 5 minutes The new process eliminates all but the manual insertion of the tool to the tool holder and setting of the tool height Mean time during testing for swapping a tool setting height and obtaining tool offset values was 3 2 minutes 36 6 Conclusion A machine vision based optical tool setter was developed for checking the radius of lathe tools The system had an average repeatability of 8 2 um with an accuracy average of 2 5 The standard deviation of the measurements was improved by not applying thresholding to the image data and using a blue light source software based calibration routine was used to calibrate the camera to the full field of view While the overall system did not meet the goal of 5 micron repeatability there was still potential for improved part tolerance control The system was also capable of automatically updating the CNC control tool offset register to reflect the measurements Additionally the system could eliminate 8 3 minutes from the manual tool process Finally a source of human error was eliminated as it was no longer necessary to use judgment in determining the tool nose radius value
8. represents the portion of the image that was not calibrated due to the spacing of the grid dots From Figure 17 the calibration error map predicted at most 4 um of error centered at Position 2 For a comparison to a distorted lens and corresponding error map from National Instruments see Appendix C Zshift Zshift cam P aa X 8 X shift 2 5 Camera Field of View i X shirt i 1 i un Lid 3 4 Figure 16 Vision System test locations 33 Figure 17 Calibration Error Tool Radius and Offset Performance The largest portion of testing was devoted to performing actual radius and offset measurements in order to characterize the entire system including errors from the lathe order to get an idea of the attainable accuracy and repeatability For this reason no attempt was made to separate the machine errors from the vision system errors In the first series of tests tool inserts with radius values of 203 2 um 428 75 um and 792 48 um were measured 30 times at the 5 positions used in the calibration tests Deviations were calculated for each test and are presented in Table 4 The data showed systematic errors for all measurements performed see Mean Deviations listed in Table 4 When the data was analyzed at the individual
9. the axes The result of this step was a calibrated vision system 24 capable of outputting real world units with rotation of the camera accounted for the calibration process This was a benefit to using the built in calibration as no additional math operations were needed to calculate true position of the tool Figure 11 Calibration with Rotation The second requirement was to calculate and set the offset from the machine tool coordinate system to the machine tool X centerline and Z offset locations This was accomplished by using a tool insert that had a known tool nose radius and moving the tool until contact is made on X and Z surfaces of the faceplate of the spindle This position was set as the new machine origin as shown in Figure 12 The radius and the offset from the camera coordinate system origin were calculated Xt and Zt in Figure 12 The offset from the set machine origin is Xm and Zm respectively The calculation of the camera offset is then of the form X m X a 2 25 N m Xm Machine Xcam Origin 1 X Camera cam Origin i 2 3 Camera Field of View 4 X gt Figure 12 Coordinate Systems Once this was calculated it was stored in the vision system software for use when generating tool offsets 4 3 Data Collection The data collection from the measurement program was per
10. the field of view to be used as a calibration gauge see Figure 8 He also used a histogram to programmatically check if the light source was active He found the accuracy of the system to be about 2 5 micron for a field of view of 1 4mm The drawbacks of this system were that the calibration was only good for a small percentage of the viewable area and thus larger radius tools could present problems Also the gauge object would need to be protected from any physical contact which could alter the shape of the gauge Further the tool could be damaged if it were to contact the gauge object The system being presented eliminates these drawbacks No gauge object is necessary and a full field calibration is performed to remove any non linearity from the lens system Scale gauge Figure 8 Doiron s Calibration Object Reddy 16 developed a non contact tool setting system to accurately position a micro diamond tool using a CCD camera and a microscope He tested various 16 disturbances to system such as defocus inclined tool and lighting on accuracies of the system and applied Chauvenet s criterion to eliminate spurious results He found that digitization errors from not using sub pixel processing errors in the vision system calibration from lens distortion and machine vibration were the cause of most inaccuracies and non repeatability of the system The system being presented utilizes subpixel processing is full field cali
11. Ci Bea 35 6 CONELUSION Prope RR 37 6 1 RECOMMENDATIONS terere ee i ee e t erre etri ees 37 APPENDIX NI VBAI CODE erronee trae etra no evo Fo re eva sors UY eee ela 0a 41 APPENDIX TCP SOCKET SERVER 46 APPENDIX C EXAMPLE OF CALIBRATION ERROR 50 APPENDIX D TOOL OFFSET AND TOOL NOSE RADIUS ERRORS ae sase sane oonc onse enseoneens 51 iv 1 Introduction 1 1 Motivation A commercial fast and accurate way of determining the tool nose radius necessary for machining complex profiles on a lathe via an in process technique does not exist There are many different compensation methods that can be used on the machine i e leadscrew and volumetric compensation the part i e dynamic fixture offsets and the tool 1 e laser tool setters contact tool setters offline tool measurement to improve part tolerance and finish Commercially available contact and non contact tool setters do not facilitate the measurement of the actual tool nose radius and in high precision contour work the tool nose radius is a critical factor Further by utilizing the Application programming interface API of an open control system the measurements and the corresponding values can be programmatically entered into the Computer Numerical Control CNC without user intervention By developing and implementing
12. Date Twesday May 20 2008 10 58 01 AM 44 Transition always true Name default Transition always true ee To State Acquire from Pricetty 1 NENNEN Transition 19 Name default From State Update Display To State Results OK Pricetty 1 File Path C Program FllesiNational Instruments Vision Bullder AI 3 51 Vision Tool Setter Completeplusaveraging vbal Creation Date Wednesday April 09 2008 1 00 30 PM Last Modification Date Thursday April 24 2008 2 59 08 Print Date Tuesday May 20 2008 10 58 01 AM 45 APPENDIX B TCP Socket Server Code MDSI Socket Server Process Command Set up server on IPADDRES address SOCKET port Connection Requested Connection YES Closed Accept Client Connection Data from Client YES Receive Data ona cot or MDS YES Data cO or MDSI Init MDSI Library MID machinelD MDSI Libray Init Succeded YES Command succeded YES Send Result to Client Send ERR to Client 46 Imports Imports Imports Imports Imports Imports Sys Sys Sys Sys Sys mds tem tem IO tem Net tem Net Sockets tem Text iMacroSupportVBX Public Class MDSITcpServer Shared Shared Shared Shared Shared Shared Shared Shared Public End Th Dim Dim For Nex Sub machineID As String IP As String Port As String bDebug As Boole
13. ING LOOP Catch e As SocketException Console WriteLine SocketException 0 Finally server Stop End Try End Sub Main Public Shared Function parse Data ByVal value As String String Executes a statement if data contains mdsiMacroObj and B Evaluates a statement if data contains mdsiMacroObj C If cO is present attempt to remove tool offset Dim exec cmd As Boolean value Contains parse Data Err If value Contains mdsiMacroObj initret InitMdsiLibrary Connect to MDSI Variable Database If initret Then Dim Script Engine New MSScriptControl ScriptControl Script Engine AllowUI False Script Engine Language VBScript Script Engine AddObject mdsiMacroObj mdsi Try If exec cmd Then mdsi mdsiReadFreshDataVB False 48 Script Engine ExecuteStatement value parse Data OK 1 mdsi mdsiReadFreshDataVB False parse Data Format Script Engine Eval value 000 00000 End If Catch ex s Exception Console WriteLine Script Engine Returned 0 ex Message ToString parse Data Err End Try Reset scripting engine and uninitialize Script Script Engine Nothing mdsi unInitialize End If ElseIf value Contains cO Then Format is form 0 where is tool offset register to remove Do this to ensure data is o k value value Remove 0 2 Console WriteLine Clearing Offset Table Row 0 value Dim j As Integer Dim
14. Optical Tool Setting and Control for Precision Lathe BY Charles Geoffrey Haning Submitted to the graduate degree program in Mechanical Engineering and the Graduate Faculty of the University of Kansas in partial fulfillment of the requirements for the degree of Master of Science Chairperson Committee members Date Defended The Thesis Committee for Charles G Haning certifies that this is approved Version of the following thesis Optical Tool Setting and Control for Precision Lathe Committee Chairperson Date Approved Acknowledgements I would like to thank my employer for the opportunity to participate in the Technical Fellowship Program and the countless employees who helped along the way I must especially thank John Fines and Todd Mills for the guidance they have provided I must thank my family for their love and support Tiffany I couldn t have done this without you and your dedication I Love You 2 T2EMOTIVATION 4 ee E eed e eure er andreas d te OE Fen e EE dede 1 LO RESEARCH METHODOEOGY 5 A dee e ctt tet eure eee duo Ose Sandee ioe os 1 153 LHESIS STRUCTURE strengen leenen Tao Rs 2 2 1 VISION HARDWARE inerte cite hice deter benne 3 2 2 3 2 3 OPTICS SYSTEM tere ie eed
15. a vision based measurement approach tool nose radius can be calculated in addition to tool offsets with the goal of reducing tolerance loss from incorrect tool nose radius and tool offset values 1 2 Research Methodology The focus of this research was on the development of a machine vision system to provide automatic tool offset and tool nose radius compensation on a lathe used on the shop floor To accomplish this work a T bed lathe which utilized a PC based control was used for development and testing National Instrument Labview 8 5 with the Vision Development System as well as National Instruments Vision Builder for Automated Inspection was utilized to develop the machine vision software Microsoft Visual Basic 2005 was used to write the interface between the vision software and the PC based controller 1 3 Thesis Structure This thesis is organized into six chapters The first chapter discusses motivation and introduces the methodology used to perform the research The second chapter discusses the background information including the systems materials and machines used in addition to theory of operation The third chapter is a summary of finding during the literature review The fourth chapter discusses the Experimental Design and explains how data was recorded The fifth chapter discusses the results of the work performed The sixth chapter includes Conclusions and Recommendations of the current work 2 Background The foc
16. ailable from http www baslerweb com beitraege beitrag en 23777 html 4 Basler Inc Basler Pilot User Manual 5 Zuech N Understanding amp Applying Machine Vision Second Edition 2000 New York NY Marcel Dekker Inc 6 National Instruments Inc National Instruments GigE Information 7 Navitar Inc Lens System Performance Data 2007 8 Menn N Practical Optics 2004 Elsevier Academic PRess Burlington MA p 61 73 9 Edmund Optics Infinity corrected and Plan objectives cited Available from http www edmundoptics com techSupport DisplayArticle cfm articleid 277 10 Laikin M Lens Design Third Edition Revised and Expanded 2001 Marcel Dekker Inc p 36 39 133 134 265 11 MellesGriot Optics Info 2007 cited 12 MDSI Inc OpenCNC DataSheet 2008 13 Maali F et al VISION BASED TECHNIQUE FOR AUTOMATIC TOOL SET UP IDENTIFICATION AND WEAR MEASUREMENT IN CNC LATHE 1984 Las Vegas NV USA ASME New York NY USA 14 Nobel G M Tool nose radius deviation error compensation simulation and implementation 1988 15 Doiron T D Computer vision based station for tool setting and tool form measurement Precision Engineering 1989 11 4 p 231 238 16 Reddy G G Accuracy of a vision based tool setting system for micro diamond tools 1994 Louisiana Tech University 17 Varga G Advantages of the CCD camera measurements for profile and wear of cutting tools Journal of Physics Conference S
17. an True ret As mdsiMacroReturnTypes mdsi New mdsiMacroSupportX units As mdsiUnitBasisModes initret As Boolean Shared Sub ParseCmdLine is currently isn t checking for correctness of the command line arguments As String 3 As Integer 0 Each s In My Application CommandLineArgs If s m Then machineID My Application CommandLineArgs j 1 ElseIf s a Then IP My Application CommandLineArgs j 1 ElseIf 3 p Then Port My Application CommandLineArgs j 1 End If j 1 t Public Shared Sub Main Dim ser Con Try server s TcpListener ver Nothing sole TreatControlCAsInput True ParseCmdLine Get IP Port and machine ID from Command Line Dim localAddr IPAddress IPAddress Parse IP server New TcpListener localAddr Port Start listening for client requests server Start Buffer for reading data Dim bytes 1024 s Byte Dim data String Nothing Enter the listening loop This loop continues until program is exited While True LISTENING LOOOP Console WriteLine Waiting for a connection 0 server LocalEndpointw ToString Perform a blocking call to accept requests You could also user server AcceptSocket here Wait for Connection here Dim client TcpClient server AcceptIcpClient Console WriteLine Connected from Client 0 client Client v LocalEndPoint ToString If coming back into loop make sure data No
18. brated and the shutter time of the CCD camera is very low resulting in vibration being less of a concern Varga 17 investigated the use of a CCD camera system to compare the in use grinding wheel profile based on a known good profile This was implemented on a machine with a separate computer to run the vision application No geometrical accuracies were presented He proposed this system could be used for tool wear monitoring 3 2 Tool wear monitoring There have been several authors who have worked with machine vision for the purpose of monitoring tool wear This is not the focus of the presented research and is provided for further reference to machine vision applications Kurada and Bradley 18 used a CCD camera to monitor wear of a lathe tool by monitoring wear pattern reflections and their corresponding grey level differences Wong 19 used a CCD camera to monitor tool flank wear similar to Kurada and Bradley Jurkovic 20 used a CCD camera and a scanning laser projected onto the tool to detect wear by analyzing the line pattern generated by the laser thus enabling height detection of the tool wear Kerr 21 developed and verified Kurada and Bradley s work 17 3 3 Tool setting system Laser The principle of operation for commercially available non contact laser tool setters is simply electronics monitoring for beam interruption Renishaw NC4 non contact laser tool setter can have accuracies as good as 1 0 micron 22 The b
19. culated radius was 1 5 um at a focus point of 0 0508 mm X maximum deviation was 0 5 at 0 0508 mm and Z maximum deviation was over 2 4 at 0 0508 mm At 0 0508 mm out of focus the inspection process of the vision software had trouble identifying a suitable edge so more than ten measurements were attempted to get ten good measurements When this occurred the circle fit algorithm used a reduced set of data to perform its circle fit This reduction in points is illustrated in Figure 15 where the captured data points appear as the bright dots at the edge of the dark radius representing the curve fit The circle fit algorithm uses a minimum change in pixel value as a cutoff for determining what is an edge and as a result of the out of focus condition this cutoff wasn t met Outof Focus Points Focus Points Figure 15 Depth of Focus Data Points Calibration Checks To check the calibration of the system tools whose radius had been measured to five decimal places on a separate calibrated optical vision system were inspected in 32 five different regions of field of view as shown in Figure 16 Thirty measurements were taken at each of the five positions and the data was analyzed for variation between the measurement positions The error map produced by the calibration routine was evaluated to determine if there was any correlation between the data This is shown as a graphic in Figure 17 The dark outer region
20. dardized digital interfaces used for industrial machine vision such as Firewire IEEE 1394a and 1394b and Camera Link Camera Link The GigE standard ensures the behavior of the host and the camera and also provides a standard approach to access any GigE compliant camera GigE makes it possible for data transfer up to 100 meters 6 GigE was used because of its relative low interface cost a standard gigabit Ethernet adapter card 15 all that 15 needed and for the distance of data transfer needed 2 3 Optics System The optics system is made up of three components The first and second is a Zoom 6000 assembly from Navitar Inc that included the zoom lens which creates adjustability in the field of view FOV and zoom of the system as well as a tube lens to focus the image onto the CCD sensor of the camera The third is a Mitutoyo Inc 5X Infinity Corrected Flat Field Microscope objective Together these three components created a range in the system field of view FOV of 3 21mm x 2 41mm at a pixel resolution of 2 1 microns to 0 77mm x 0 58mm with a pixel resolution of 0 6 microns respectively 7 The Mitutoyo lens is an infinity corrected objective This means the image it produces is projected to infinity parallel rays and an additional lens 15 required to focus image onto camera sensor 8 This additional lens is called a tube lens A diagram labeling the components is shown in Figure 3 and a ray diagram of t
21. e tool nose radius to 40 0014 inches for the radius range of 0 002 in to 0 03125 inches 2 8 Open Architecture Control The controller on the T bed lathe is a software product by MDSI named OpenCNC OpenCNC is an unbundled modular all software based CNC control system 12 OpenCNC runs on a generic IBM compatible PC running Microsoft Windows 2000 or Windows XP operating system and uses the add on Real Time Extensions from Ardence Inc to provide the deterministic response necessary for 13 machine tool control One of advantages of OpenCNC is a published Application Programming Interface API allowing access to internal functions and variables for development of additional functionality such as the vision tool setter The development tools are available through Microsoft and Microsoft Visual Basic 14 3 Literature Review During the literature review several works were examined but relatively few dealt with either tool offset or TNR measurement Many focused on measuring tool wear only For this reason the literature review is split into three sections The first is Tool Measurement which presents an overview of tool offset and TNR measurement systems The second section presents an overview of systems which perform tool wear monitoring with machine vision systems The third section is an overview of commercially available devices and discussion of other concepts that were considered in the place of the currently pres
22. enefits of this approach are that it is a commercially available product it has high accuracy and can perform in a harsh environment In order to perform tool nose radius measurements the tool would have to be incrementally moved to trigger the beam to generate a series of points along the radius of the tool as shown in Figure 9 Figure 9 Example Search Pattern However laser non contact sensors are incapable of detecting object below a certain size The investigated tool setter claimed a minimum radius of 30 micron 001 inches For a 002 inch radius lathe tool this would result in the tool having to be moved in more than 0001 inches past the actual end point of the tool As this 18 error would vary for different sized tools and different tool geometries it would complicate any attempt to accurately measure the tool nose radius Optical Micrometer Commercially available optical micrometers provide high resolution up to 0 5 micron The principle of operation is that a led source is used with optical lensing to create a beam of light think of a 2 D laser The detector is generally a line scan CMOS or CCD camera 23 An approach similar to the laser tool setter would be used with the difference being the iteration would only have to use one axis of motion and then record both the distance to the first point and the thickness recorded by the optical micrometer Drawbacks to this approach are the same as the laser t
23. ented system 3 1 Tool Measurement Maali 13 proposed using a 2048 pixel line scan CCD camera and a microcontroller to provide tool set up capability by moving the tool thru the response line of the camera to build a 2048 x 2048 pixel image By doing so and then looking for the highest pixel in the array the Z and X offsets could be calculated from the set of data Drawbacks were the amount of time required to iterate through the small displacements to generate a good data set and the tight integration between the control and the camera system The system being presented eliminates these drawbacks by utilizing a full frame CCD camera to capture a full 2 D view of the tool edge Nobel 14 interfaced a standalone vision system with an Allen Bradley 8200 control thru the use of a custom interface computer to perform real time compensation of tool nose radius from a true radius The limitations were the program could only react to current machine position 1t had no access to the path 15 planning portion of motion control system This resulted in discontinuity errors on abrupt changes of motion such as corners and axis reversals Through cutting test parts that avoided abrupt changes of motion he was able to produce a small reduction in the overall form error of the part of approximately 10 um Doiron 15 conducted his research using a CCD camera connected to a microscope and used the novel approach of including an additional object in
24. eries 2005 13 p 159 162 18 Kurada 5 Bradley Machine vision system for tool wear assessment Tribology International 1997 30 4 p 295 304 19 Wong Y S et al Machine vision monitoring of tool wear 1998 Boston MA USA SPIE Bellingham WA USA 20 Jurkovic J M Korosec and J Kopac New approach in tool wear measuring technique using CCD vision system International Journal of Machine Tools and Manufacture 2005 45 9 p 1023 1030 21 Kerr D J Pengilley and R Garwood Assessment and visualisation of machine tool wear using computer vision The International Journal of Advanced Manufacturing Technology 2006 28 7 p 781 791 22 Renishaw Inc Renishaw NC4 Data Sheet cited Available from http www renishaw com en 6099 aspx 39 23 24 25 26 27 28 Optical Micrometer Data Sheet 2007 Blum 76 Contact Tool Setter Info 2007 Heidenhain TT 40 Contact Tool Setter Info 2007 Marposs 718 Contact Tool Setter Info 2007 Renishaw RP3 Contact Tool Setter Info 2007 National Instruments Inc National Instruments Vision Concepts Manual 40 APPENDIX A NI Code Flow Chart 41 Run LabVIEW VI 1 Set Variatie 1 B xem RE o dem LN vac Find Edges Bisector 142 pe e Step 5 Line2 Angle Geometry Step 6 ROT Position Min Max Angles Calculator jr sce a en
25. formed by adding a step to the inspection in VBAI which enables measurements to be logged to file The first experiment was to check the advantage of blue wavelength light over white multi wavelength and to compare the effects of using thresholding techniques to enhance edge contrast of the acquired image Data sets of 100 individual measurements were used for each test The second experiment was to evaluate the effect of defocus on the measurement system Five different positions were evaluated 26 corresponding to 0 0 001 0 002 inches representing offset from correct focus distance A third experiment was performed to check effect of light intensity on the measurement system The final experiment was designed to check the accuracy of the calibration performed as well as analyze the performance of the vision system After the vision system was calibrated three measurements were taken each measurement is an average of 10 individual calculations with the tool at five different locations within the field of view The calibration was evaluated based on the error of the measurements 27 5 Results White Light and Light Source To evaluate the effect of light wavelength and the use of thresholding on the results 100 individual measurements were conducted for four cases e White light source without thresholding Blue light source without thresholding e White light with thresholding e Blue light with t
26. he lens system is shown in Figure 4 fp toom Assembly rm Tbe lers Odjiective lers Figure 3 Lens Assembly Camera Tube Lens Objective Lens Sensor Figure 4 Lens Diagram The Mitutoyo objective is also a plan achromat design A plan objective corrects for color and spherical aberrations better than other designs and has a flat field over 95 of the imaging area 9 The 95 area radius equals R J 95 R or 9767 There were several parameters that had to be determined during the design process A tradeoff had to be made between the FOV and system resolution because the two are inversely related The optics system was designed to resolve features to 2 um for the desired field of view 2 6mm x 2 0mm This allowed the largest radius tool up to 1 5mm to fully be in the field of view of camera Additionally this provided user larger window to approach when positioning tool within the FOV According to Zuech accuracy and repeatability of a machine vision system can be improved by averaging several images together and the improvement equals the square root of the number of images averaged 5 This assumes negligible outside factors such as vibration and lighting variation Depth of focus DOF was an important consideration as it was 2 orders of magnitude smaller than the FOV or 20 um 0 000787 inches The calculated DOF indicated the completed system would be dependant on accu
27. hresholding The data shown in Table 2 is separated into three sections radius X and Z Within each table the average max min maximum deviation max min and standard deviation sigma was calculated for each of the four cases The maximum deviation was almost double for the experiments using thresholding In the same way the standard deviation was at least double for the thresholding cases 28 Table 2 Light Source Data Results Blue 470nm White 550nm No Threshold Threshold Threshold Threshold aay ae El 802 421 804 166 801 816 2 350 0 388737125 1 510792212 0 75888073 1 508002603 BE EE Average 172 1282 176 6138 172 1822 175 0963 Max 171 4014 174 3567 171 0979 172 5606 173 863 2 461 4 696 0 421452865 1 40213634 4653 lt gt 366 256 354 953 366 571 355 962 365 382 353 746 1 189 igma 0 18910262 The data was analyzed normal distribution was fit to data Results can be seen as a plot versus a Normal distribution in Figure 13 Blue light and no thresholding had considerably less variation than any other experiment data Using the threshold function more than doubled the standard deviation of the measurements in every case so thresholding was abandoned as a technique to be used Further the blue light source was used exclusively as a way to further improve measurement data 29 Blue Threshold White
28. ill factor Column Column maman Column Column Column Column Column Column Vert Vert Vert Vert Vert ert Vert Vert Ve Shift Pixels Shift Pixels Shift Pixels Shift Pixels Shift Pixels Shift Pixels Shift Pixels Shit Pixels Reg Re Reg Reg Reg Reg Reg Reg Add dy BE EEE amp amp di kb amp kihi4A4RAkikkkdhdh Ee i i i d E A amp iiiiEiisididdh idhkd bd TEES EG GI C E E ER FR ET REIR ERES TTITITTTITTTITTTIT Figure 2 CCD Sensor Digital Interface In previous generations of machine vision cameras output was analog video and required a frame grabber to digitize the signal Many cameras now output data thru a digital interface because the camera performs the analog to digital conversion process internally The result is improved noise immunity of the transmitted data The particular camera used is capable of outputting quasi 16 bit 65536 gray levels 12 4096 gray levels and 8 bit pixel depth 256 gray levels images at up to 12 frames per second fps at its full resolution of 2058x2456 pixels It outputs this data over a digital interface called GigE which is an implementation of gigabit Ethernet for industrial machine vision applications GigE for vision applications has been standardized by the Automated Imaging Association AIA and represents several stan
29. k As Integer initret InitMdsiLibrary Connect to MDSI Variable Database If initret Then Dim num axes mdsi axisCount Dim done As Double For j 0 num axes 1 For k 1 4 Select Case Case 1 2 mdsi Offset j k value ToString False 0 0 done mdsi Offset j value ToString False Used to check if row actually clear Case 3 4 mdsi Offset 0 k value ToString False 0 0 done mdsi Offset 0 value ToString False Used to check if row actually clear End Select Next Next If done 0 Then parse Data OK End If End If mdsi unInitialize End If End Function Public Shared Function InitMdsiLibrary As Boolean Dim sinit As String mid machineID amp jsID 0 locale 0 If mdsi initialize sinit False False 1 Then 1 Failure Console WriteLine MDSI VBX Init Failure InitMdsiLibrary False Else InitMdsiLibrary True Console WriteLine MDSI VBX Init Success End 15 End Function End Class MDSITcpServer 49 APPENDIX Example of Calibration Error For a further example of calibration error mapping here is an example from National Instruments 50 APPENDIX D Errors Tool Offset and Tool Nose Radius Green Nominal Red X Shift 51 Green Nominal Red X Shift Green Nominal Red Z Shift 52 Green Nominal Red Z Shift Green Nominal Red TNR Large 53 Green Nominal Red TNR Small
30. mated Non collimated light source The result is a reduction in perspective errors independent of viewing distance and an improved edge contrast thus reducing size error By considering the peak response of 10 Basler camera 1 Figure 6 optics system Rayleigh limit was determined to be 2 05 um 0 00008 inches Relative Response 0 9 0 8 0 7 0 6 0 5 0 4 0 3 0 2 0 1 0 0 400 500 600 700 800 900 Wave Length nm Figure 6 Camera Response 1 2 5 Vision Software NI Labview w Vision Toolkit for Automated Inspection VBAI was used for its compatibility with a wide variety of machine vision cameras and the flexibility it provided in the design of the vision National Instruments Labview 8 5 with the Vision Toolkit and Vision Builder 1000 application Many tools are available thru the vision toolkit providing an easy way to create and deploy machine vision gauging applications VBAI provides many of the tools available with Labview and the vision toolkit in a graphical programming environment by the use of flow charting and sequential functions Utilizing VBAI 11 reduced complexity of programming involved and presented an improved software solution that involved less custom generated code than creating a similar application using a text based programming language such as Microsoft Visual Basic 2 6 T Bed Lathe
31. oolsetter the minimum detectable object size is greater at 002 inches The entire tip of smaller tools those with 0 002 inch radius would be ignored before the optical micrometer would finally register that the tool had crossed the plane of measurement and this would be unacceptable Contact Renishaw Blum Marposs and Heidenhain all have physical contact based tool setting devices on the commercial market The principle of operation is sensing a displacement applied to the unit either thru mechanical strain gauge or optical sensing methods 24 25 26 27 The drawbacks include the force required to detect the tool can damage fragile diamond tools Additionally only approximations of the tool nose radius can be made unless many touches are made along the edge of the tool using a perfectly round physical detector with a known radius This would be 19 time consuming and further exaggerates any error present in measurement device to the tool nose radius Offline The currently employed method of checking the tool nose radius on the shop floor is to use an optical comparator to measure the radius of their tools offline before use There are many drawbacks including 1 Inaccuracies when installing the tool on the machine 2 Operator is still forced to touch off the part to set X and Z offsets 3 Operator must use judgment on the radius of his tool 20 4 Experimental Setup In order to get relevant results from the vision
32. or while setting the touch off points on the part 6 1 Recommendations In order to obtain better calibration results a calibration grid with a much finer grid spacing and dot size would need to be utilized The ideal grid for the current camera and field of view according to the National Instruments documentation would be 15 um dots spaced every 30 um Further development of the software also an increase in complexity would allow current tool shapes to be stored in memory so it can later be recalled to check for tool wear and automatic shape recognition A zoom lens with powered zoom could be used to make more accurate measurements for smaller radiused tools Alternatively custom lens development 37 services could be contracted to develop a lens solution to address resolution and telecentricity concerns to provide a very robust lens design cost control would be a major concern Integration of techniques such as Pierce s criterion to identify spurious data or Hough transforms to identify curve data from line data Investigate White Light Interferometery for inclusion of tool height measurement concurrent with radius and offset measurement capabilities 38 Bibliography 1 Basler I Basler Pilot User Manual 2008 2 Jahne H Becker and Geibler Hand Book of Computer Vision amp Application Vol 1 1999 San Diego CA Academic Press 93 97 188 195 3 Basler 1 Image Sensor Technology 2007 cited Av
33. positions a trend did 34 emerge in repeatability with mean repeatability of radius value of 6 3 um and a standard deviation of 33 um Table 4 Tool deviations Measured Tool Deviations from actual um 203 2 428 75 792 48 Radius Radius Radius Average of 10 runs um X Z um um X um Z um um X Z um Position 1 1 1288 2 6197 4 2518 o eL p Ll Position 2 1 5488 1 3980 3 6998 3 4106 2 5357 2 3369 2 8801 1 1119 4 3641 Position 3 Position 4 Position EERE ee EN EN EEE SEE Mean Deviation Deviation The data was further analyzed over each position to get a representation of repeatability max error min error at any one location This is shown in Table 5 Table 5 Position Repeatabilit Repeatability at Radius each Position um X um 2 um Posiionf 6 8715 5 9585 Mean Repeatability 6 3412 3 3647 4 6040 22 pf 0 33021 1 0788 1 8395 Finally the mean was found to get an overall view of the accuracy and repeatability of the system shown in Table 6 Table 6 Overall Accuracy amp Repeatabilit Radius All Positions um X um Z um Repeatability 8 2472 4 9550 6 5002 Mean Deviation 0 8681 0 8170 2 3064 Deviation 2 7775 1 5736 2 2965 Time Study The previous method of determining tool insert radius was a manual process that involved walking to the nearest
34. rate tool height adjustment prior to making tool nose radius and offset adjustment 2 10 In all optic systems there is a fundamental limit to the resolving power of the system imposed by diffraction Diffraction occurs because of interference from obstacles result in secondary waves interfering with the original wave The Rayleigh criteria describes the resolving power of the system as the minimal distance where two objects separated by some distance can still be distinguished as two objects through the imaging system The Rayleigh criteria for microscope objectives is defined as R where NA is equal to the numerical aperture of lens given by the lens manufacturer This was found to be 2 05 um for the designed system NA 14 4 470 nm 10 2 4 Lighting In gauging applications backlighting 15 the preferred method of illumination when the object being measured has a 2 D profile that is planar Backlighting provides high contrast differentiation for object being measured by placing light source inline with the camera but behind the object This type of lighting can be further enhanced by making the light output collimated A collimated light source provides a parallel unidirectional beam of light This concept is illustrated in Figure 5 Lens Camera Lens Camera System System Collimated Non Collimated Light Source Light Source 22 Figure 5 Colli
35. sents several advantages mainly Higher Signal to Noise S N ratio lower dark noise and higher fill factor A higher signal to noise and lower dark noise ratio meant less noise would be introduced to the image data 3 The higher fill factor can result in sharper pictures due to the ability of the camera to read the image from the sensor producing less motion blur Additionally CCD cameras have better response at lower wavelengths of light which help improve overall system resolution A plot of a CCD and CMOS image sensor relative response to different wavelengths of light can be seen in Figure 1 CMOS vs CCD Relative Response Relative Response Wavelenth of light CMOS Figure 1 Camera Response Curve CCD imagers acquire an image in a three step process First photodiodes or MOS photo capacitors are exposed to light Second Charge Transfer occurs thru the use of shift registers which moves the packets of charge within the silicon substrate Finally output amplification via Variable Gain Control VGC and Charge to Voltage conversion via Analog to Digital conversion ADC take place 4 5 An example diagram of a CCD sensor is shown in Figure 2 With the particular CCD sensor utilized in the camera the imager is broken down into left and right areas and pixels are clocked out from both the left and right side of the sensor simultaneously to aide in a high f
36. spection data and displays it on the screen The Results step presents the operator a dialog box to accept or decline updating the tool offset table This dialog also includes a number input box to set which offset register to store the data If the operator accepts the tool offset values this data is transferred to the appropriate tool offset register via the TCP Socket Server See Appendix A for a flow chart of the code execution states and a diagram of the various steps TCP Socket Server The second application developed was a synchronous socket server written using Microsoft Visual Basic 2005 The purpose of this program was to handle communication between VBAI and OpenCNC using VBAT s internal TCP socket capability to communicate with MDSI OpenCNC The program allows a connection from a remote client via TCP sockets It also creates a connection to MDSI OpenCNC thru the use of an MDSI developed ActiveX control It then facilitates communications between OpenCNC and the client VBAI A flow chart and the code itself are presented in Appendix B 4 2 Calibration There were three requirements of the calibration in order to make accurate tool offset measurements These were Calibration to real world units the ability to 23 account for angular errors non squareness to lathe s XZ plane and physical offset between the machine coordinate system origin and the camera coordinate system origin Two of these factors were acco
37. system the appropriate software had to be created One application was developed to communicate between VBAI and the MDSI controller The other application performed all of the vision acquisition data analysis and displaying of results Secondly a calibration process had to be developed in order to get relevant information from the vision system Experiments were then performed in order to check the results of the calibration process and once this was satisfactory actual tool offset and TNR measurements were made to judge the accuracy of system Experiments were performed with different sized tools as well as different tool materials to determine the accuracy and repeatability of the system 4 1 Software The largest portion of the research was devoted to programming the vision software Two applications were developed to support the project goals The first and largest application was the vision capture and analysis component This was written using National Instruments Vision Builder for Automated Inspection VBAI NI VBAI Code The VBAI code development took a considerable amount of time and effort with a lot of trial and error to determine an inspection sequence that would work in almost all circumstances The process was broken up into small tasks Setup Acquisition Calibration Inspection Display and Results Setup consisted of starting the TCP Socket Server if it was not already running Secondly this step initializes local
38. thing data Nothing Get a stream object for reading and writing Dim stream As NetworkStream client GetStream 47 Dim i 11 32 0 Loop to receive all the data sent by the client Use try Catch to handle forced disconnect from client First read is done outside while loop Additional stream Read are done inside while loop Try i stream Read bytes 0 bytes Length Catch IOException Console WriteLine Socket Forcibly disconnected 0 While i lt gt 0 PROCESS BYTES Translate data bytes to ASCII string data System Text Encoding ASCII GetString bytes 0 i Console WriteLine Received bytes 0 1 data Length ToString data ToString Process Data thru Function parse Data parse Data handles all checks on what type of response to give data parse Data data Dim msg System Text Encoding ASCII GetBytes data Send back a response stream Write msg 0 msg Length Console WriteLine Sent 0 bytes 1 data Length ToString ToString Loop to receive all the data sent the client Use try Catch to handle forced disconnect from client Try stream Read bytes 0 bytes Length Catch serr s IOException Console WriteLine Socket Forcibly disconnected 0 End Try End While PROCESS BYTES Close IO stream garbage collenct and end connection stream Close stream Dispose client Close End While LISTEN
39. unted for simultaneously by using the built in calibration tool available in VBAI 4 2 1 Calibration of vision system The built in routines from VBAI were used to calibrate the vision system in order to account for angular errors in the alignment of the vision system as well as to provide real world units capability The built in nonlinear calibration was chosen to provide a high degree of robustness in the calibration According the National Instruments documentation on nonlinear calibration The nonlinear algorithm computes pixel to real world mappings in a rectangular region centered around each dot in the calibration grid and then estimates the mapping information around each dot based on its neighboring dots 28 To most effectively calibrate the camera to the machine tool the optics assembly and the light source assembly were aligned such that the centerlines were parallel to each other to reduce any errors from the light source not being parallel with the optics Secondly the optics system was installed on the T bed lathe and then aligned to the machine in X and Z directions Finally the built in calibration routine within VBAI was used to calibrate the system to account for any remaining squareness or angular errors 4 2 2 Calibration of Camera Offset By aligning the calibration grid shown in Figure 11 to be square with the machine axes the calibration algorithm 15 also able to eliminate error from the camera not being square to
40. us of this research was two fold 1 Develop a machine vision system capable of resolving to 5 um for the tool nose radius TNR X and Z offsets 2 Develop interface software to allow the machine vision software to automatically transmit tool nose radius and offset values to a machine tool control CNC 2 1 Vision Hardware There are three main components that make up the hardware necessary for a vision system Camera Optics and Lighting Relevant theory and the basis for choosing the current hardware are explained here 2 2 Camera The Basler Pilot pia2400 12gm is a monochrome interline CCD camera with a 5 megapixel resolution 2458x2040 It uses gigabit Ethernet as its digital interface with the PC It is capable of 12 fps at full resolution 1 It was selected because of its sustained frame rate its sensitivity to light and for the digital interface which facilitated the long cable distance between the camera and the CNC control computer Camera Theory There are two main technologies behind current imaging electronics These are Charge Coupled Device CCD and Complimentary Metal Oxide Semiconductor CMOS 2 The main differences in performance between CCD and CMOS sensors according to Basler are shown in Table 1 Parameter CCD CMOS Fill Factor HIGH MED Dark Noise LOW MED S N Ratio HIGH MED Dynamic Range MED HIGH Table 1 Adapted from 2 For the purposes of gauging CCD pre
41. variables to their appropriate values 21 Acquisition requested an image from camera second step was included that evaluates the intensity of the image and based on this data could adjust the image until the intensity was within some target value In this way the intensity adjustment step would loop back to the Acquire step iteratively Calibration simply loaded the calibration data from an image and applied it to the most recent image Inspection is the largest step it contains all the steps necessary to determine the radius and offset values from the calibrated image First it finds 4 points around the perimeter of the image generates lines bisectors and the intersecting point for those lines as shown in Step 1 of Figure 10 32 gt Figure 10 VBAI Steps In Step 2 it rough calculates the tool nose radius uses the bisector and the point at the end of the tool to find the edge of the radius uses this information to create a new region of interest This consists of setting a tolerance band for the radius 22 measurement setting angular ranges for circle fit algorithm shown in Step 3 of Figure 10 Finally it performs the final tool nose radius fit as shown in Step 4 of Figure 10 Further the Acquisition Calibration and Inspection steps are repeated a set number of times configured by a local variable on startup this was set to 10 for the purpose of this study The Display step averages the in
Download Pdf Manuals
Related Search