Design for Casting


Although this was written many years ago, the techniques described are still relevant now. In fact a new generation of designers have graduated from university knowing less about casting processes than their predecessors 20 years ago.



Paper Presented by Steve Prentice at "Solid Modelling 2000 Conference"

on Wed 8th March 2000 at National Motorcycle Museum, UK. 


How to reduce the time and money you spend with your pattern maker


Solid models are terrific at giving visual feedback on your design and now your CAD models will link direct to analysis software and manufacturing processes. One of the most obvious of the manufacturing links is "Rapid Tooling" for sand and die-casting, where tooling for dies and sand cores are cut or formed direct from CAD data. The techniques described here extend beyond casting, as forging and plastic moulding can be treated the same way.


There are considerable productivity gains, quality improvements and cost benefits to be made with rapid tooling but to benefit from this your design must be complete. Which means that on your model split lines must be identified, the part must be modelled with draft on all relevant surfaces, internal cores must be separated into another file and all edge radii and significant fillet radii must be present.


If a model is rapid tooling ready the pattern maker can take the data and cut the pattern without changing or interpreting any surface or feature that you have modelled. The castings that you get back should then perfectly r eproduce your CAD design model (which you might or might not view as a benefit).


So why are less than 10% of CAD models that are presented to pattern makers genuinely ready for rapid tooling? It also appears that this problem gets worse as models get more complex. Draft and split lines are rarely applied, is it that they are not understood? Cores usually have to be extracted from a monolithic model and reassembled as surfaces by the pattern maker. Radii are often applied to make the model look complete, unfortunately these can sometimes interfere with the pattern makers efforts to apply draft surfaces. I don't accept that this is because of software limitations, or because of designers' abilities. I see it as a problem of design management.


Design managers are not asking engineers to work with draft, in fact some managers have a policy of working without draft. At least they do have a policy, but it's one that will cost them on price and quality. Do you have a policy on casting design? Would you like to formulate one while you read the rest of this article?


Traditionally pattern makers hand carved wooden patterns from drawings. Information on the drawings was never totally definitive so the pattern maker used his license to interpret the design as he applied draft and radii. Manual pattern makers are still being kept busy with designs from 2D CAD and curiously with 2D drawings produced from 3D CAD systems. It seems strange to solid model components when the only output is to be a 2D drawing, but having done this myself very recently I accept there are sometimes reasons for it (well, no one's perfect).


3D CAD pattern making uses a CAD model to create a cutter tool path that machines the pattern from a block of resin. We all imagine that when we give 3D CAD data to the pattern maker that is rapid tooling in action. But as I've said before in over 90% of cases the pattern maker needs to make changes to the design model, when he does this it shouldn't be called rapid tooling.


There are 2 CAD methods that the pattern maker can use to modify the design model data. The most popular is to use a surfacing CAD system (e.g. NC Graphics or Delcam), to import the design model as surfaces and attach or hinge surfaces at the draft angle to create the "surface CAD pattern". This leaves large gaps between these draft surfaces and the rest of the model, provided the gaps are less than the radius of the cutter used to form fillet radii (typically 3 or 4 mm) the surface CAD pattern can be used to programme the cutter path. These disjointed surfaces will prevent the surface CAD pattern from being returned usefully into the design CAD system. The design CAD model will then be a different shape to cast components in the real world, which must impact on total quality of your design effort.


To overcome this problem of one-way data transfer some pattern makers have started using native CAD systems to add draft features. This is often at the request of a dominant customer who insists on modelling without draft. The pattern maker uses the same CAD system that the model was designed in, so he can edit the design model to add draft features to create the "native CAD pattern". When complete this can be returned to the customers design department as the new design model, it is still an editable solid model, it exactly replicates real world castings and has added value to the customers design knowledge, but this approach has its drawbacks. The pattern maker has to purchase the CAD system, train his staff and could have to invest in 3 or 4 different systems to suit different customers. This would be enormously expensive as this approach is only being requested by users of high-end CAD systems. Editing the design model to add draft and split lines will be a considerable task, even impossible on some models as it relies on features having been created from an understanding of the split lines. It is easier in these cases for the pattern maker to start again on a new model and work with draft from the split lines. Obviously the native CAD pattern method is going to be expensive, but it has advantages in quality and design knowledge.


Preparing your own models for rapid tooling is the method to maximise the value of your design knowledge and retain total ownership of the design process. Your design models will accurately represent your real world castings and the data will be exactly what the pattern maker wants. It should be the quickest and cheapest route from start of design to delivery of castings. To achieve this all you need to do is understand split lines and model with draft. Talk to your pattern maker before you start, show him a simple sketch of what you want, discuss split lines and draw directions, then start modelling from the split line and build with draft as you go. You may be spending more time and money in your own design department but you will be spending less with your pattern maker and this fits well with the business philosophy of maximising your added value.


This method of working with draft may come easily to the older generation of engineers who have wide experience of products and processes, but the younger generation with apparently better CAD skills may have difficulty due to lack of experience of working with pattern makers in the workshop. We have found that CAD models that are modelled with draft, by an engineer with an understanding of pattern making, can work out much simpler (fewer features, smaller file size, more elegant design) than if the CAD model were built without draft, by an engineer lacking in experience. The illustrated examples show how we have built models with every feature present in very short design times.



The motorcycle crankcase looks complex, but as a casting each half is a simple 2-direction draw. Our customer had a part finished 2D design, his pattern maker had quoted 12 weeks lead time for castings from manual patterns and £27000 tooling charge. Our project objective was to accelerate the lead time for cast prototypes. The upper crankcase was modelled as cast and machined configurations in 120 man hours, rapid tooling ready data was delivered to the pattern maker in 8 days from the project start and castings were received from the foundry 3 weeks after that. The lower crankcase was similarly modelled in 140 man hours, rapid tooling ready data was delivered in 10 days from the project start and castings were received 4 weeks later. Our customer was willing to pay extra to accelerate the casting lead times, but he didn't have to - the actual tooling charge was £7500.


Each of these example castings has been modelled with separate CAD models for cores or internal shapes; this is crucial to our method of working to simplify CAD models and separate the cores, as pattern makers want. We start with the pattern model, this is the external shape of the casting, it has no internal features or machining. We then take an independent copy of this model to create the core model, which is the internal shape of the casting. Wall thickness is formed by editing features of the core model to make it smaller. The pattern and core models are placed in an assembly and the core is subtracted from the pattern giving the as cast CAD model to which machining features are added to complete the part. The pattern maker is given IGES data of the pattern and core models that he uses directly for tool path generation, we also supply data of the machined part for him to check machining allowances.




The V6 cylinder block shows how this technique of modelling the sand cores separately gives us real benefits in easing the modelling task. Each of the cores is built as a separate model and subtracted from the pattern model in an assembly. The water jackets and bore cores are identical on both banks, so each has only been modelled once and loaded twice into the assembly to form cavities on both banks. The 3 crank space cores are similar but slightly different, so the centre core was modelled first, then independent copies made, which were edited for the different shapes at front and rear. Editing core and pattern models is simple, regeneration of models is fast as file sizes are small. A design team can work on the block, with each engineer designing a different core. This is concurrent design. The block is shown at rough machining stage, total design time for cast and machined models is 220 man hours.


When working on core models without a lot of experience it is not always obvious where the split line should be. The split line illustration shows 4 acceptable options of where to place a split line and how to work draft and radii from it. If you are in any doubt about split lines consult your pattern maker before you start. In general remember to first identify where to place the split line, start modelling from there, always model with draft as you go, always model with edge radii, you may omit fillet radii if you must as they will be added by default by the cutter tool radius.


Cylinder heads are usually the most complex parts that engine designers have to work with. Wall-by-wall construction is a popular technique of building a monolithic cylinder head model. It appears at first to be very simple, it could be suitable for a concept design study, but it should not be taken forward as a method for production intent design. The monolithic model results in massive file sizes (over 120 MB is common), slow regeneration times and it limits design work to one engineer at a time. Building a model like this with draft is enormously difficult and adding later split lines, draft and radii in-situ is almost impossible. The core data can be extracted from the model using a surfacing package that can also add draft and radii to form a surface CAD pattern but then the design CAD model is going to be different from real world castings.




This in-line-3 cylinder head model has been constructed from pattern and core models, it is complete with all machining, separated cores, split lines, draft and radii. The design CAD model will be identical to real world castings. Design and modelling has taken only 400 man hours. The largest file size is 35 MB. A team of 7 could work concurrently on this head, each designing a separate core. The core assembly can be used as a digital mock-up checking clearances and fits as the cores come together. The close-up view of the water jacket core shows that it is a one-piece sand core with 3 draw directions, these are coloured in the illustration to help identification, but the CAD model of the jacket has been modelled in 3 pieces. These are almost identical , so like the cylinder block crank space cores, the centre core was modelled first, then independent copies made, which were edited for the different shapes at front and rear. The same method was used for the oil space core (under the tappets) and the camshaft core (above the tappets). The intake and exhaust ports appear simple but are actually varied irregular sections lofted along a complex compound spline using further complex splines as guide curves. We are confident of applying this method to any port shape imaginable.


We design using "mid-range" solid modelling tools only, without needing a surfacing package to create complex shapes or manipulate data for the pattern maker.

So have you been thinking about a design policy for castings? Does it include these ideas?


  • Maximise added value of design knowledge

  • Employ engineers who understand what they are designing

  • Consult with pattern makers on split lines before you start modelling

  • Model cores separately

  • Design with draft

  • Start CAD modelling from the split line


Do you want to create CAD models that are helpful to the foundry?

Or do you enjoy spending too much time and money with your pattern maker?

Print Print | Sitemap
© Steve Prentice Design Limited