WO1994014029A1 - Measuring system suitable for strain analysis of three-dimensional components - Google Patents

Measuring system suitable for strain analysis of three-dimensional components Download PDF

Info

Publication number
WO1994014029A1
WO1994014029A1 PCT/BE1993/000075 BE9300075W WO9414029A1 WO 1994014029 A1 WO1994014029 A1 WO 1994014029A1 BE 9300075 W BE9300075 W BE 9300075W WO 9414029 A1 WO9414029 A1 WO 9414029A1
Authority
WO
WIPO (PCT)
Prior art keywords
points
grid
camera
images
measuring
Prior art date
Application number
PCT/BE1993/000075
Other languages
French (fr)
Inventor
Michel Vermeulen
Herman Bruneel
Original Assignee
Lk Limited
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Lk Limited filed Critical Lk Limited
Publication of WO1994014029A1 publication Critical patent/WO1994014029A1/en

Links

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01B—MEASURING LENGTH, THICKNESS OR SIMILAR LINEAR DIMENSIONS; MEASURING ANGLES; MEASURING AREAS; MEASURING IRREGULARITIES OF SURFACES OR CONTOURS
    • G01B11/00—Measuring arrangements characterised by the use of optical techniques
    • G01B11/16—Measuring arrangements characterised by the use of optical techniques for measuring the deformation in a solid, e.g. optical strain gauge
    • G01B11/165—Measuring arrangements characterised by the use of optical techniques for measuring the deformation in a solid, e.g. optical strain gauge by means of a grating deformed by the object
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01B—MEASURING LENGTH, THICKNESS OR SIMILAR LINEAR DIMENSIONS; MEASURING ANGLES; MEASURING AREAS; MEASURING IRREGULARITIES OF SURFACES OR CONTOURS
    • G01B11/00—Measuring arrangements characterised by the use of optical techniques
    • G01B11/002—Measuring arrangements characterised by the use of optical techniques for measuring two or more coordinates
    • G01B11/005—Measuring arrangements characterised by the use of optical techniques for measuring two or more coordinates coordinate measuring machines
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01B—MEASURING LENGTH, THICKNESS OR SIMILAR LINEAR DIMENSIONS; MEASURING ANGLES; MEASURING AREAS; MEASURING IRREGULARITIES OF SURFACES OR CONTOURS
    • G01B5/00—Measuring arrangements characterised by the use of mechanical techniques
    • G01B5/004—Measuring arrangements characterised by the use of mechanical techniques for measuring coordinates of points
    • G01B5/008—Measuring arrangements characterised by the use of mechanical techniques for measuring coordinates of points using coordinate measuring machines

Definitions

  • the present invention relates to a measuring system suitable for strain analysis of three-dimensional components to which a pattern has been applied, consisting of a camera mounted on a coordinate measurement machine (CMM) which has a built-in computer linked to an image processing system.
  • CCM coordinate measurement machine
  • the purpose of the invention is to measure displacements and strains in the surface of a mechanical component after deformation in a non-machining mechanical or hydraulic process, such as deep drawing, drop forging or forging of steel and/or aluminiur..
  • the component to be examined is provided with a pattern of regular, well known geometric form and size before deformation.
  • the grid is transferred to the component in the familiar electrochemical, notochemical, printing or photographic way.
  • the component, and more particularly a blank, is subsequently deformed in the process, which also deforms the grid.
  • the strains are identified by measuring the deformed grid.
  • Circle Grid Analysis the circle grid is transformed to a set of ellipses. The extent of the strains can be derived from the length of the minor and major axes of each ellipse. Then the strains are displayed in a strain diagram as principal main strains.
  • Square Grid Analysis the displacement of the points of intersection of the grid's mutually perpendicular lines is measured.
  • the strain analysis is also meant to optimize the interaction between plate, tool and lubricants. This allows the die designer to check the safety margin of a certain point on the deep drawn part against the Form Limit Diagram. This method also gives an idea of the distribution of the strain in a specific zone, the direction in which the material flows and the entire strain at a specific point.
  • a method to measure the distribution of strain on a deformed surface is described in document USA-4969106 (J.H. VOGEL) .
  • the method uses a CCD-camera and a vision system to process two images received through stereo ⁇ scopic recording. It consists of the following steps :
  • - processing in case of automatic image treatment basically consists of two successive filtrations (average noise reduction and high pass convolution with an elementary cell 7 x 7) ;
  • the two sets of two-dimensional coordinates are merged into a set of three-dimensional coordinates (Combining) .
  • At least one corresponding point in the two sets should be indicated by the operator MANUALLY;
  • a first disadvantage of this method is that it requires frequent manual intervention by the operator during the measuring process, as the image processing algorithms are very rudimental, for example for :
  • a second disadvantage is that the accuracy of the above described stereoscopic system decreases when the size of the component increases.
  • the main size of the component can be divided into a maximum of 512 or 1024 parts only. If an acceptable level of accuracy is to be reached, the size of the component has to be limited. There is no method or instrument that can subdivide the component automatically into smaller zones for which an acceptable degree of accuracy was possible.
  • the present invention aims at solving these problems. It is a measuring system especially suited for strain analysis of three-dimensional patterned components, consisting of a camera built on a coordinate measurement machine (CMM) which has a built-in computer linked to an image processing system. Typical of this measuring system is that the CCD-camera has at least five degrees of freedom and that the computer is in charge of the entire measuring and calculation process, i.e. :
  • the method has the advantage that, for economic and technical reasons, it complies with the increasingly stringent requirements to the material itself and to the corresponding forming processes. It allows for accurate and reliable methods of analysis possible. It is not impaired by the complexity of components, generating a large number of data. The methods of analysis applied, quickly and automatically yield results that can be used in practice with minimal operator intervention (ergonomic objective) .
  • the measuring system has a measuring head which can comprise both a touch probe and a camera, which are mutually and automatically interchangeable.
  • a measuring head which can comprise both a touch probe and a camera, which are mutually and automatically interchangeable.
  • An innovation is that the touch probe is controlled by appropriated scanning hardware and software during the process of measuring the shape of complex components.
  • a control hardware and software move and position the camera around the component, avoiding contact between camera and component, both in the positions from which the images are taken and during intermediate movements.
  • a particular design of the coordinate measurement machine is of the bridge type, fit to move in three orthogonal directions, with a measuring head that can revolve around a vertical and a horizontal axis respectively, a computer to control the measuring stand meant for scanning the deformed component and with specific procedures linking and controlling the above mentioned image processing system, and for calculating and displaying the results.
  • the material of the component does not impose any restriction on the application of the present invention. It can either be metal, plastic or composite material or any other which can be applied in solid, plate or layer shape.
  • the material should be suitable for the application of any speci ic pattern on the surface before the forming process, after which the (also deformed) pattern must still be recognizable. The nature of the recognizability is unimportant.
  • the presently introduced machine is based on visual recognition using a pattern. Any way of recognition, such as infrared, magnetic, ultrasonic, ultraviolet etc., with which the same objective can be reached following the same procedure, also belongs to the scope of the present invention.
  • the expressions "vision” and “image” cover the general concepts of "recognition” and “result of recognition”. Other characteristics of the invention will be dealt with in the following description of procedures and equipment of the invention.
  • the enclosed drawings, relating to the description, are given as examples only. - figure 1 shows a perspective drawing of a measuring system along the lines of the invention;
  • FIG. 2 shows a large scale picture of a touch probe during the process of measuring a component
  • FIG. 4 schematically shows the camera scanning a component
  • - figure 6 is an image after processing
  • - figure 7 is an image with a crack after processing
  • FIG. 9 is a diagram on which the strains are represented by order of size and direction;
  • the system relating to the invention comprises the following components, building together one integrated system to perform a number of consecutive subtasks :
  • CMM three-dimensional coordinate measuring machine 1
  • a three-dimensional coordinate measuring machine 1 e.g., of the HA80 type of LK Ltd. ® (England) , with at least five degrees of freedom and an appropriate touch probe 2 (of the TP2 type of Renishaw ®, England) to measure the global shape of the component 3, consisting for example of a heavy granite table 4 as basic element, with on it a bridge moving in longitudinal direction (X) .
  • X longitudinal direction
  • a unit with inside a vertically movable (Z-direction) quill 6 with the probe head 9 at the free end moves in Y-direction;
  • - computer(s) 7 for controlling the measuring stand and processing all the different data in the different stages of the measuring process;
  • a camera 8 with appropriate lenses, automatically interchangeable with the probe 2 of the measuring machine;
  • the probe head 9 is of type PH10M of Renishaw ®. It offers two additional rotational possibilities about a vertical and horizontal axis respectively. The probes can be placed under different angles, so that almost every point of the component to be measured becomes accessible.
  • a non-contacting laser probe e.g., of type OP5 of Renishaw ®
  • the measuring stand moves over very precisely defined paths and air bearings and is driven by servo-controlled electric motors. The position in each direction of motion is measured with high precision digital measuring rulers (Heidenhain ®, Germany) .
  • An automatic switch system makes it possible to choose between the different probe configurations from the measuring programme.
  • the machine includes a monochrome CCD-camera 8 focused on that spot of the component which is to be analyzed.
  • Fig. 2 shows a touch probe 2 during measurement. Laser probes can be used too. If the range of the touch probe 2 should require so, the operator has the possibility to divide the component 3 in different zones (ZONING) .
  • the measurement yields the 3D-coordinates of a large number of points of the registered surface, but not those of the points of intersection of the pattern lines (because they are not perceptible for the probe) .
  • the software package for the different measuring tasks consists of different levels :
  • CMES Continuous Measuring System
  • prismatic components consisting of both teaching and off-line programming. All state-of-the-art devices are present, e.g., automatic component alignment;
  • This software is a special application for scanning randomly curved surfaces, such as those used in car manufacturing (parts of the body) .
  • This module makes it possible to develop subroutines accessible from CMES. This can be done on the basis of functions that are not accessible at level 1, such as extensive communication with peripheral machines and/or the internal computer structure of the MicroVAX ®.
  • An automatic procedure divides the component into triangular zones, the dimensions of which are chosen so that they can later be taken in one general view by the camera 8.
  • the direction perpendicular to the surface is calculated for each triangle, as is the future position of the camera 8. 4.
  • Images of the deformed surface (CAMSCAN : CAMera SCANning)
  • Fig. 4 schematically shows the camera 8 scanning a component 3. In reality, the camera is much smaller : length ⁇ 150 mm, diameter ⁇ 30 mm.
  • Different lenses can be mounted on the camera (manually) or several automatically interchangeable cameras can be used, each with its own lens. Furthermore, it is possible to use zoom lenses, which can be controlled automatically.
  • Tests have already been carried out with lenses with a range of 25, 16 and 10 mm 2 respectively.
  • the lenses have an excellent depth of field, so that the curved areas of the component 3 very sharply visible.
  • the images are transmitted to the image processing system.
  • the number of images recorded depends on the size of the component 3 to be examined. This number is basically unlimited for off-line processing of images. For example, if an optical disc is used for image storage there is actually only a theoretical limit. One single image of 512*512 pixels with 8 grey shades requires 262 Kb. Therefore, an optical disc of 900 Mb can contain approximately 3400 images, which means that in reality there are no strict limits, not even with on-line processing.
  • Each of the recorded images is dealt with separately in an image processing programme in order to calculate the characteristics of the elementary cells.
  • the pattern is a square meshing, the points of intersection of the line pattern need to be identified. In case of a circular pattern the centres of the ellipses can be treated in the same way. To be able to determine all the points of intersection automatically, without operator assistance, very powerful algorithms have been elaborated, which are also able to process "difficult" images of "poor quality". If the component is a steel plate, the contrast is very poor due to the considerable and irregular reflections resulting from the complicated component shape and its roughness.
  • the mean background of the image is determined and subtracted from the original image. This results in an image with a regular mean background. This image is made binary.
  • the two directions of the grid lines are determined by means of the two largest local maxima of the co-variance, calculated for the different directions in the image.
  • a family of grid lines is extracted by filtering out the other family a ⁇ ci the background noise by means of an OPENING function with a linear, structural element in the direction of the grid lines determined under 5.2.
  • Grid lines that are broken off after this procedure, are patched again by means * of a CLOSING function with a larger linear structural element.
  • OPENING and CLOSING functions are described in the literature on mathematical morphology. See the publication of Jean Serra - Image Analysis and Mathematical Morphology, Volume 1 , Academic Press Inc., San Diego, California (1989) .
  • the filtering in 5.1 and the operations in 5.3 cannot detect local differences in contrast. Consequently, parasitic short line segments are sometimes created. Some of these line segments connect grid lines of the other family and therefore cause parasitic meshes in the deformed pattern 10. Others cause free projections that intersect maximum one grid line of the other family. These parasitic lines are eliminated by successive application of the OR function to both images from 5.3; the SKELETON function applied to the new image; the PRUNE function, which removes the free projection; and the linear OPENING function in both directions, after inversion of the image obtained after pruning, which eliminates the parasitic meshes.
  • the determination is done in two different ways.
  • the points of intersection are determined by taking the logical intersection using the AND function of the two families of line segments in the two directions, followed by the SKELETON function that reduces all objects in the resulting image to one pixel.
  • the resolution will never be higher than one pixel.
  • a typical output file of SIGPAD, which supplies all the necessary data, is printed beneath.
  • the first four lines contain general information.
  • the first line contains the name tbOOl.img of the file housing the image concerned.
  • the second line mentions the image number of the lens used.
  • the third line gives the image number.
  • the fourth line lists the slope of both directions.
  • the next line is "D 1 3", which describes how many segments there are in the first direction, in our case three.
  • D 1 3 describes how many segments there are in the first direction, in our case three.
  • S 1 2 3 For each segment there is a line defining the number of points belonging to that segment, e.g., "S 1 2 3", which means that segment 1 contains two points. Finally, the coordinates of these points are given. The same procedure is repeated for the second direction.
  • TABEL 1 Typical output file of SIGPAD
  • the format used implies that the coordinates of each point occur twice, which allows the reconstruction of the connections between adjoining grid points.
  • the line segments are described mathematically by means of the commonly known regression methods.
  • the points of intersection are the algebraic intersections of the line segments in both directions. In this way, a subpixel resolution is used and a number of non-detected points can be interpolated very accurately.
  • the display of the information is the same as in the first method, but the coordinates are no longer necessarily integers.
  • Fig. 5 shows an example of a "difficult" image before processing.
  • the image shows part of the component 3 in the steel plate.
  • the elevation in the middle is bright because of the reflection; especially at the left hand side the component fades out to the back, making the exposure very weak and the contrast between the lines and background very vague.
  • All the points of intersection can be traced : Fig. 6.
  • Fig. 7 gives an example of an image with a crack after processing.
  • First conversion factors are applied to translate the pixel coordinates into real dimensions (mm) .
  • the conversion factors are obtained through stringent calibration procedures applied to a pattern 10, the dimensions of which are exactly known by measurement with a measuring microscope.
  • a pattern 10 the dimensions of which are exactly known by measurement with a measuring microscope.
  • control and calculation processes record is kept of which (plane) image displays match each camera position. This makes it possible to convert them into absolute coordinates (world coordinates) in a three-dimensional space.
  • Special attention is payed to the exact connection (“knitting") of the separate images, so that the mutual orientation (east-west) , (north-south) of all points of intersection is exactly determined.
  • the objective of the LINKING module is to reconstruct the deformed grid through the aid of the information displayed in the individual images.
  • the parametric description of the form of the measured surface takes place with splines (NURBS) .
  • the image coordinates (u,v) are the pixel coordinates of a point in the image. The origin is the centre of the image.
  • the world coordinates (x,y,z) are the coordinates of a point within the CMM's system of reference.
  • the parameter coordinates (s,t) are the coordinates of a point on the surface, in the spline-parameter space representing this surface.
  • the grid coordinates (i,j) are the discrete coordinates of a grid point on the measuring raster.
  • the SIGPAD module extracts all data concerning the grid from an image via mathematical morphology filterings. For each image the grid point image coordinates (u,v) in that image are given. When the grid is a square one, the two directions of the deformed grid lines are given as well. To demonstrate how these points are mutually linked by these lines, we assume the following. There are four links in each point.
  • the northern link (N-link) is situated along the most vertically oriented grid line in the sense of a mounting v; the eastern link (E-link) is located in the other direction in the sense of the mounting u; the definitions of the southern link (S-link) and western link (W-link) are clear from the above mentioned.
  • the SIGPAD module extracts the line segments in the deformed grid's two directions after several image filtering operations. We are interested in the points of intersection of the line segments in the two directions. Two types of mistakes that are not easily intercepted by the image filterings, but that can be detected and eliminated easily afterwards are : multiple points of intersection and points of intersection as a result of parasitic short line segments which intersect only one segment in the other direction. This is exemplified in 5.5, where there are multiple points of intersection between the line segment "SI" in the first direction and "SI" in the second direction. Multiple points of intersection are always mutually linked in both directions. The intersections on short parasitic segments are linked in one direction only after LINKING of all images, and therefore do not belong to the grid. The complexity of the calculation of this part is proportional to n,. 6.2 The projection of the image coordinates.
  • the camera position for the k m image is determined by :
  • -vk / f k is the position of the tool centre point (TCP) of the camera 8, i.e., the centre of the image in the focal plane of the camera's lens system, in other words the point we are looking at.
  • p k is the point of optical action (POA) , i.e., the projection centre during the image formation and f is the focal distance, e ⁇ , e vk , and e wk are the unit vectors of the camera coordinate system.
  • the coordinates of a point u, of the image k are (u,v) .
  • x is the projection to the surface of X, from p k , it is sufficient to perform this projection to establish the world coordinates of the grid point.
  • a parametric representation through splines is irreversible, which means that it is possible to calculate the parameter coordinates from the world coordinates but not vice versa. Therefore, the projection is an iterative operation. It is not necessary to elaborate on this, as those processes are amply clarified in the existing literature.
  • the parameter coordinates of the grid points are a side product of this operation. The complexity of calculation of this part is proportional to n p .
  • the complexity of the calculation of this part is determined by the sorting and is proportional to n p .log 2 n p .
  • the aim is to reduce all data in a list of doubles to the main element in the list of grid points.
  • the coordinates this is done by means of a weighted average that depends on the distance of the grid point to the focal plane and the slope of the projection beam and the normal to the surface in the projected grid point.
  • the mean coordinates are stored in the main element.
  • the different images have been recorded from different angular positions of the CMM's probe head. Consequently, the northern direction of the different images does not necessarily match the same direction in the deformed and non-deformed grid. Linking information of the different images can only be used to reconstruct the grid if the link directions of these images are compatible.
  • the method used in the invention also offers the possibility to derive information from the distances between the points and lines. Because the surface can be entirely arbitrarily in the thre-__-dimensional space, it is useless to use euclidean distances in this space. However, it is useful, to use the integrals of the lines over the surface, although this proves to be very time- consuming.
  • the only space that can be used is the two- dimensional parameter space (s,t) of the spline description.
  • the grid coordinates of the points immediately emerge from the grid coordinates of the line segments to which they belong.
  • the grid coordinate of the line segments from both families of line segments is determined in the same way. This happens for the NS-family.
  • a reference line segment is taken, named and marked as having been processed. All line segments to which the reference line segment refers to most closely, or which refer to the reference line segment of reference most closely, can be named. In the east the grid coordinate is higher by one; in the west, it is lower by one. All these line segments are marked as having been processed and are used in turn as reference line segments. The process stops when all line segments have been named. Even now mistakes may occur because line segments are missing or because there are parasitic line segments. The mistakes can be eliminated by insisting that all line segments in the east refer to a line segment with a higher grid coordinate and by renaming possible deviations.
  • the points are first sorted against the composite key i,j and are kept sorted by moving the points, the coordinates of which have been adapted, up in the list.
  • the result at this point is that the grid is correctly named as graph, i.e. the line segments in the neighbourhood of the missing line segments may have received the number of a missing line segment and parasitic line segments also have a number.
  • the continuity of the strain allows a correction of these mistakes by an ultimate renaming, taking account of the distances in the three-dimensional space.
  • the grid coordinates equal the grid coordinates of the line segments to which they belong.
  • the complexity of the calculation of this part is n p .log 2 .n p .
  • the grid coordinates of the different connected parts have been determined, apart from a translation in the (i,j) space. These translations are not important for the calculation of the strain. But they are required if one wants to have an overview of how the material flows. They are determined by fitting the different pieces in the parametric space (s,t) together as the pieces of a puzzle.
  • the output of the STRAINCALC module for each grid point consists of :
  • the coordinates of all points of intersection of the deformed pattern 10 are the basis for this calculation.
  • Measuring the displacements is one of the original possibilities of the system.
  • For the calculation of the strains we use the proven methods of SHEDIN & MELANDER and of SOWERBY et al. Shedin E & Melander A. The evaluation of Large Strains from Industrial Sheet Metal Stamping with a Square Grid . J. Applied Metalworking, Vol . 4, no. 2, Jan. 1986. Sowerby R. , Duncan J.L. & Chu E. , The Modelling of Sheet Metal Stamping , International Journal of Mechanical Science, Vol. 28, 1986, p. 415-430.
  • Fig. 9 shows the results of strain measurement with the principal strain depicted in direction and size.
  • Other ways of presenting basically the same information are always possible, e.g. "iso-strain lines" for equivalent strains (Fig. 10) .
  • the presented system is of an original concept because of the LOGICAL and MODULAR approach.
  • the system is furthermore able to treat components with erased zones in the same automated way as ordinary components thanks to linking technology.
  • the accuracy of the system's measuring results does not depend on the accuracy of the pattern applied to the component, because the "inaccurate" patterns can be measured in the same way before the component is formed.
  • the system is superior because of its accuracy regarding the measurement of strain. Contrary to the stereoscopic system described in document USA-4969106, the size of the image is small. For example, if the size of the image is 16mm, the pixel resolution is +30 ⁇ m. If the cell in the pattern is approximately 5 mm, this means that the strain's resolution is 0.6%.

Landscapes

  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Length Measuring Devices By Optical Means (AREA)

Abstract

The present invention relates to a measuring system suitable for strain analysis of three-dimensional patterned components (10). The system features a camera (8) with five degrees of freedom, mounted on a coordinate measurement machine (CMM) (1) which has a built-in computer (7), linked to an image processing system. The computer is responsible for all measuring and calculation processes, i.e.: preliminary measurement of the global shape of the component (3) using sensor technology; positioning and orienting of the camera; communication with the image processing system to determine the number of images; transmitting this number of images to the image processing system, so as to calculate the parameters of the pattern; actual calculation of the parameters; application of several calculation algorithms to these parameters in order to convert the image output to a connective mesh of three-dimensional world coordinates of the points of intersection; calculation of the field of displacement and strains as well as graphical and/or numerical presentation of the results.

Description

MEASURING SYSTEM SUITABLE FOR STRAIN ANALYSIS OF THREE-DIMENSIONAL COMPONENTS.
The present invention relates to a measuring system suitable for strain analysis of three-dimensional components to which a pattern has been applied, consisting of a camera mounted on a coordinate measurement machine (CMM) which has a built-in computer linked to an image processing system.
The purpose of the invention is to measure displacements and strains in the surface of a mechanical component after deformation in a non-machining mechanical or hydraulic process, such as deep drawing, drop forging or forging of steel and/or aluminiur..
To performstrain analysis in case of deep drawing, the component to be examined is provided with a pattern of regular, well known geometric form and size before deformation. Frequently used patterns include a grid of circles, completely filled up or the circumference only (Circle Grid Analysis = CGA) , completely separated or intersecting and a lattice of mutually perpendicular lines forming a square mesh (chequered pattern) (Square Grid Analysis = SGA) .
The grid is transferred to the component in the familiar electrochemical, notochemical, printing or photographic way.
The component, and more particularly a blank, is subsequently deformed in the process, which also deforms the grid. The strains are identified by measuring the deformed grid. In Circle Grid Analysis, the circle grid is transformed to a set of ellipses. The extent of the strains can be derived from the length of the minor and major axes of each ellipse. Then the strains are displayed in a strain diagram as principal main strains. However, with Square Grid Analysis the displacement of the points of intersection of the grid's mutually perpendicular lines is measured.
Measurement of the strains as mentioned above is a time- consuming and laborious enterprise. Yet, deformation analysis - particularly the measurement of strain - is a very important tool to evaluate the application of the material used to fabricate the component. The determination of the displacements is important for the evaluation of the deformation process necessary to produce the component. Its importance is dual. On the one hand, it allows for assessing the process itself. On the other hand, there is a growing need for numerical deformation process simulation so as to reduce the number of time-consuming and costly try-outs. But numerical simulation is a practical tool only if the input data (boundary conditions) correspond to reality. Today's simulation technology requires a realistic verification of the calculation methods applied and the data used. The strain analysis is also meant to optimize the interaction between plate, tool and lubricants. This allows the die designer to check the safety margin of a certain point on the deep drawn part against the Form Limit Diagram. This method also gives an idea of the distribution of the strain in a specific zone, the direction in which the material flows and the entire strain at a specific point.
Thus it is possible to verify whether tool concept, blank design, steel quality or lubricant need to be reviewed.
A method to measure the distribution of strain on a deformed surface is described in document USA-4969106 (J.H. VOGEL) . The method uses a CCD-camera and a vision system to process two images received through stereo¬ scopic recording. It consists of the following steps :
- application of a traditional pattern on the component to be deformed;
- recording of two pictures of the deformed surface from two different angles, so that there is a geometric relationship between the two;
- digitalization of the grid points of the deformed grid for each image to obtain two sets of two-dimensional coordinates of these grid points;
- processing the individual images, either manually using a cursor to identify the intersection point of the pattern lines, or by means of image processing software which identifies the relationship between the points of intersection and the adjoining points;
- processing in case of automatic image treatment basically consists of two successive filtrations (average noise reduction and high pass convolution with an elementary cell 7 x 7) ;
- if the lines are broken, there is a need for the operator to interfere MANUALLY and to reconstruct the lines using the cursor (patching) . This results in two sets of points of intersection, the common part of which should be indicated MANUALLY by the operator;
- the two sets of two-dimensional coordinates are merged into a set of three-dimensional coordinates (Combining) . At least one corresponding point in the two sets should be indicated by the operator MANUALLY; - calculation of one set of three-dimensional grid point, coordinates as a function of the geometric relation between the two images and the two sets of two- dimensional coordinates of the connected grid points;
- calculation of the distribution of the strain over the surface as a function of the three-dimensional coordinates of the grid points.
A first disadvantage of this method is that it requires frequent manual intervention by the operator during the measuring process, as the image processing algorithms are very rudimental, for example for :
- the determination of the common zone in the two individual images (boundary of regions) ;
- the correspondence between two identical points in the images matching one point of reference of the component
- the restoration of the broken lines to enable independent operation of the calculation programme.
A second disadvantage is that the accuracy of the above described stereoscopic system decreases when the size of the component increases.
Because a camera's resolution (usually up to 512 pixels, sometimes 1024) is limited, the main size of the component can be divided into a maximum of 512 or 1024 parts only. If an acceptable level of accuracy is to be reached, the size of the component has to be limited. There is no method or instrument that can subdivide the component automatically into smaller zones for which an acceptable degree of accuracy was possible.
The combination of all of the disadvantages listed makes the described USA-4969106 method unsuitable for systematic analysis of larger components, i.e., those found in practice in industry.
The present invention aims at solving these problems. It is a measuring system especially suited for strain analysis of three-dimensional patterned components, consisting of a camera built on a coordinate measurement machine (CMM) which has a built-in computer linked to an image processing system. Typical of this measuring system is that the CCD-camera has at least five degrees of freedom and that the computer is in charge of the entire measuring and calculation process, i.e. :
- preliminary measurement of the global shape of the component using probe scanning;
- positioning and orienting of the camera, which has a probe head with two degrees of freedom; - communication with the image processing system to determine the number of images; transmitting the number of images to the image processing system so as to calculate the parameters of the pattern;
- application calculation of the parameters; - application of several algebraic algorithms to these parameters for the conversion of image results to an unbroken meshing of three-dimensional world coordinates of the points of intersection;
- calculation of field of displacement and strains as well as graphical and/or numerical display of the results.
The method has the advantage that, for economic and technical reasons, it complies with the increasingly stringent requirements to the material itself and to the corresponding forming processes. It allows for accurate and reliable methods of analysis possible. It is not impaired by the complexity of components, generating a large number of data. The methods of analysis applied, quickly and automatically yield results that can be used in practice with minimal operator intervention (ergonomic objective) .
According to a particularity of the invention, the measuring system has a measuring head which can comprise both a touch probe and a camera, which are mutually and automatically interchangeable. An innovation is that the touch probe is controlled by appropriated scanning hardware and software during the process of measuring the shape of complex components. A control hardware and software move and position the camera around the component, avoiding contact between camera and component, both in the positions from which the images are taken and during intermediate movements.
A particular design of the coordinate measurement machine is of the bridge type, fit to move in three orthogonal directions, with a measuring head that can revolve around a vertical and a horizontal axis respectively, a computer to control the measuring stand meant for scanning the deformed component and with specific procedures linking and controlling the above mentioned image processing system, and for calculating and displaying the results.
The material of the component does not impose any restriction on the application of the present invention. It can either be metal, plastic or composite material or any other which can be applied in solid, plate or layer shape. The material should be suitable for the application of any speci ic pattern on the surface before the forming process, after which the (also deformed) pattern must still be recognizable. The nature of the recognizability is unimportant. The presently introduced machine is based on visual recognition using a pattern. Any way of recognition, such as infrared, magnetic, ultrasonic, ultraviolet etc., with which the same objective can be reached following the same procedure, also belongs to the scope of the present invention. The expressions "vision" and "image" cover the general concepts of "recognition" and "result of recognition". Other characteristics of the invention will be dealt with in the following description of procedures and equipment of the invention. The enclosed drawings, relating to the description, are given as examples only. - figure 1 shows a perspective drawing of a measuring system along the lines of the invention;
- figure 2 shows a large scale picture of a touch probe during the process of measuring a component;
- figure 3 shows the measured surface represented by means of SPLINE techniques;
- figure 4 schematically shows the camera scanning a component;
- figure 5 displays a difficult image before processing;
- figure 6 is an image after processing; - figure 7 is an image with a crack after processing;
- figure 8 shows a component;
- figure 9 is a diagram on which the strains are represented by order of size and direction;
- figure 10 shows a diagram with equivalent strains.
The same symbols in the illustrations indicate identical or similar elements.
As shown • in figure 1, the system relating to the invention comprises the following components, building together one integrated system to perform a number of consecutive subtasks :
- a three-dimensional coordinate measuring machine 1 (CMM) , e.g., of the HA80 type of LK Ltd. ® (England) , with at least five degrees of freedom and an appropriate touch probe 2 (of the TP2 type of Renishaw ®, England) to measure the global shape of the component 3, consisting for example of a heavy granite table 4 as basic element, with on it a bridge moving in longitudinal direction (X) . On the portal, a unit with inside a vertically movable (Z-direction) quill 6 with the probe head 9 at the free end, moves in Y-direction;
- computer(s) 7 for controlling the measuring stand and processing all the different data in the different stages of the measuring process; - a camera 8 with appropriate lenses, automatically interchangeable with the probe 2 of the measuring machine;
- image processing system for the operation of images received via the camera 8; - all software and hardware, e.g. , MicroVAX ® for controlling the probe 2 and/or camera 8 and linking the image processing system and measuring machine 1;
- all software and hardware, e.g., MicroVAX ® for the calculation and displaying of displacements and strains.
- At the bottom of the quill 6 (Fig. 2) is a measuring head 9, with the measuring probe 2 . The probe head 9 is of type PH10M of Renishaw ®. It offers two additional rotational possibilities about a vertical and horizontal axis respectively. The probes can be placed under different angles, so that almost every point of the component to be measured becomes accessible.
For specific applications where the traditional touch probe with ruby ball could lead to deformation (permanent or elastic) of the component because of the very contact during measurement, a non-contacting laser probe has been provided (e.g., of type OP5 of Renishaw ®) , which operates by the principle of triangulation. By using this probe, the scanning speed can be increased, which is particularly interesting for scanning large curved surfaces. The measuring stand moves over very precisely defined paths and air bearings and is driven by servo-controlled electric motors. The position in each direction of motion is measured with high precision digital measuring rulers (Heidenhain ®, Germany) .
An automatic switch system makes it possible to choose between the different probe configurations from the measuring programme.
The machine includes a monochrome CCD-camera 8 focused on that spot of the component which is to be analyzed.
Clearly defined modules, which are logically and mutually integrated, are used in a special design of the measuring system :
1. Preparation of the components (GRIDDING) (Fig. 4)
By analogy with the system described in USA-4969106, a pattern 10 is applied on the component in the usual way before deformation. However, the choice of a pattern with square meshing in no way restricts the application of the device.
2. Determination of the global shape of the component.
If the global shape of the component is not known in advance, it is possible to measure this shape with the coordinate measuring machine, ut ^ng suitable scanning software for the measurement of complex surfaces. Fig. 2 shows a touch probe 2 during measurement. Laser probes can be used too. If the range of the touch probe 2 should require so, the operator has the possibility to divide the component 3 in different zones (ZONING) . The measurement yields the 3D-coordinates of a large number of points of the registered surface, but not those of the points of intersection of the pattern lines (because they are not perceptible for the probe) . The software package for the different measuring tasks consists of different levels :
- CMES (Continuous Measuring System) is a traditional component programming software for prismatic components, consisting of both teaching and off-line programming. All state-of-the-art devices are present, e.g., automatic component alignment;
- Scanning software This software is a special application for scanning randomly curved surfaces, such as those used in car manufacturing (parts of the body) .
- High Level Programming. This module makes it possible to develop subroutines accessible from CMES. This can be done on the basis of functions that are not accessible at level 1, such as extensive communication with peripheral machines and/or the internal computer structure of the MicroVAX ®.
3. Mathematical description of the surface (FORMCONSTRUCT) . Through the measurement points of the measured surface a mathematically defined surface is drawn by means of Spline Methods (Fig. 3) . These mathematical methods are described in a handbook by Gerald E. Farin - Curves and Surfaces for Computer Aided Geometric Design, A Practical Guide , Academic Press, Inc., San Diego, California (1990) .
An automatic procedure divides the component into triangular zones, the dimensions of which are chosen so that they can later be taken in one general view by the camera 8. The direction perpendicular to the surface is calculated for each triangle, as is the future position of the camera 8. 4. Images of the deformed surface (CAMSCAN : CAMera SCANning)
The mathematical description of the surface to be analyzed provides the data for the control of the 3D- measuring stand (direction and position) and the positioning of the camera 8. Special procedures have been worked out to optimize the path of the camera 8 in time and to avoid collision with the component 3. Fig. 4 schematically shows the camera 8 scanning a component 3. In reality, the camera is much smaller : length ± 150 mm, diameter ± 30 mm.
Changing over from touch probe 2 to camera 8 can be done automatically, because of the special construction of the camera 8.
Different lenses can be mounted on the camera (manually) or several automatically interchangeable cameras can be used, each with its own lens. Furthermore, it is possible to use zoom lenses, which can be controlled automatically.
Tests have already been carried out with lenses with a range of 25, 16 and 10 mm2 respectively. The lenses have an excellent depth of field, so that the curved areas of the component 3 very sharply visible.
When the camera 8 reaches a preset position, the images are transmitted to the image processing system. The number of images recorded depends on the size of the component 3 to be examined. This number is basically unlimited for off-line processing of images. For example, if an optical disc is used for image storage there is actually only a theoretical limit. One single image of 512*512 pixels with 8 grey shades requires 262 Kb. Therefore, an optical disc of 900 Mb can contain approximately 3400 images, which means that in reality there are no strict limits, not even with on-line processing.
5. Single image processing (SIGPAD : Single Image Grid Parameter Determination)
Each of the recorded images is dealt with separately in an image processing programme in order to calculate the characteristics of the elementary cells. If the pattern is a square meshing, the points of intersection of the line pattern need to be identified. In case of a circular pattern the centres of the ellipses can be treated in the same way. To be able to determine all the points of intersection automatically, without operator assistance, very powerful algorithms have been elaborated, which are also able to process "difficult" images of "poor quality". If the component is a steel plate, the contrast is very poor due to the considerable and irregular reflections resulting from the complicated component shape and its roughness.
The strength of the algorithms is their adequate succession and combination of different procedures, all based on the principles of mathematical morphology. The procedures use the entire image and this without any prior knowledge of the characteristics of the image itself. The different procedures successively take care of :
- smoothing the irregular background, caused by non- homogenous exposure and reflection; - determining the directions of the two families of lines in the pattern;
- extracting the two families of grid lines;
- eliminating parasitic lines, caused by local differences in contrast; - determining the points of intersection of the two families of grid lines and the grid lines to which they belong.
5.1 Smoothing of the background.
By means of the SMOOTH function (from mathematical morphology) the mean background of the image is determined and subtracted from the original image. This results in an image with a regular mean background. This image is made binary.
5.2 Determination of the directions of the two families
The two directions of the grid lines are determined by means of the two largest local maxima of the co-variance, calculated for the different directions in the image.
5.3 Extraction of the two families of grid lines
A family of grid lines is extracted by filtering out the other family aπci the background noise by means of an OPENING function with a linear, structural element in the direction of the grid lines determined under 5.2. Grid lines that are broken off after this procedure, are patched again by means * of a CLOSING function with a larger linear structural element. The definition and operation of the OPENING and CLOSING functions are described in the literature on mathematical morphology. See the publication of Jean Serra - Image Analysis and Mathematical Morphology, Volume 1 , Academic Press Inc., San Diego, California (1989) .
.4 Elimination of parasitic meshes
The filtering in 5.1 and the operations in 5.3 cannot detect local differences in contrast. Consequently, parasitic short line segments are sometimes created. Some of these line segments connect grid lines of the other family and therefore cause parasitic meshes in the deformed pattern 10. Others cause free projections that intersect maximum one grid line of the other family. These parasitic lines are eliminated by successive application of the OR function to both images from 5.3; the SKELETON function applied to the new image; the PRUNE function, which removes the free projection; and the linear OPENING function in both directions, after inversion of the image obtained after pruning, which eliminates the parasitic meshes.
5.5 Determination of the points of intersection of the two families of grid lines and the grid lines to which they belong.
The determination is done in two different ways. In the first one, the points of intersection are determined by taking the logical intersection using the AND function of the two families of line segments in the two directions, followed by the SKELETON function that reduces all objects in the resulting image to one pixel. Here, the resolution will never be higher than one pixel. A typical output file of SIGPAD, which supplies all the necessary data, is printed beneath. The first four lines contain general information. The first line contains the name tbOOl.img of the file housing the image concerned. The second line mentions the image number of the lens used.
The third line gives the image number.
The fourth line lists the slope of both directions. The next line is "D 1 3", which describes how many segments there are in the first direction, in our case three. For each segment there is a line defining the number of points belonging to that segment, e.g., "S 1 2 3", which means that segment 1 contains two points. Finally, the coordinates of these points are given. The same procedure is repeated for the second direction.
TABEL 1 : Typical output file of SIGPAD
d:tb001. img 2 001
0 5 5 0 D 1 3
S 1 2
118 465 73 467
S 2 2
407 369 416 474 S 3 3 282 72
D 2 4
S 1 2
118 465 73 467
Ξ 2 2
416 474 310 477 Ξ 3 2 407 369
304 376 S 4 1
282 72
The format used implies that the coordinates of each point occur twice, which allows the reconstruction of the connections between adjoining grid points.
In the second case, the line segments are described mathematically by means of the commonly known regression methods. The points of intersection are the algebraic intersections of the line segments in both directions. In this way, a subpixel resolution is used and a number of non-detected points can be interpolated very accurately. The display of the information is the same as in the first method, but the coordinates are no longer necessarily integers.
Fig. 5 shows an example of a "difficult" image before processing. The image shows part of the component 3 in the steel plate. The elevation in the middle is bright because of the reflection; especially at the left hand side the component fades out to the back, making the exposure very weak and the contrast between the lines and background very vague. After processing, all the points of intersection can be traced : Fig. 6.
For each image a set of (pixel) coordinates is obtained for all points of intersection in the image concerned after processing. Cracks or parts with an erased pattern do not disrupt the image processing methods used.
Fig. 7 gives an example of an image with a crack after processing.
6. Conversion to world coordinates and combining of all images (LINKING) .
The results of the separate images need to be combined. First conversion factors are applied to translate the pixel coordinates into real dimensions (mm) . The conversion factors are obtained through stringent calibration procedures applied to a pattern 10, the dimensions of which are exactly known by measurement with a measuring microscope. During the control and calculation processes record is kept of which (plane) image displays match each camera position. This makes it possible to convert them into absolute coordinates (world coordinates) in a three-dimensional space. Special attention is payed to the exact connection ("knitting") of the separate images, so that the mutual orientation (east-west) , (north-south) of all points of intersection is exactly determined. If there is a crack 11, it is possible to restore the unbroken pattern 10 with the applied linking techniques, under the condition that at least along one siαe of the crack, a part of the pattern is still intact. If the pattern 10 has been erased over a part of the component (e.g., due to the forming process itself) , it is also possible to restore the broken pattern 10.
The objective of the LINKING module is to reconstruct the deformed grid through the aid of the information displayed in the individual images. The parametric description of the form of the measured surface (FORMSCAN) takes place with splines (NURBS) . Four coordinate systems are referred to. The image coordinates (u,v) are the pixel coordinates of a point in the image. The origin is the centre of the image. The world coordinates (x,y,z) are the coordinates of a point within the CMM's system of reference. The parameter coordinates (s,t) are the coordinates of a point on the surface, in the spline-parameter space representing this surface. The grid coordinates (i,j) are the discrete coordinates of a grid point on the measuring raster. The different steps necessary to determine the LINKING module are :
- generation of the linking information in each image;
- determination of the parameter and world coordinates of the grid points through the projection of the grid point image coordinates on the surface;
- identification of the "doubles", i.e., the points occurring in two or more images and necessary to link the different images;
- elimination of the doubles, but with preservation of all information they contain;
- correction of the linking information of the individually connected images if the images have been taken from different camera positions, i.e from different angles of the camera; - searching in a group of connected points for the grid lines composing that group;
- determination of the grid coordinates of these lines, and the points they contain;
- relating the grid coordinates of the different connected parts of the deformed grid to each other;
- generating them as output for the STRAINCALC module.
These problems are briefly dealt with one by one. The complexity of the calculation is described against the number of grid points np and the number of images n,.
6.1 Generating the linking information in an individual image.
The SIGPAD module extracts all data concerning the grid from an image via mathematical morphology filterings. For each image the grid point image coordinates (u,v) in that image are given. When the grid is a square one, the two directions of the deformed grid lines are given as well. To demonstrate how these points are mutually linked by these lines, we assume the following. There are four links in each point. The northern link (N-link) is situated along the most vertically oriented grid line in the sense of a mounting v; the eastern link (E-link) is located in the other direction in the sense of the mounting u; the definitions of the southern link (S-link) and western link (W-link) are clear from the above mentioned. It is very important to realize that points can be missing while the link between the adjoining points is not missing, and that links can be missing between points that can be detected. Of course, it is also possible that parts of the grid were erased during the process or that parasitic points are detected due to contrast effects and reflection.
The SIGPAD module extracts the line segments in the deformed grid's two directions after several image filtering operations. We are interested in the points of intersection of the line segments in the two directions. Two types of mistakes that are not easily intercepted by the image filterings, but that can be detected and eliminated easily afterwards are : multiple points of intersection and points of intersection as a result of parasitic short line segments which intersect only one segment in the other direction. This is exemplified in 5.5, where there are multiple points of intersection between the line segment "SI" in the first direction and "SI" in the second direction. Multiple points of intersection are always mutually linked in both directions. The intersections on short parasitic segments are linked in one direction only after LINKING of all images, and therefore do not belong to the grid. The complexity of the calculation of this part is proportional to n,. 6.2 The projection of the image coordinates.
The camera position for the km image is determined by :
Pk = k + f -wk
-vk / fk is the position of the tool centre point (TCP) of the camera 8, i.e., the centre of the image in the focal plane of the camera's lens system, in other words the point we are looking at. pk is the point of optical action (POA) , i.e., the projection centre during the image formation and f is the focal distance, e^, evk, and ewk are the unit vectors of the camera coordinate system.
The coordinates of a point u, of the image k are (u,v) .
Thanks to calibration we know the scaling factors in the u and v directions and we can calculate the coordinates of the same point U, in the camera coordinate system. They equal (scu. u,, scv . v,, 0) . The world coordinates X, of that point are given by the following transformation
Figure imgf000022_0001
Since x, is the projection to the surface of X, from pk, it is sufficient to perform this projection to establish the world coordinates of the grid point. A parametric representation through splines is irreversible, which means that it is possible to calculate the parameter coordinates from the world coordinates but not vice versa. Therefore, the projection is an iterative operation. It is not necessary to elaborate on this, as those processes are amply clarified in the existing literature. The parameter coordinates of the grid points are a side product of this operation. The complexity of calculation of this part is proportional to np.
6.3 Identification of doubles.
To identify the doubles we transform the set of all image grid points displayed in the images to a set of equivalence classes by means of an equivalence relation. The high accuracy of the CMM easily allows us to define such a relation. The obvious definition is that all image grid points that are located within a distance eCMM from each other, are all elements of the same grid point. e MM is a distance depending on the accuracy of the CMM. To avoid the explicit calculation of euclidean distances we apply the slightly different criterium :
I Xι~Xj I - eCMM x, = Xj « l y.-y ≤ CCM
I Zι- Zj I - eCMM
Testing all these points two by two with this criterium would make the complexity of the calculation quadratically dependent on the number of points. To avoid this, we sort all image grid points on the x-coordinates and index them, taking account of the first inequality of the criterium, in other words as long as the x-coordinates of the successive points in the sorred list comply with a first inequality, they receive the same whole index. It suffices to run through the list linearly. The same operations are applied to the y en z coordinates, which results in three indices for each point. All points with the same three indices are elements of the same grid point. By linking the different sortings in an appropriate way, we need to run through the sorted final list linearly only once. During the process the list is transformed to a linear list with one element of each grid point while sublists of other elements can be added. The entire procedure is basically a variation on sorting against several keys, a procedure which is described elaborately in literature.
The complexity of the calculation of this part is determined by the sorting and is proportional to np.log2np.
6.4 Elimination of the doubles.
The aim is to reduce all data in a list of doubles to the main element in the list of grid points.
For the coordinates this is done by means of a weighted average that depends on the distance of the grid point to the focal plane and the slope of the projection beam and the normal to the surface in the projected grid point. The mean coordinates are stored in the main element.
For the links this is done by tagging all links with a refe ence to the main copy of the class of doubles to which the link belongs. As the number of links remains the same, we keep the data structure of the linear list of main elements with the connected list of doubles, but only the link information is relevant.
The complexity of the calculation of this part is proportional to np. 6.5 Correction of rotations.
The different images have been recorded from different angular positions of the CMM's probe head. Consequently, the northern direction of the different images does not necessarily match the same direction in the deformed and non-deformed grid. Linking information of the different images can only be used to reconstruct the grid if the link directions of these images are compatible.
To achieve this objective, we take one image as reference and indicate that this image has been processed. For all images that have doubles in common with this image, we examine whether they have a link in common from that double as well. If this is the case, the two images have doubles in common, as well as the link between those doubles, which does not necessarily imply that the direction of the links is the same. If this is the case, we must rotate the second image over one, two or three quadrants, which means that the links swap places as if this were a rotation. For example, a rotation of one quadrant means :
nl *- el el •«- si si *- wl wl +- nl
The swaps are analogous for the other rotations.
If necessary, several images can be rotated with the first image. All of the rotated images are marked to have been processed and can in turn be used as references. The process stops once all images of reference have been exhausted. This does not mean that all images are now compatible. It is possible that subsets of images exist that do not overlap anywhere. If so, the entire process starts all over again with an image that has not yet been processed. This kind of problem is often encountered in questions of connectivity. We refer to the existing literature on topology and graphics for further details.
If an image has only doubles in common with an image of reference and not the matching link, it cannot be rotated. Then the doubles have to be undoubled to avoid mistakes later, since the points of both images may be connected but the directions may be incompatible.
The complexity of the calculation of this part is proportional to np.
6.6 Determination of the connected parts.
To determine all points of a connected part of the grid, we take one point of reference and indicate that it has been processed. We then determine all the points connected with the point of reference and unprocessed so far. We mark these points as having been processed and use them in turn as points of reference. The process is ended when all points of reference have been exhausted. Again we refer to the literature on topology and graphics for further details.
To determine the next connected part of the grid, we take a point that has not been processed yet and repeat the entire procedure. When all points have been processed, all the connected parts are identified.
The complexity of the calculation of this part is proportional to np. 6.7 Determination of the grid lines within a connected part.
To determine the line segments of grid lines within a connected part in two directions, we apply the same procedure as described above, but use the north-south or east-west links only. Each of the line segments is added to a list of line segments. The operation results in two lists of line segments, also called families : one with north-south (NS) and one with east-west (EW) line segments.
The complexity of the calculation of this part is proportional np.
6.8 Determination of the grid coordinates within a connected part.
It is rarely possible to reconstruct the grid coordinates (i,j) of a quadrangular grid, starting from linking information only, if not all points and links are present. The method used in the invention also offers the possibility to derive information from the distances between the points and lines. Because the surface can be entirely arbitrarily in the thre-__-dimensional space, it is useless to use euclidean distances in this space. However, it is useful, to use the integrals of the lines over the surface, although this proves to be very time- consuming. The only space that can be used is the two- dimensional parameter space (s,t) of the spline description.
The grid coordinates of the points immediately emerge from the grid coordinates of the line segments to which they belong. The grid coordinate of the line segments from both families of line segments is determined in the same way. This happens for the NS-family.
We basically need to change from the connected grid of points to a connected grid of line segments preserving all the useful information. The EW links for the NS line segments need to be determined. For reasons of analogy we limit ourselves to the E links. Because not all of the points of an NS line segment refer to the same NS line segment in the east direction, all references of the line segments need to be extracted from all references of the points.
To be able to name the NS line segments, a reference line segment is taken, named and marked as having been processed. All line segments to which the reference line segment refers to most closely, or which refer to the reference line segment of reference most closely, can be named. In the east the grid coordinate is higher by one; in the west, it is lower by one. All these line segments are marked as having been processed and are used in turn as reference line segments. The process stops when all line segments have been named. Even now mistakes may occur because line segments are missing or because there are parasitic line segments. The mistakes can be eliminated by insisting that all line segments in the east refer to a line segment with a higher grid coordinate and by renaming possible deviations. As this may lead to new deviations, this can be done adequately only if the line segments are sorted on the grid coordinate first. The NS segments have been arranged topologically in a linear list. The same happens with the EW line segments. The grid coordinates equal the grid coordinates of the line segments to which they belong. Usually the naming will still be ambiguous. To solve this ambiguity, the distances have to be taken in account. If two points have the same grid coordinates, we determine how many line segments are intersected by the linking line between these points, both in NS and in EW directions, and we increase the grid coordinate of the most northern and most eastern line segments by the respective number of intersections plus one. This procedure is repeated until no two points have the same (i,j) coordinates. To detect points with the equal (i,j) coordinates, the points are first sorted against the composite key i,j and are kept sorted by moving the points, the coordinates of which have been adapted, up in the list. The result at this point is that the grid is correctly named as graph, i.e. the line segments in the neighbourhood of the missing line segments may have received the number of a missing line segment and parasitic line segments also have a number. The continuity of the strain allows a correction of these mistakes by an ultimate renaming, taking account of the distances in the three-dimensional space.
The grid coordinates equal the grid coordinates of the line segments to which they belong. The complexity of the calculation of this part is np.log2.np. We assume the number of line segments to be ΞQRT(np) .
6.9 Relating grid coordinates of non-connected parts.
The grid coordinates of the different connected parts have been determined, apart from a translation in the (i,j) space. These translations are not important for the calculation of the strain. But they are required if one wants to have an overview of how the material flows. They are determined by fitting the different pieces in the parametric space (s,t) together as the pieces of a puzzle.
6.10 Generating the output
The output of the STRAINCALC module for each grid point consists of :
- a sequence number of the grid point;
- the grid coordinates of the point; - the world coordinates of the point;
- the sequential numbers of the linked points in the four directions.
Once all the different steps of this module have been run through, all points of intersection of the deformed meshing are known.
7. Calculation of the field of displacements and strains
(STRAINCALCulation)
The coordinates of all points of intersection of the deformed pattern 10 are the basis for this calculation.
If the original coordinates of the meshing on the deformed component are x,y,z and the coordinates of the meshing after the deformation are x', y', z', the displacements dx, dy, d7 for all mesh points are to be calculated as follows :
d = x' -x dv= y'-y d7.= z ' -z
Measuring the displacements is one of the original possibilities of the system. For the calculation of the strains we use the proven methods of SHEDIN & MELANDER and of SOWERBY et al. Shedin E & Melander A. The evaluation of Large Strains from Industrial Sheet Metal Stamping with a Square Grid . J. Applied Metalworking, Vol . 4, no. 2, Jan. 1986. Sowerby R. , Duncan J.L. & Chu E. , The Modelling of Sheet Metal Stamping , International Journal of Mechanical Science, Vol. 28, 1986, p. 415-430.
These are basically obtained from the set of world coordinates in phase 6. These methods have been implemented on computers that are part of the system of the present invention.
8. Presentation of the Results PRESRES / PRESentation of RESults) .
The results are presented through the traditional methods available in software packages such as CAD-software, or those taking care of preprocessing for FEM (e.g. I-Deas) . To demonstrate the system, an analysis was carried out for part of the component represented in fig. 8.
Fig. 9 shows the results of strain measurement with the principal strain depicted in direction and size. Other ways of presenting basically the same information are always possible, e.g. "iso-strain lines" for equivalent strains (Fig. 10) .
The presented system is of an original concept because of the LOGICAL and MODULAR approach.
It can measure strains and fields of displacements by linking all the recorded images. The latter is a considerable improvement in the scientific study of forming processes. The concept is also original because of another aspect, viz. that for the first time the possibility is created to analyze larger components BOTH in an industrial AND scientific way and on a FULL SCALE, since there are no limits to the number of images.
It is also the first system that can treat components with cracks in the same automated way as components without cracks, thanks to the applied image processing algorithms.
The system is furthermore able to treat components with erased zones in the same automated way as ordinary components thanks to linking technology.
The accuracy of the system's measuring results does not depend on the accuracy of the pattern applied to the component, because the "inaccurate" patterns can be measured in the same way before the component is formed. The system is superior because of its accuracy regarding the measurement of strain. Contrary to the stereoscopic system described in document USA-4969106, the size of the image is small. For example, if the size of the image is 16mm, the pixel resolution is +30 μm. If the cell in the pattern is approximately 5 mm, this means that the strain's resolution is 0.6%.
If even smaller resolutions are necessary, it is possible to use another lens with larger magnification. However, in reality there is a limit to the accuracy with which a pattern can be applied. But this problem can be solved as well if the "inaccurate" pattern is measured with the same equipment before the deformation of the plate. The system is highly automated. The need for the operator to interfere is limited to fixing the component on the three-dimensional measuring stand and to defining the areas to be examined during the measurement of the global shape. Both operations take place at the beginning of the analysis, and the operator is no longer required until the results are displayed or plotted. The image processing algorithms applied are very powerful and do not require manual intervention.

Claims

C L A I M S
1.- Measuring system suitable for strain analysis of three-dimensional components (3) to which a pattern has been applied (10) , consisting of a camera (8) mounted on a coordinate measurement machine (CMM) wich has a built- in computer, linked to an image processing system, characterized in that the camera has at least five degrees of freedom and that the computer is responsible for all measuring and calculation processes, i.e. :
- preliminary measurement of the global shape of the component 3 using probe scanning (2);
- positioning and orienting of the camera which has a probe head (9) with two degrees of freedom; - communication with the image processing system to determine the number of images; transmitting this number of images to the image processing system, so as to calculate the parameters of the pattern;
- actual calculation of the parameters; - application of several algebraic algorithms to these parameters for the conversion of the image output to an unbroken meshing of three-dimensional world coordinates of the points of intersection;
- calculation of the field of displacement and the strains as well as the graphical and/or numerical presentation of the results.
2.- Measuring system according to claim 1, characterized in that it is equipped with a probe head (9) which can contain both a touch probe (2) and a camera (8) , each automatically interchangeable.
3.- Measuring system according to claim 2, characterized in that the touch probe (2) is controlled by appropriate scanning hardware and software during the process of measuring the shape of complex components.
4.- Measuring system according to claim 2, characterized in that the camera (8) is moved and positioned by a control hardware and software around the component (3), avoiding collision between camera (8) and component (3) , both in the positions from which the images are taken and during the movements in-between.
5.- Measuring system according to claim 1, characterized in that the image processing application treats individual images in order to determine the points of intersection of the line pattern on the deformed square grid.
6.- Measuring system according to claim 1, characterized in that a special routine elimines the strong local contrasts, so that cracks (11) and erased parts can be operated upon correctly without manual intervention by the operator.
7.- Measuring system according to claim 1, characterized in that it has a special linking routine to determine the original grid coordinates for the points of intersection of the lines in the individual images, so that one connective deformed pattern (10) is generated, from which the displacements and strains can be calculated, even if there are cracks (11) or erased areas.
8.- Measuring system according to anyone of the previous claims, characterized in that the coordinate measurement machine is a three-dimensional bridge type measuring stand, able to move in three orthogonal directions, with a probe head (9) able to rotate in two ways, i.e., around a horizontal and vertical axis respectively, a computer to control the measuring stand by means of software dedicated for scanning the component plate, and with specific software for linking and controlling the above mentioned image processing system and for calculating the results and their presentation.
PCT/BE1993/000075 1992-12-10 1993-12-03 Measuring system suitable for strain analysis of three-dimensional components WO1994014029A1 (en)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
BE9201091 1992-12-10
BE9201091A BE1006447A3 (en) 1992-12-10 1992-12-10 Measuring system suitable for distortion analysis of three dimensional workpieces.

Publications (1)

Publication Number Publication Date
WO1994014029A1 true WO1994014029A1 (en) 1994-06-23

Family

ID=3886579

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/BE1993/000075 WO1994014029A1 (en) 1992-12-10 1993-12-03 Measuring system suitable for strain analysis of three-dimensional components

Country Status (2)

Country Link
BE (1) BE1006447A3 (en)
WO (1) WO1994014029A1 (en)

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2009005468A1 (en) * 2007-07-03 2009-01-08 Gcoder Systems Ab Pre tension monitoring solution
CN103884303A (en) * 2014-03-31 2014-06-25 青岛麦科三维测量设备有限公司 Multifunctional measuring machine
EP2749840A1 (en) * 2012-12-28 2014-07-02 Canon Kabushiki Kaisha Measurement apparatus and measurement method
EP3182057A1 (en) * 2015-12-17 2017-06-21 General Electric Company Components with array-based strain sensors and methods for monitoring the same
WO2018048872A1 (en) * 2016-09-09 2018-03-15 Gleason Metrology Systems Corporation Measurement of toothed articles utilizing multiple sensors

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO1990010194A1 (en) * 1989-02-27 1990-09-07 Camsys, Inc. Computerized method of determining surface strain distributions in a deformed body

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO1990010194A1 (en) * 1989-02-27 1990-09-07 Camsys, Inc. Computerized method of determining surface strain distributions in a deformed body

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
W.H. PETERS E.A.: "DIGITAL IMAGINGTECHNIQUES IN EXPERIMENTAL STRESS ANALYSIS", OPTICAL ENGINEERING, vol. 21, no. 3, May 1982 (1982-05-01), BELLINGHAM US, pages 427 - 431 *

Cited By (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2009005468A1 (en) * 2007-07-03 2009-01-08 Gcoder Systems Ab Pre tension monitoring solution
RU2467284C2 (en) * 2007-07-03 2012-11-20 ДжиКоудер Системз АБ Pretension monitoring
US8391561B2 (en) 2007-07-03 2013-03-05 G-coder Systems AB Pre tension monitoring solution
EP2749840A1 (en) * 2012-12-28 2014-07-02 Canon Kabushiki Kaisha Measurement apparatus and measurement method
CN103884303A (en) * 2014-03-31 2014-06-25 青岛麦科三维测量设备有限公司 Multifunctional measuring machine
EP3182057A1 (en) * 2015-12-17 2017-06-21 General Electric Company Components with array-based strain sensors and methods for monitoring the same
CN106895816A (en) * 2015-12-17 2017-06-27 通用电气公司 Component with the strain transducer based on array and the method for monitoring it
WO2018048872A1 (en) * 2016-09-09 2018-03-15 Gleason Metrology Systems Corporation Measurement of toothed articles utilizing multiple sensors
CN109661561A (en) * 2016-09-09 2019-04-19 格里森计量系统股份有限公司 Utilize the toothed article of multiple sensor measurements
CN109661561B (en) * 2016-09-09 2021-04-16 格里森计量系统股份有限公司 Measuring toothed articles with multiple sensors
US11092430B2 (en) 2016-09-09 2021-08-17 Gleason Metrology Systems Corporation Measurement of toothed articles utilizing multiple sensors

Also Published As

Publication number Publication date
BE1006447A3 (en) 1994-08-30

Similar Documents

Publication Publication Date Title
DE102015011914B4 (en) Contour line measuring device and robotic system
DE102008041523B4 (en) Method for three-dimensional measurement and device for three-dimensional measurement
DE102018109774B4 (en) Image processing system, image processing device and image processing program
Martins et al. Camera models based on data from two calibration planes
EP1522051A2 (en) Discrete linear space sampling method and apparatus for generating digital 3d models
Mason et al. Automatic sensor placement for accurate dimensional inspection
Motavalli et al. A part image reconstruction system for reverse engineering of design modifications
Spitz et al. Accessibility analysis using computer graphics hardware
CN114972421B (en) Workshop material identification tracking and positioning method and system
Zhao et al. Camera calibration with three noncollinear points under special motions
DE112022000320T5 (en) Processing method and apparatus for generating a cross-sectional image from three-dimensional position information detected by a visual sensor
WO1994014029A1 (en) Measuring system suitable for strain analysis of three-dimensional components
Galanulis Optical measuring technologies in sheet metal processing
EP1098268A2 (en) Method for the threedimensional optical measurement of object surfaces
CN116091401A (en) Spacecraft assembly part identification positioning method based on target detection and composite target code
Maeda et al. Measurement of the geometric features of a cutting tool edge with the aid of a digital image processing technique
Gronle et al. Model-based, active inspection of three-dimensional objects using a multi-sensor measurement system
CN108734763B (en) Method and system for digitizing micro-assembly space of micro-assembly system
DE112011104780T5 (en) Error display device and error display method
EP3889890A1 (en) Method for object detection on the basis of image data
CN111612071A (en) Deep learning method for generating depth map from shadow map of curved surface part
DE102019110185A1 (en) Method and system for registering a design data model in a space
DE102014209342A1 (en) Method for determining geometric data of an object using a measuring microscope and a measuring microscope
JPS6393076A (en) Form measuring instrument
Hecht et al. Triangulation based digitizing of tooling and sheet metal part surfaces-Measuring technique, analysis of deviation to CAD and remarks on use of 3D-coordinate fields for the finite element analysis

Legal Events

Date Code Title Description
AK Designated states

Kind code of ref document: A1

Designated state(s): JP US

AL Designated countries for regional patents

Kind code of ref document: A1

Designated state(s): AT BE CH DE DK ES FR GB GR IE IT LU MC NL PT SE

121 Ep: the epo has been informed by wipo that ep was designated in this application
122 Ep: pct application non-entry in european phase