A short cut to style

Improved hardware and software technology makes the process of creating a mathematical model from a physical part very straightforward. Martin Oakham looks at some of the reverse engineering options on offer.

Most new designs are variations on existing themes. Take the Vauxhall Calibra: is it a brand new design – or just a sporty Cavalier with a sexy rear end? Whatever you decide, you will probably agree that the change is usually too small to be classed as a complete re-design. Like most designs, cars undergo the vehicular equivalent of a new pair of Armani jeans and Calvin Klein underwear, just to keep up with changing fashions.

If there is no need to redesign the entire part, there must be a simple way of collecting enough information to make modifications easily. This method is generally known as reverse engineering – logical when you consider that you are working in reverse from the norm.

In addition to recording data about physical parts, or their tooling, reverse engineering is used whenever a part modelled in clay, wood or foam needs translating into a CAD model for modification. It is also used when all that is left of a previous job is the tooling, eg a mould, or when a competitor’s product needs examining or when final parts need verifying against the original CAD design.

Reverse engineering is an efficient process giving time savings as high as 90% when compared to traditional methods of data collection. But in order to take full advantage of the available technology it is important to look at your application very carefully. Just as a powerful CAD station is not used to design the odd screw, neither is a precision scanning system needed to reverse engineer simple prismatic surfaces. The power is needed when you’re reverse engineering complex curved surfaces such as an exhaust manifold.

Today’s computerised digitising methods include powerful algorithms which give users freedom to take measurements in a way that best suits the component, rather than forcing them to work within the limitations of manual techniques. Most reverse engineering systems used today are intelligent enough to determine the data density needed to define relatively complex 3D objects quickly and accurately. Main surfaces are scanned rapidly, since relatively few points are needed, while high densities of points are only taken where detailed information needs to be gathered. Measured data can be combined and used as source data to develop CAM programs for machining or stereolithography files for rapid prototyping – all with relative ease. It is now practical, for example, to design the main outline of a product with CAD, but use handmade models for complex details or decoration. The hand modelled sections can then be digitised and combined with the main CAD model in the computer.

The terms, ‘Scanning’ and ‘Digitising’ associated with reverse engineering are often interchanged, as both amount to collecting 3D data from a model or pattern. Digitising, though, actually refers to taking readings of 3D co-ordinates at predetermined discrete positions using a touch-trigger probe. Scanning, on the other hand, describes the production of a continuous stream of data points based on the deflection of a constant contact probe or laser strike triangulation. Of the two, scanning is the more efficient.

A criticism often levelled at scanning probes is that the physical size of the stylus prevents details such as sharp internal corners from being recorded correctly and that the pressure required by the probe prevents soft materials such as modelling clay from being scanned. However, this is only partly true. Renishaw, for example, claims that the combination of extremely low contact force (10-80g) and fine stepover offered by its SP600 allows materials like plasticine to be scanned with extremely small styli. For fine detail work such as coining dies, pointed styli as small as 0.1mm are typically used. Having said that, non-contact laser scanning has a clear advantage over probing when measuring extremely soft materials such as fabric and foam. These materials do not require clamping so are not subjected to any stresses. The only solution for probing is to freeze the material, but water expands when frozen, upsetting accuracy.

Laser-based systems can scan very tight corner radii, ie smaller than 0.1mm. The Surveyor 3D laser digitising system, supplied by GKS, emits a Class H beam at less than 1mW. Using triangulation-based technologies, 3D co-ordinates are captured according to scan density and pattern parameters set by the user. These co-ordinates are stored in a file that can be fed directly into Laser Design’s DataSculpt software: a point data editing package which allows users to manipulate data files, or convert them for input into a CAD/CAM system. The Surveyor systems use a line range profiling sensor (RPS 450 and 150 from Geometric Research) which spreads the laser beam into a line and collects data up to 100 times faster than point range sensors. GKS says the inaccuracies associated with mechanical probe offsets, deflection, vector analysis, and probe size or shape are eliminated.

As laser scanners start with a true surface, the equivalent of a zero diameter stylus, there are no errors introduced due to the contact position. Sign errors result when the probe contacts the workplace off centre – a problem with pivoting probes.

Although profiles with concave details finer than 0.1mm can be scanned, they can be a problem to machine. So the cutters needed should be considered before starting. There is little point in scanning at a resolution of 0.005mm if the finest cutter obtainable is 0.1mm – remember, it’s the application that’s important. Using a high resolution scan grid to record a relatively flat surface would be a waste of time. The more points taken, the longer it takes. This is why the driving software is becoming more sophisticated, taking into account the profile it is measuring so that the density of points can be automatically increased over areas of rapid change and reduced over relatively flat sections. This is achieved by using a chordal tolerance, a feature found in most contact scanning systems but which is uncommon in laser systems. The method uses data from the previous scanned strip to control its speed. Surface co-ordinate data used in Renishaw’s Tracecut scanning/CAM software is taken from the stylus tip rather than the ball centre, allowing cutters of the same diameter as the probe tip to be used without compensation.

Deep, near vertical faces are frequently encountered when scanning. While representing no difficulty for a contact system, this situation can cause problems for lasers. For optimum operation the laser beam needs to be nominally at 90 degrees to the unknown surface. This is suitable for patterns with low relief, but for steep vertical faces the laser needs to be reorientated and successive passes made. Additional data must be merged: this takes time as considerable computing power is needed to filter any redundant data, and mis-match errors often result in harsh environments. Most laser scanning and digitising systems use a fixed grid format, defined prior to digitising. This takes the form of specifying a distance between passes (stepover), and data points in the direction of scanning (pitch).

Consider the side view of a cube, for example. A contact scanning system will collect data from the vertical face all the way from the base to the top face of the cube. The stylus radius generated at the base of the resulting shape can be removed by profiling with a standard end mill, or by extending the shape down by slightly more than the radius and machining back. However, if the laser approaches the same vertical face at a constant pitch, at one instant the Z value is at the base of the cube and next – one pitch distance further on – it is on the top face of the cube with no Z information between the two points, ie a large change in Z for a very small movement in X or Y. This will give an error. In order to minimise these errors, very small pitch values are used, giving rise to large volumes of data irrespective of shape complexity.

Both systems are clearly an improvement on the earlier digitising systems, but which is better is a purely academic argument; both systems rely heavily on the software and computing power available. As computing power increases, so will the power of scanning systems. What is important is that the right system is used for the job.

Digitising or scanning an existing object or model can provide you with enough data to create tool paths and machine an exact copy. Here we see how the tool paths to machine the fish are built up from data collected from digitising the original.

LK’s new Micromeasure optical freeform scanner aligns point clouds to produce a true image of the shape. The system can measure 14,000 points/sec and the data is continuously fed into the hard disk of an on-line PC where patch matching capabilities in the software convert co-ordinates into a defined surface.

{{INFORMATIONRENISHAW. Tel: Gloucestershire (01453) 524524.GKS. Tel: Birmingham (0121) 622 3636.DELCAM. Tel: Birmingham (0121) 766 55443D SCANNERS. Tel: London (0171) 922 8822.LK. Tel: Derby (01332) 811349.}}

Martin Oakham did his apprenticeship in tool making & design, has a BSc and an MSc and is currently Technical Editor of Metalworking Production magazine.