use of digital checking fixtures and scanning techniques for reverse ...

Session 3BUse of Digital Checking Fixtures and ScanningTechniques for Reverse Engineering PurposesSean DerrickWestern Michigan University, s4derric@wmich.eduDr. Mitchel KeilWestern Michigan University, mitchel.keil@wmich.eduAbstract - This case study will examine and teachmethods for reverse engineering components usingincomplete digital scan data, CAD models and a digitalchecking fixture technique. In this example acommercially available truck leaf spring had itsgeometry captured in both a deflected and non deflectedstate. However due to time and scanning constraints, thedeflected spring was only partially captured makingdirect Reverse Engineering (RE) impossible. Once thenon deflected spring was modeled the partial scan datawas used to generate a digital checking fixture. Usingthis fixture the non deflected model was altered togenerate a very accurate representation of the deflectedspring.Index Terms – Center for Advanced Vehicle Design andSimulation (CAViDS), Computer Aided Design (CAD),Geometric Dimensioning and Tolerance (GD&T), ReverseEngineering (R.E).INTRODUCTIONOptical and digital scanners have made Reverse Engineering(R.E.) relatively easy compared to conventional manualmeasuring techniques. Structured light or lasers are used tomap the geometry of a component and that geometry can bedigitally recreated using CAD systems. These systems allowfor measurement accuracy to be within the thousands, if nottens of thousands, of the inch. However with any newtechnology there can be draw backs or imperfections. Thedraw back with scanners is that if the data is incompletethen RE becomes extremely difficult if not impossiblewithout the original component present. This situation couldarise if time constraints prevent all of the geometry frombeing captured by the scanner or if line of sight access isn’tpossible. Problems also arise if data is lost or corruptedduring file transfers offThis paper will present a technique to help reverseengineer a component with partial scan data. Specifically,this paper will show how to reverse engineer a deflectedcomponent using partial data and a non-deflected model.For teaching purposes this paper will present this techniqueas a practical case study which was used in an actualresearch project.METHODOLOGYThe principles of this technique are already firmly groundedand used in the non digital realm. The principles are used inGeometric Dimensioning and Tolerance (GD&T) as well asdimensional metrology. In most practical cases, todetermine the dimensions of a deflected part, a fixture iscreated using known dimensions. Then a non deflectedcomponent is placed into the fixture and deformed to meetthe desired dimensions. Once deflected, the part can beanalyzed and used for various purposes. This processusually involves time intensive operations to generate anactual fixture and also requires the deformed part asreverence. Furthermore, to generate extremely accurateresults precision machining and joining operations arerequired to reproduce a deflection of a thousandths of aninch. Significant cost will go into creating a fixture accurateenough to replicate the circumstances.A similar methodology is being implemented in digitalCAD media. Virtual CAD models are paired with partialscan data to generate a checking fixture in the computer.Then a free state CAD component is modified andmanipulated to meet the fixture, thereby accuratelyrepresenting the deflected component. This technique isfaster and cheaper than that of the conventional means andcan be conducted numerous times.For this digital technique primary and secondarydatums are located in the object that is to be reproduced.Following this step, CAD primitive shapes are made torepresent these datums. Once the primitives are created theyare then assembled in a form of a digital checking fixture.The fixture, to be presented in the coming example, issimply a CAD assembly produced in Solidworks.[1] Thescan data to be duplicated is overlaid into the fixture forverification to act as both a tertiary datum as well as ameasurement reference. The scan data is also used to checkdeviation after modification. A completed CAD model isthen inserted into the fixture and modified till it matches thescan geometry. Along with being much faster and cheaperthan actually fabricating a real component fixture, there isone additional bonus. Scanner technology capturesextremely tight tolerances and surface deviations thereforemaking this methods final product extremely precisecompared to manual and conventional means.Pittsburgh, PA March 26 - 27, 2010ASEE North Central Sectional Conference3B-1

The following steps are used in order to perform thefixture technique: Generate digital X,Y and Z sections of the object Locate critical datum features Generate parametric primary datum Generate parametric secondary datum Place datums in CAD assembly Overlay digital sections on assembly Place part to be deformed in fixture Deform component to match datums Verify final product for accuracy using scan dataSession 3Bincomplete or obstructed from the scanners view bysurrounding components of the truck. The lack ofinformation made direct duplication of the deflected springsdirectly from the scan data impossible. There was no longeraccess to the truck itself to allow for rescanning of thedesired geometry. Therefore, a method was developed toreproduce the deflected spring using its basic geometry andthen modified to meet known dimensions of its deflectedcounter parts. A digital picture taken of the non deflectedspring can be seen below in Figure 1 along with its reverseengineered CAD model.CASE STUDYIn this case study the overall project involved the reverseengineering of a production model Peterbuilt truck. Thetruck’s geometry was captured using digital white lightscanning technology and then reproduced into a CADmodel. The model would then be used in a dynamicsimulation program called ADAMS-car. Once in ADAMS-Car roll over analysis of the truck could be conducted andthen analyzed. The resulting data would then be use by thePeterbuilt Company, which produces the truck.The research was conducted at Western MichiganUniversities Center for Advanced Vehicle Design andSimulation (CAViDS) laboratory. The project was tasked toreverse engineer components, verses simply retrieving priorCAD files from the company, in order to factor inmanufacturing differences and variances which the idealCAD files neglect or cannot account for. Those variancesmade our simulations as accurate as possible to the realmanufactured trucks. [2]Towards the end of the project, several componentsneeded to be clarified or re-created prior to simulation. Oneof the components which needed to be worked on was thetruck’s leaf spring assembly. Three scans of the springassembly were taken for the project. From these scans threemodels would be generated. One model was of the overallspring assembly in a non-deflected state, resting on theground, while fully assembled. The second model was of thespring assembly in place on the truck. This second assemblywould be deflected but only under assembly conditions. Thetruck was lifted into the air so that no other weight wasplaced onto the spring. The third and final model would beof the spring in its deflected resting state, with the truck onthe ground and the weight of an unloaded truck bearingdown upon it. The three models would be used to helpcalculate and verify the computer generated spring’s motionand deflection as well as to create the initial spring CADmodel.The challenge of creating the computer models wouldbe in the lack of provided information. Scans of the leafspring assembly were conducted months prior to its CADduplication and before the need for multiple models wouldbe required or even requested. Complete scans of the leafspring assembly, in its not deflected state, were conducted.However scans of the mounted spring were eitherFigure 1Non Deflected Spring (top) RE CAD Model (bottom)Though the scans did not allow for direct duplication,they did however capture key areas where the spring was tobe mounted relative to other components in its deflectedstate. Also some geometry of the spring was captured thatallowed for referencing based upon mounting points. Inother words the scanner revealed datum points and key crosssections of the spring. Enough key points existed that amodel of the non deflected spring could be modified andaligned with these points to reproduce its deflectedcounterpart.To modify the non deflected model, a form of digitalchecking fixture was created using the datum pointscaptured in the scan. Like a real life measuring fixture, thedatum points would hold the work piece in place while themain cross sections of the spring were bent and manipulatedto match the known cross-sections which were captured inthe partial scan. The non deformed spring would bemanipulated till it would align with the known sections ofthe deformed spring. Once finalized the deformed springwas then re-inserted into the known scanned geometry to beverified and to calculate percentage error. This techniqueallows for accurate reproduction because the softwarechanges the model according to how the metal woulddeform under loading conditions.Pittsburgh, PA March 26 - 27, 2010ASEE North Central Sectional Conference3B-2

Technique StepsFor this project an ATOS II white light scanner was used toscan part geometry. [3] Once scanned, the surface geometryof a part is stored inside the attached software. The ATOSsoftware allows a user to divide up the scanned surface intosections. Too accurately model components in this way,sections along the X, Y and Z axis were made at fivemillimeter increments. An example of the spring sectionsalong the X-axis can be seen in Figure 2. In this figure thesections are overlaid on the part. When these sections areoverlaid on top of one another the entire part can be seen inthree dimensions. This is the starting point to this reversetechnique.Figure 2Scan Sections Overlaid on ModelThe second step in creating the digital fixture was tolocate and identify the main datum reference points, then tocreate a three dimensional solid primitive to act in its place.In the case of the leaf spring, these points were the two endsof the spring and the top assembly mounts. All three ofthese primary points are bolt locations and regardless of theamount of deflection, these points will always be bolted inthese locations. An example of the described location canbe seen highlighted in Figure 3.Session 3Bboth a known size and had their geometry captured in allthree scans. An example of the finished first step can beseen in Figure4.Figure 4Primitive Cylinders with Spring for ReverenceThe third step in creating the fixture was to locate andrepresent secondary datums. In the springs case thesecondary datum would be where surfaces of the springwould come in contact with known surfaces of the truck.For example the leaf spring included a mounting platewhich not only reinforced the bolt for the center axle mountbut also came in full contact with the axle itself. Also, twomounting plates, which are perpendicular to their mountingbolts at either end of the spring, were located. These arepoints which, like the bolts from the first step, will neverchange being in contact. However they are not as critical asthe bolt locations because there have larger variances. Thesesurfaces can shift and slide depending on the amount ofdeflection and how they were assembled. For example theends of the spring will bolt to the frame of the vehicle andprevent the spring from translation along the X, Y, and Zaxis. The surface of the vehicle frame, that the spring isbrought up against, prevents translation in only the Z axisand prevents rotation. Because the surface allows for alarger degree of translational movement and is not as criticalto the overall design intent, as the bolts, the surfaces are asecondary reference. After locating these points, they wererepresented in the CAD model by solid surfaces which thespring primitive would be joined against using matecommands. An example of the third step can be seen belowin Figure 5.Figure 3Primary Datum LocationsThese points represented the primary datum points. Theprimitives where made as cylinders, which represented thebolts that affix the spring in reality. All three of the bolts areFigure 5Primary and Secondary Datums with Spring for ReferenceThe final step in creating the fixture was to import thescanned sections of the deflected leaf spring into the fixturemodel. The imported sections were transferred into themodel as an IGES format. The points were then overlaidPittsburgh, PA March 26 - 27, 2010ASEE North Central Sectional Conference3B-3

onto the already created solid primitives from the previoussteps. This newly imported data would serve to act as aguide when transforming the model and to verify thelocations found in the previous steps. These cross sectionsalso would act as a tertiary datum point.Once the fixture was complete and verified using theIGES file, by overlay, the completed assembly of the nondeflectedleaf spring was then added to the fixture assembly.Once oriented, one of the main datum points were alignedand affixed to the fixture model as described in Figure 6.From this point the next set of bolt locations were placed.Using several computerized commands, which allowthe model to be deformed without changing its overalldimensions, the other main datums were then aligned andchecked for interference. Below in Figure 8 shows thespring after it has been formed to meet the primary andsecondary datum points. Once interferences were eliminatedfor the main datum points, the deforming process wasrepeated matching the cross section of the spring to thefixture. The cross section allows for very fine geometricdeflections to be captured.Figure 6Spring Before DeformationSession 3BAfter the desired component is created and verified itcan be compared against its counterpart in a Finite ElementAnalysis (FEA) program in order to determine stress andstrain characteristics. In the case of the leaf spring both thenon-deflected and deflected models were overlapped in FEAand combined with the known loading of the truck. Fromthis information a stress strain curve was generated andmapped along with a stress distribution along the geometry.Once the stresses and strains are known they can be used toin an extremely accurate representation of how the springwill behave can be made in the dynamic simulation.CONCLUSIONIn the case of the spring the overall process of taking theCAD model and creating the fixture took approximately onehour. An additional hour was needed to both modify thescan sections so that they could be overlaid and tomodifying the part for deflection. From start to finish thisre-engineering technique took about two hours time makingthis technique much faster than making doing it manually.Above all this technique allowed the project to be completedwithout the actual spring present and was far more accuratein its results then manually checking.This method can be used for any component which hasto be deformed for analysis as long as there is a preexistingCAD file along with deflected scan data. Using thistechnique parts such as beams, springs, cantilevers orirregular geometry solids can be reverse engineered usingminimal data.Figure 6Spring After Deformation (Primary and Secondary Datums)After this process had been concluded a secondinterference check was run to see if the spring model cameinto contact with the imported IGES file. Following a seriesof minor adjustments, the newly deformed model of the leafspring now lined up with the reference geometry. The finalinterference check, with the sections, found that the springwas within less than one thousand of an inch of deviationfrom the IGES overlay. Given that the scanner being usedallowed for greater inaccuracy than the deviation foundbetween the scan data, the model was deemed an acceptablereproduction. This process was repeated to successfullygenerate the second partially deflected spring.With the deflected spring complete the CAD model wasbrought into the ATOS software for a final verification. Theraw scan data holds many more data points for comparisonthan the sections making the verification in ATOS evenmore reliable than relying on the sections alone. Thesections were used due to data transfer and ease ofpreliminary use however a lot of data is lost in thesectioning process. Therefore it is highly recommended thatthe actual scanned surfaces be used for final modelverification. If an extreme degree of accuracy is not neededfor the component then this step can be neglected.REFERENCES[1] Lombard, M, SolidWorks Surfacing and Complex Shape ModelingBible, Indianapolis, Indiana, Wiely Publishing Inc, 2008.[2] Ingle, K, Reverse Engineering, Washington, DC, McGraw Hill, inc.1994.[3] Optical Measuring Techniques. 2006. ATOS User Manua v6. OrderNumber: D-38106 Braunschweig, Germany.AUTHOR INFORMATIONSean Derrick Masters Student, Western MichiganUniversity, College of Engineering and Applied Science,Kalamazoo, MI, 49007, s4derric@wmich.eduSean studied engineering design and graduated fromWestern Michigan University in Spring 2009. He iscurrently a graduate student studying manufacturingengineering at Western Michigan University specializing inreverse engineering, design for manufacturability andconceptual machine design. After his masters degree, heplans to begin a career in military research anddevelopment.Pittsburgh, PA March 26 - 27, 2010ASEE North Central Sectional Conference3B-4