1、1Enhancing moulds manufacturing by means of reverse engineeringAbstract :Modern CAD/CAM techniques together with five-axis high-speed milling allow to reduce moulds manufacturing time and costs.Nevertheless, in order to put a mould into use,operations of manual finishing and fitting are still always
2、 required. Such operations, performed manually by mould makers, modify the milled surfaces of moulds. Reverse engineering techniques can be employed in quality control to evaluate dimensions and geometrical tolerances on moulds after machining and fitting. Changes in the products shape are sometimes
3、 decided after a mould has already been machined. In such cases, if possible, the mould maker directly modifies the mould. Thus the final real geometry of the mould does not reflect the one of the original CAD model. The aim of this paper is to point out the benefits of non-contact qualitycontrol an
4、d to illustrate a procedure, based on reverse engineering techniques, to reconstruct and update the mathematical model of the mould after it has been polished and fit. The procedure was tested on a mould for the production of a plastic camera body that was previously inspected by means of a structur
5、ed light scannerKeywords :CAD model . Reverse engineering . Mould . Surfaces reconstruction . Computer aided inspection1 IntroductionProduction tools (better known as moulds) are a essential element in the manufacturing of wide consumption goods.The design and the fabrication of moulds are a very lo
6、ng and expensive step in the developmen of a new product.For such reason, any effort improving mouldsmanufacturing helps to reduce times and costs.Nowadays, modern CAD/CAM technologies an fiveaxis high-speed milling 1 allow:& Generation of the 3D CAD model of the mould.This paper aims to show the ad
7、vantages of contactless inspection compared to point-wise measurements. After operations of polishing and fitting, the updating of the 3D CAD model of a mould should also become a good practice among mould makers. A procedure to fulfil that task was defined using common reverse engineering technique
8、s and software packages. The proposed procedure was tested and validated on a mould for injection moulding of plastic camera body.2 Inserts digitising and inspection (CAD model comparison)The inserts of the mould are made of light alloy (Fig. 1).After rough milling on an NC machine, the female figur
9、e was finished by EDM of the tiny grooves located on the camera handle. Because of their size, in fact, those grooves cannot be machined by milling. The left half of the male mould cavity was not treated after the finishing milling operation so I still presents evidence of the 45 crossing mill path.
10、 The other half of the cavity was instead polished after machining (Fig. 2). Such manual operation was introduced to evaluate if the noncontact scanner is able to discriminate the different finish of the two halves properly. In order to point 6out the advantages of non-contact inspection, mould inse
11、rts were digitised using the optical scanner ATOS Standard produced by GOM GmbH (Fig. 3). Contactless scanning of large moulds or products is less time consuming than contact scanning and undercuts can be easily digitised by changing the relative position of the optical sensor to the part. Moreover,
12、 current optical Fig. 1 Photo of the mould inserts (overall dimensions: 2009835 mm) 552 Int J Adv Manuf Technol (2009 43:551562 scanners accuracy and resolution allow a good definition of part edges as a great number of points are measured on them. The selected scanning device exploits stereoscopic
13、vision as it has two Sony XC75 built-in CCD cameras (resolution of 768 572 pixel/8 bit), which store images of the light fringes projected on the scanned object. The projector, placed in the centre of the sensor, projects a sequence of ten slides: four interference patterns (phaseshift technique 9)
14、followed by six Gray coded 10, 11 binary images. The six-bit code allows distinguishing between 26 = 64 columns in the field of view. Considering the overall size of mould inserts, the scanner was calibrated on a working area of 160 200 mm at a distance of 600 mm. The accuracy of the scanner in such
15、 configuration is 0.06 mm as declared on the device datasheet. Before digitising, a thin layer of white opaque powder was sprayed on the inserts in order to avoid light reflection problems typical of metallic surfaces. Furthermore, reference points are applied on the inserts sticking adhesive target
16、s (i.e. markers) on areas that contain less detail. Multiple scans are automatically registered in one point cloud, as the scanning software recognises the fixed reference grid created by markers. There is a lack of data in the areas covered by markers, but GOMs software allows to complete the virtu
17、al model of the object by closing such holes. The registration errors shown in Table 1 were computed by GOM software during the alignment procedure based on the reference grid of markers.Rapidform software by Inus Technology Inc. was employed for the mould inspection. The result of the comparison be
18、tween the triangulated point cloud model and the original CAD one were computed by the software and are shown in Table 1 in terms of average distance, standard deviation and maximum distance.2.1 Female mould inspectionThe digitising of the female mould cavity has required twelve scans. The compariso
19、n 7between the digitised data and the original CAD model shows a maximum deviation of 0.84 mm on the left superior pocket. Such value was verified by point-wise measurements on a CMM: measuring the depth of that pocket, the real deviation resulted of 0.85 mm.In the superior half of the female cavity
20、, the deviation is about 0.30 mm, while the error is positive (about 0.20 mm) in the other half. The deviation map with relative errorsFig. 2 Different finish of the male mould surfaceFig. 3 3D scanner ATOS Standard and female mouldwith reference point markers Intscanners accuracy and resolution all
21、ow a good definition of part edges as a great number of points are measured on them.The selected scan ning device exploits stereoscopic vision as it has two Sony XC75 built-in CCD cameras (resolution of 768 572 pixel/8 bit), which store images of the light fringes projected on the scanned object. Th
22、e projector, placed in the centre of the sensor, projects a sequence of ten slides: four interference patterns (phaseshift technique 9) followed by six Gray coded 10, 11 binary images. The six-bit code allows distinguishing between 26 = 64 columns in the field of view. Considering the overall size o
23、f mould inserts, the scanner was calibrated on a working area of 160 200 mm at a distance of 600 mm. The accuracy of the scanner in such configuration is 0.06 mm as declared on the device datasheet.Before digitising, a thin layer of white opaque powder was sprayed on the inserts in order to avoid li
24、ght reflection problems typical of metallic surfaces. Furthermore, reference points are applied on the inserts sticking adhesive targets (i.e. markers) on areas that contain less detail.Multiple scans are automatically registered in one point cloud, as the scanning software recognises the fixed refe
25、rence grid created by markers. There is a lack of data in the areas covered by markers, but GOMs software allows to complete the virtual model of the object by closing such holes. The registration errors shownin Table 1 were computed by GOM software during the alignment procedure based on the refere
26、nce grid of markers.Rapidform software by Inus Technology Inc. was employed for the mould inspection. 8The result of the comparison between the triangulated point cloud model and the original CAD one were computed by the software and are shown in Table 1 in terms of average distance, standard deviat
27、ion and maximum distance.(Fig. 4) allows to clearly notice the relief of the inferior half of the cavity, whereas the mould parting plane is almost aligned correctly with the CAD data.The deviation different sign of the two halvessuggests that the EDM operation was not done correctly. The yellow lin
28、e in the middle of the deviation map can be thought as an imaginary axis around which the cavity surface may be rotated (in the direction shown by the arrows) to coincide again with the original CAD model.Probably during EDM, the electrode motion was not perfectly perpendicular to the parting plane
29、of the female mould. The little inclination to the normal might be due to an incorrect mounting of the mould on the EDM machine.For this reason, the electrode machined the piece eroding too much material in the superior half of the female insert.The inclination is calculated considering the normal o
30、f the parting plane on the digitised data and on the CAD model: the angle between them measures 0.63.2.2 Male mould inspectionNine scans were necessary to completely digitise the male mould insert. In the comparison between the scan data and the original CAD model, the maximum error is localised on
31、the superior right prominence of the cavity (Fig. 5).In that zone, there is a graze caused by the fall of the milling tool during a tool change.In the CAD data, the bottom boundaries of the handle tiny grooves have sharp edges. The male mould insert was machined by a ball-nose finish cutter having a
32、 diameter of 4 mm and the CNC mill path followed crossed 45 diagonal trajectories in parallel planes. Such finishing operation does not allow to create sharp edges on the grooves bottom.Hence, the scan data show a great deviation on the grooves of the handle. Figure 2 allows to observe the graze and
33、 the marks left by the 45 diagonal trajectories of the milling tool.Moreover, the deviation map of the male mould insert exhibits a difference between the left half and the right one. 9That is the evidence of the manual finish of the right half.The diversity is exalted by limiting the representation
34、 scale up to an error of 0.20 mm(Fig. 5). The polishing operation has removed a very thin layer of material, resulting in a deviation of less than 0.10 mm. In order to confirm such result obtained by contactless inspection of the whole male insert, the difference between the polished half of the cav
35、ity and the other one was checked by means of a CMM. Two zones were selected on the superior surface of the male cavity, one on each half.The first one, labelled A, refers to the half that was not manually finished, so it lays on the left part of male insert.The other one, labelled B, refers to the
36、polished half (Fig. 6). Traditional point-wise measurements were repeated on the two zones by means of Renishaw TP-20 probe mounted on a coordinate measuring machine DEA Global Image model 07.07.07. After the alignment with the CAD model coordinate system, 83 points were inspected on zone A and 106
37、points on zone B. CMM inspections result are reported in the second column of Table 2 and 3 in terms of average distance, standard deviation and maximum distance between measured points and the corresponding ones on the mould CAD model.To further compare CMM inspection results with the ones of non-c
38、ontact scanning, two limited sets of points, corresponding to zones A and B, were isolated from the whole scan data of the male insert. The results of the comparison of those two sets with the matching points of mould CAD model are shown in the last column of Tables 2 and 3. As a matter of fact, suc
39、h results are very similar to those obtained by traditional quality control, but are referred to a higher point density.This means and allows to state that the optical scanning device employed can substitute the CMM, exploiting the advantages of contactless inspection mentioned in the introduction.
40、The only requirement is that the scanner accuracy must be consistent with the dimensional tolerances of the inspected mould. Since CMM results were almost equal to those obtained by optical scanning, it is also important to notice that the layer of opaque powder sprayed on the insert is influence fr
41、ee on inspection results because its thickness is inferior to the scanner accuracy. Moreover, the registration error computed by GOM software (Table 2) and the effect of point cloud 10processing are not additive to scanner accuracy but are rather included in the value declared by the constructor.3 P
42、rocedure for the reconstruction of the mould CAD modelIn the reconstruction of the mould CAD model, geometric tolerances (symmetry, concentricity,parallelism, orthogonality, etc.) have to be considered as they represent an essential information for the piece fabrication, assembly and correct working
43、. During the design phase, as well as applying reverse engineering, the function of the piece is very important and it cannot be disregarded.For example, the angle between two planes may measure about 90 in the point cloud data of a scanned mould for injection moulding. Without any consideration of
44、the technological kind, the displacement to 90 might be interpreted as a measurement inaccuracy and one could decide to draw two perpendicular planes in the final recreated CAD model. Small draft angles are instead necessary for the correct extraction of the manufactured piece and the creation of tw
45、o perpendicular planes could be wrong. For such reason, the process of reconstruction depends very much on the operators experience. For the creation of a correct CAD model starting from scan data, standard or automated procedures do not exist. The most important step is the segmentation process, wh
46、ich is the subdivision of scan data into surface areas or zones. Shape features support product redesign and advanced manufacturing technologies. In literature several methods have beenFig. 5 Deviation map of themale mould insert 0.200.180.160.140.120.100.080.060.040.02110.00POLISHED HALF(mm)ZONEAZO
47、NEBFig. 6 Zones A and B selected for the point-wise measurements by CMMTable 2 Comparisons of measures on the milled half (Zone A) of male mould cavityInspection deviation analysis result CMM Optical scanner Number of points measured 83 420Average distance () 0.01 mm 0.01 mmStandard deviation () 0.0
48、1 mm 0.01 mmMaximum distance 0.04 mm 0.05 mmInt J Adv Manuf Technol (2009) 43:551562 555investigated to enhance the segmentation process by means of automatic or semi-automatic edge detection and shape feature recognition. Robust results were obtained for classic geometries and some kind of mechanic
49、al parts, but not for freeform shapes and features. Automatic procedures of specialised software packages are not yet adequate to give satisfactory results without user intervention: they generate a collection of smoothly connected untrimmed patches globally reflecting the topological structure of the shape, but surface borders will seldom run along feature lines, sharp edges, fillet profiles, etc. The so-called arbitrary topology surfaces 12 do not subdivide the part in a sensible manner to redesign its shape, since the basis of this reconstru
