The concept of net shape for castings
The concept of net shape for castings
J. Campbell
IRC in Materials, The University of Birmingham, Birmingham B15 2TT, UK
Available online 9 March 2000.
Abstract
A quantitative study of the factors influencing the variability of casting dimensions has been attempted. The variability of casting size limits the achievement of a finished (i.e. net) shape in one operation. The casting processes include lost wax, lost foam, sand casting, gravity die (permanent mould) casting, and high-pressure die casting. Lost wax (the so-called precision casting process) is found to be highly repeatable in dimensions only for small castings, and rapidly loses repeatability as casting size increases, becoming as poor as a low technology sand casting process at sizes of approximately 1 m. The other casting processes are less influenced by casting size. Pressure die casting and high technology sand casting are found to have similar excellent performance with regard to control over casting dimensions over the whole spectrum.
Author Keywords: Castings; Net shape; Accuracy
Article Outline
- 1. Introduction
- 2. The casting processes
- 3. The metals
- 4. The analysis
- 5. The results
- 6. Conclusions
- References
- 4. The analysis
1. Introduction
A ‘net shape’ product is one whose shape requires no further modification, and thus is ready for use. The concept is one, which is usually applied only to parts, which are required to assemble and fit closely with other parts.
Most soft plastic toys are made net shape, but the shape is often not critical, and the concept has no relevance. However, it is critical in the case of components such as Lego, where each injection moulded part is required to fit accurately into other parts without any further processing; clearly, in this case, no machining can be contemplated on grounds of cost. The component must be net shape.
For many metal components, with perhaps the exception of some powder forming routes, the part can rarely be finished exactly to the required final tolerance in a single forming operation. Thus, in general, a forming operation such as forging or casting is carried out to produce a ‘near-net-shape’ product, which is subsequently brought into the required tolerance by a finishing operation of some kind.
The purpose of this study is to evaluate the various casting processes for their capability in the production of net or near-net shape products. Thus, the problem resolves to ‘what accuracy can the various casting processes achieve?’
This is a task not easily limited to a manageable scale. For instance accuracy is required in terms of straightness, flatness, concentricity, etc. For this study only the ability to reproduce a length will be considered.
There are also the well-known twin aspects of errors of length: (i) systematic errors associated with, for instance, the length of the pattern equipment being made incorrectly; and (ii) random errors associated with the lack of reproducibility of the processing steps. It is important to consider both these factors.
Sources of error also in two forms: (i) values, which are not functions of length but apply uniformly to parts or to the whole of the product. For instance, these sources include such features as mould or core assembly errors. In particular this includes the important factor of mis-match between mould halves. Also included in this category are features like the thickness of a mould coating when applied uniformly (well or badly) to the whole or part of a mould by spraying or dipping; and (ii) errors which are due to expansion and contraction during processing, and which are thus a function of the size of the component.
There is also the fundamental difficulty associated with the fact that an intrinsically poor processing route could be carried out with extreme care by one manufacturer, giving a satisfactory product in spite of the technology. This contrasts with the situation where an intrinsically reproducible process is carried out with ineptitude or even crass irresponsibility by another, therefore giving an unacceptable product. Thus, although in this study the fundamental capability of the process will be emphasised, the ability of the manufacturer to overcome deficiencies in the process by intelligent and diligent effort must not be underestimated. Even so, the view is taken here that processes, which are intrinsically reproducible, are to be favoured over those, which are not. This study is based on that premise, although occasional note will be taken of known exceptions. Also, some attempt will be made to place limits on the achievements of both good and bad practice to assess the extent of the tolerance problem.
Because of the paucity of data on the accuracy of castings produced by different processes (notwithstanding the existence of an ISO standard [1. ISO Standard 8062. Castings — system of dimensional tolerances, 1984..1] and one notable experimental attempt [2]) many of the estimates of accuracy have been based merely on the author’s experience. This, clearly, is hardly satisfactory, but is judged to be better than nothing at this stage. The paper is offered as a preliminary study, outlining the concepts involved, and providing a framework in which the mechanics of the problem can be understood and into which better data can be placed as it becomes available.
Clearly, the concepts described here are capable of considerable sophistication. For instance computer studies using such techniques as the Monte Carlo random addition of sources of error to ascertain distributions of dimensions would be valuable to quantify the expected standard deviations of critical dimensions of castings produced by different processes. This would be a welcome advance over this rough quantification.
2. The casting processes
Inaccuracy can be introduced into the product at several stages of the casting process, which typically, usually involves the sequential production of a series of shapes, one formed intimately against the other, thus producing a series of positive and negative forms which finally give the desired positive shape. As the number of steps increases there is an increasing chance to introduce error. The main processes are considered in order of increasing complexity, and by implication, vulnerability to error, below.
2.1. Exterior shapes
Die casting is the most direct of the casting processes. Here, a metal mould (a near-net-negative of the required shape) is filled with liquid metal, which is allowed to solidify to give the final desired positive form.
Pressure die casting involves the injection of the metal at high speed and high pressure into the die cavity, whereas low pressure die casting and gravity die casting use similar cast iron or steel dies which are covered with a protective refractory coating, and which are filled relatively quiescently.
Sand casting involves the additional step of having a positive pattern, which is used to produce a negative sand mould to produce the positive cast form. Two main types of sand mould are used: (i) greensand; and (ii) chemically bonded sand. Although sand has a poor image, and is in fact often used badly and under poor control, the fact is that it has considerable potential for reproducibility. It often does not require a protective mould coat, and, thus, retains the accuracy of the pattern. Because the pattern never contacts the liquid metal it is usually long-lived and suffers little wear, retaining its original accuracy for up to 10 times longer than die casting processes.
Lost wax (i.e. investment) casting involves a further additional step, because a negative die is required, usually accurately machined from an aluminium alloy. This die to form the wax pattern is long-lived and retains its accuracy well. The die is filled with liquid wax to make a positive pattern. A ceramic shell mould is formed around this, and the wax melted out, to produce a negative into which metal is finally poured to create the final positive shape. Significant scope for variability exists in the stresses of the dewaxing operation, in the firing of the shell where sintering and shrinkage of the shell occurs, on preheating prior to casting, the expansion and phase changes which occur on casting (particularly when pouring high temperature alloys and steels), and the variable restraint of the mould on cooling as a result of the prior variations in chemistry and sintering of the ceramic shell.
Lost foam casting is in some ways similar to lost wax, but where the disposable positive pattern is formed from polystyrene expanded inside a machined aluminium die. Again, the die is accurate and is the least subject to errors of wear compared to all the casting processes. The foam pattern may be subject up to 0.8% shrinkage (depending on the type of foam) [3] and may be required to be assembled and glued together from individual foam parts. The whole is then coated with a ceramic-slurry, which is allowed to dry. The assembly is finally placed in a box, loose dry sand is vibrated into place around the pattern to support it during casting, and extract the heat during freezing. This action of pouring and vibrating sand into place in this way usually introduces measurable distortion of the flimsy polystyrene pattern. The metal is poured into the foam, displacing it to form the final positive shape. (The negative shape of the mould is not revealed in the normal operation of this process.)
2.2. Interior shapes
Hollow parts of castings, which can be formed by simple withdrawable shapes, can be introduced into die castings. In this case the core is a permanent feature of the die tooling and is usually made from steel, and is sometimes water-cooled. Pressure die cast engine blocks are among the most ambitious aluminium alloy castings made by this process.
More complex interior shapes cannot be withdrawn, and thus require to be formed by disposable cores. The most common type is the resin-bonded sand core, in which the resin is designed to break down after casting, allowing the sand to flow out of the cored cavity. Alternatively, in a few instances, cores can be dissolved out (such as salt core by water or silica core by hydrofluoric acid).
The core is supported inside the mould cavity by its prints. These are (positive) core extensions, which locate in (negative) print locations in the mould or die. For sand casting the location of the core can be precise because assembly is at room temperature, and because the core-to-mould interface is usually kept clear of uncertainties such as a variable thickness of core coating. For gravity die casting, the sand core is often poorly located because the thermal distortion of the die causes the print negatives to be out of location. In addition, the die has a variable thickness of die coat, which has been sprayed on, some of which finds its way onto the print locations, causing further deterioration of their precision. Worse still, especially if core loading of the die is lengthy or is delayed, the heat of the die can sometimes soften, or even cause to break down, the resin binder in the sand core, causing the core prints to disintegrate and the core to sag.
Once surrounded by liquid metal the core will be subject to buoyancy forces, which will tend to make it float. This is not so bad in liquid magnesium where the density difference with silica sand is near zero. Similar neutral buoyancy applies for the aluminium vs. zircon sand system. Buoyancy becomes increasingly problematic for the common systems such as aluminium vs. silica sand and liquid iron (or other dense metals such as copper and steel) vs. silica. The latter systems more particularly so because of the higher casting temperatures involved. Flotation forces have to be withstood by adequate mechanical support of the core, usually by good print design and location, or, often less desirably, by metallic supports (chaplets) which are cast-in to the product as permanent features. To ensure good fusion between the insert and the casting is not always easy. Sometimes such features are subsequently machined out.
Although these problems related to internal cores are noted in passing here, they have not been incorporated into the formal exercise, which constitutes the basis of this paper. Threats to accuracy from cores are only one of the many aspects, which illustrate the complexity of a formal study of dimensional problems in castings.
2.3. Errors due to expansion and contraction
In greensand moulding systems the sand is bonded automatically by the compaction of the sand. However, at compaction pressures above approximately 1 MPa (10 bar) the sand mould deforms elastically, and, on withdrawing the pattern, is subject to ‘spring back’. This general distortion of the mould leads to numerous difficulties relating to core assembly and mould closure. This, to the author’s knowledge, is the only common system exhibiting distortion due to stress. Most distortions in casting processes arise because of thermal expansion as is discussed below.
When the metal first enters the mould, the mould, and more particularly, the cores, heat up and expand. Thus, the mould cavity enlarges and will probably suffer some distortion. This is reasonably reproducible in sand moulds because the pouring temperature is normally under good control (usually ±10°C or better) and the moulding sand is always close to room temperature. Thus, the starting conditions are usually reproducible. This is less true for such processes as die casting, where the die temperature of perhaps 300°C is often poorly controlled, varying by as much as ±150°C.
As the casting cools it contracts, with the result that some parts of the casting are subject to tensile extension or compression because of geometrical constraint of the mould. The constraints exerted by the disposable sand mould are perhaps less severe than those of the metal die, but in general the casting stays in the sand mould longer and the constraint, therefore, operates for longer.
Although an attempt will be made to estimate these expansions and contractions, the final result corresponds to the Patternmaker’s Contraction Allowance. This is the value based on hundreds of years of experience by patternmakers and toolmakers, and is the allowance they have to provide, making the pattern a little larger than the final casting. Unfortunately the factor is sensitive to geometry of the casting, and mistakes are often made in the correct choice of the allowance, which can vary between 0 and 1.3% for aluminium castings and up to 2.4% for steel castings. Some recent work in the author’s laboratory has highlighted the uncertainties of this factor for cast aluminium and cast irons [4 and 5]. An earlier attempt by the author to estimate the allowance based on the concept of the constraint envelope has proved successful for regular shaped castings such as automotive cylinder heads and blocks, and has given reasonable accuracy for thin-walled aerospace products [6]. These semi-empirical approaches should be capable of considerable further refinement. They are not used here, however, where the current exercise is to derive the final result by adding the numerous individual contributions to ascertain the total uncertainty in the size of the finished product.
In the pressure die casting process, the casting is normally thin-walled, and, thus, rather weak. As it cools in the rigid steel die, contracting onto projections of the geometry, it is, therefore, forced to stretch plastically. The size of the casting is, thus, mainly controlled by the time to ejection.
The hot casting is cooled to room temperature in a variety of ways. This may occur as a series of individual castings on a cooling conveyor, as a heap in a bin, or by an immediate quench into water. These post-casting operations are likely to create different patterns of distortion. Likewise, heat treatment, even natural ageing, affects many products particularly heat treatable aluminium alloys. These grow slightly (of the order of 0.05%, i.e. 0.5 mm per metre) as hardening precipitates form. The post-casting operations are not included in this study.
3. The metals
Lost wax moulds (ceramic shells made by the investment technique) are unusually versatile, being capable of containing and shaping nearly all liquid metals. The discussion below centres on the other processes, which are specific to or have to be tailored to individual metals.
Zinc castings are most commonly supplied as pressure die castings. The thermal distortions of both die and casting are the least of any of the casting processes because: (i) the casting temperatures are low; (ii) the dies are constructed of steel and use no protective die coating; and (iii) the dies are supported in a close fitting and rigid steel bolster. Thus, zinc pressure die castings are intrinsically capable of meeting many net shape requirements.
The clear-cut case of near-automatic achievement of net shape by zinc pressure die castings allows us to simplify this study, and not consider such castings further.
The main cast metals considered are, therefore, the light alloys, cast irons and steels. Copper based alloys represent a smaller market, and are not separately considered. They fall into a class of accuracy close to that of cast iron as a result of the similar casting temperatures and similar moulds (although it is to be noted that many plumbing fittings — for domestic showers for instance — are cast into cast iron dies using a graphitic die coat applied by dipping).
The light alloys based on magnesium and aluminium can be cast in moulds of all types. For pressure die castings the dies are steel and no protective mould coat is used, so maximising accuracy. For low pressure and gravity die castings the die is cast iron or steel but is protected by a ceramic die coat. The thickness of the coating is not easily controlled, thus, limiting accuracy. The dies are also often subject to gross distortion partly because the die is free-standing (i.e. not supported by a surrounding steel bolster as in the case of dies for pressure die casting). For sand castings the moulds can be used without a protective coating, and thus retain their dimensions.
Cast iron is most commonly poured into greensand moulds with no protective coating. Chemically bonded sand moulds on the other hand usually require a refractory wash coat to obtain an acceptable surface finish for cast iron. Although cast iron dies are possible for cast iron, their use in western Europe is limited to specialised casting operations, where, again, a die coat is required. The process is more widely used in eastern Europe.
Steel is cast into sand moulds. For modest sizes of products greensand is widely used without a mould coat. Increasingly often, however, steel is being cast into chemically-bonded sand moulds with a protective mould coat to achieve an acceptable finish. The high temperature of steel casting leads to considerable interaction of the surface of the casting with its environment, leading to sand burn-on and oxidation, etc. Because of these interactions, or the application of mould coatings to prevent them, steel castings are normally the furthest removed from the net shape concept. For steel components, weighing up to a few kilograms, the favoured route for near net shape is, therefore, lost wax casting in vacuum.
4. The analysis
Table 1 lists the main factors, which control the dimensional variability of castings.
Table 1.
Full-size table (36K)
For simplicity the highest and lowest factors only are summed to give the most pessimistic and most optimistic estimate of accuracy. In addition to the total variation which a factor may introduce (the ‘potential’ inaccuracy figure) the most probable inaccuracy figure (the ‘probable’ figure is probably close to the standard deviations). As an example for sand castings using an accurate and stable steel pattern with excellent matching of the mould halves, this best case for an aluminium alloy casting the total potential error is ±0.10% of its length, plus ±0.25 mm error which is independent of length. This works out to be a grand total of a maximum variability in a metre long casting of ±0.13% corresponding to ±1.25 mm. A similar summation for the probable variability of a good sand casting technique to make a metre long casting yields ±0.04 % corresponding to ±0.40 mm.
5. The results
The factors are summed to give the total variability for each of the casting processes for castings of length 10–1000 mm (Fig. 1 and Fig. 2).
Full-size image (8K)
Fig. 1. Graph of casting length (mm) vs. potential variability (mm).
Full-size image (8K)
Fig. 2. Graph of casting length (mm) vs. probable variability (mm).
It is clear that the variability of process such as gravity die casting are dominated by factors which are not functions of length such as die coating. This is in contrast with processes such as lost foam, and more particularly, lost wax, whose variability increases with casting size, reflecting the importance of thermal expansion problems in this process.
Thus, lost wax can retain its common name ‘precision casting process’ only for very small castings. For large castings the process becomes no better than low technology sand casting.
The processes which shine out as having intrinsically repeatable dimensions are pressure die casting and high technology sand casting. There are good reasons for this. Pressure die casting is a simple, direct process, operated in a rigidly supported steel die, and which uses no die coat. Sand casting involves more steps, but all the steps are carried out at room temperature so that no significant expansion errors accumulate.
The position of gravity die being systematically below that of the variability of low technology sand casting was one of the major reasons for the historical choice of ‘die casting’ as opposed to sand casting for many automotive applications. Now, interestingly, recent advances in sand moulding, both in greensand and chemically-bonded sands, have overtaken the accuracy possible in gravity die. Even so, the poor image of sand casting lives on with the engineering profession, whereas ‘die casting’ has the image of cleanness and precision!
From the figures the increase in accuracy for a 500-mm long aluminium alloy casting when changing from gravity die to a precision sand process is a factor of potentially 2 but probably up to approximately 4. This is roughly in line with measurements carried out on production runs of cylinder heads by Ford of America who found on average that an accurate sand process yielded castings more than twice as accurate as their standard supplies of gravity die castings. This was a critical finding, which led to the adoption of the Cosworth Process in North America. The slight difference to be noted in the predictions of a factor between 2 and 4, and the finding of 2 in the above account may relate to the influence of cores. The analysis in this study relates only to exterior shapes, for castings containing a large volume of cores, and which are therefore substantially hollow, the results are expected to be modified to some degree.
Another cautionary note relates to the simplification of summing all the favourable features. Thus, the high technology sand prediction assumes its use for the casting of aluminium (or possibly magnesium). If cast iron were to be cast in such a mould there would be a less favourable total of variabilities, with the result that the curve for cast iron would be expected to lie somewhere intermediate between the upper and lower limits for sand casting. The ‘low tech’ sand curve assumes the casting of cast iron, as one of the features which limits reproducibility.
Finally, it is to be noted that cast products are not bought simply for their dimensional reproducibility. Surface finish, internal integrity and many other factors, not least cost, are important. However, this study has been aimed at clarifying and quantifying one aspect, the ability of the process to achieve a level of dimensional control.
6. Conclusions
Pressure die casting and high technology sand casting are the processes with the greatest capability for reproducibility of dimensions. The lost wax process is accurate for small components, but as the casting size increases it becomes rapidly poorer, eventually becoming as bad as low technology sand casting at large sizes. Gravity die casting and lost foam casting show intermediate performance.
References
1. ISO Standard 8062. Castings — system of dimensional tolerances, 1984..
2. IBF Technical Subcommittee TS71, 1979;72:46–52..
3. Brown JR. Metals and materials, 1992:550–555.
4. B.B. Nyichomba and J. Campbell, Linear contraction and residual stress of aluminium alloy sand castings. Int J Cast Met Res 11 3 (1998), pp. 163–177. View Record in Scopus | Cited By in Scopus (5)
5. B.B. Nyichomba, I.M. Cheya and J. Campbell, Linear contraction of ductile iron castings. Int J Cast Met Res 11 3 (1998), pp. 179–186. View Record in Scopus | Cited By in Scopus (2)
6. J. Campbell. In: Castings Butterworth Heinemann, Oxford (1991).
