Скачать 46.63 Kb.
An Object-Oriented Support Tool for the Design of Casting Procedures.|
Brian Knight, Don Cowell, Keith Preddy
The design of casting system for a given shape depends to a great deal upon the skill of the engineer, who must make a number of decisions. Of prime importance in the formation of these decisions is the construction of the so-called modulus model of the shape, which gives a crude approximation to the dynamic cooling and freezing order of parts of the shape. Several computer based modulus models as support tools for the engineer have been attempted, usually based on the superposition of a regular grid of cells on the shape, and iterative calculation of freezing cell by cell. In this paper, an alternative approach is described which more closely models the engineer's expertise. There are two main elements to the expertise: First, the systematic breakdown of a casting into elements with known casting properties, and the suitability for casting based on the modulus model applied to this assembly of elements. Second, extensive knowledge of past cases (both sound and otherwise) of castings. There are advantages in an approach that models this engineering expertise, and which allows for a higher degree of interaction with, and early feedback to the engineer. The system is naturally object oriented, based on traditional classifications of shape elements which have evolved over many years in casting design. This classification also provides natural keys for a case based reasoning system. Experience with a prototype system is described, and test results for a limited set of cases are discussed.
Keywords: Metal Casting Design, Expert Systems, Object Oriented, Case Based Reasoning.
The successful casting of a predesigned shape is vitally dependent upon the skill and experience of the foundry engineer, who performs a task known as methoding the design, in order to determine a number of key elements prior to manufacture. This task is concerned with the design of the casting process, i.e. filling the mould with molten metal, and the subsequent freezing of the metal. The foundry engineer must consider a variety of problems associated with this process, the most important consideration being the shrinkage which occurs as the metal freezes. Since it freezes first at the boundaries, there is a possibility that isolated pockets of molten liquid will form during freezing. Subsequent shrinkage of these pockets will give rise to porosity and other casting defects. In order to ensure that no such pockets will form at any stage, the engineer can place feeders and chills at strategic points in the mould. Feeders are reservoirs which can supply molten metal to elements of the shape as shrinkage occurs. These must freeze after the fed element and have sufficient feeding capacity as the casting shrinks. Chills are heat absorbing blocks embedded in the mould which can force parts of the shape to freeze more quickly. In addition, the design of the running system (the orientation of the shape being cast, and the positioning and sizing of the molten metal sumps and conduits for filling the mould) has an important effect on the cast, in view of the fact that freezing begins during metal filling, so that blockages can occur at any time.
Figure 1 shows the various stages of the (computer assisted) methoding process in diagrammatic form.
As can be seen, the process is an iterative one, in which tentative designs will be evaluated using various software tools. Initially there will usually be a dialogue with the client, where possible re-design of the shape for more efficient casting may be suggested (process 1). The foundryman will then decide on the orientation of the mould during filling, and design the positions and sizes of feeders and chills (process 2). For this task he/she will rely mostly on experience, together with hand calculations of the modulus of elements of the shape. The modulus of a element is the ratio of its volume to cooling surface area, modified by empirical shape specific correction factors (see Wlodawer [WLO 67]), which can give an indication of the time taken for the element to cool. The design is then evaluated against various simulation models (processes 3-6), and modifications made until the simulations are satisfactory.
Several software tools may be used to assist the methoding process. For the initial stages of methoding these tools need to be fast and easy to use: simple models based on the cooling modulus principle, or fast empirical mould-filling models. Amongst these are: CRUSADER [CRU], FEEDERCALC [FEE], and SOLSTAR [SOL], which support the preliminary design stages, and slower, more detailed numerical models such as MAGMASOFT [MAG] and SIMULOR [SIM], which support the simulation stages. CRUSADER and FEEDERCALC give numerical support on such aspects as feeder sizes and feeder-feeder distances, but do not attempt to give experiential advice on such elements as re-design for casting, or mould orientation. SOLSTAR is often used as a fast solidification model, which can check a given design or give information on feeder positioning. More advanced numerical software, using computational fluid dynamics techniques, is currently under development [CRO 89(More recent?)] but is as yet unproven.
Foundries represent an important sector of UK manufacturing industry, and there is a large benefit to be obtained by facilitating good methoding practice in terms of the reduction in reject rates, and in terms of the improvement of casting quality. The industrial problem is the provision of a system which a foundry engineer will find useful during the methoding task, and which will lead to improved casting procedures. In 1987 an audit group, commissioned by the Department of Trade and Industry, recommended a UK initiative involving companies, research organisations and universities in the field of computer aided design for casting [CRO 87]. The bulk of the research which has stemmed from this initiative has been concerned with the development of faster and more accurate numerical routines for the simulation of the complex physical processes of flow, freezing and stress involved in cast formation. These may be used in the later stages of the methoding process, to verify a given method as accurately as possible. Predictions at this stage may require re-entry to earlier stages, to adjust the given method.
There is active interest in the casting industry in the development of expert systems for the design of methoding systems. Natarajan et al [NAT 89] describe a system which uses a simple geometrical solidification model to make a preliminary assessment of castability for a limited class of shapes. This is based on the cooling modulus model, but it operates on a rectangular grid of cells, rather than on whole features - rather like SOLSTAR. This geometrical model is augmented by a small set of rules and meta-rules written in LISP. Sillen [SIL 91] describes an expert system that uses rule induction for the prediction of casting defects; this has been successfully used as an adaptive process control system for green sand casting. Sirilertworakul et al [SIR 93] describe a similar rule-based system using Turbo Prolog, with a knowledge base that facilitates the choice of alloy and casting method, given a element design and specification. Firth and Nealon [FIR 87] report on an expert process planning system developed in Prolog, for the casting of acrylic momomers encapsulating an embedment.
Other AI research has focused on geometrical feature extraction prior to such assessments. Luby et al, [LUB 88] approach this problem by defining a shape grammar, allowing the creation of designs by the use of a vocabulary of familiar geometric features. The design is then evaluated for manufacturability by the construction of the modulus model from the features. Woodward and Corbett, [WOO 90], have taken a similar approach, concentrating on design rules for aluminium alloy die casting. Chung et al, [CHU 90] describe two applications for feature-based modelling combined with geometric reasoning, one application area being a 'critic' for predicting potential defects in gating designs for investment casting (the 'lost wax' process) which suggests changes to the casting design engineer. Feature description is becoming recognised as a strong candidate for a single data representation for design, design analysis and manufacturing planning in the general context of total computer integrated manufacturing - see the survey article by Case and Gao [CAS 93].
The central idea of the work described in this paper is that the expertise of the foundry engineer is based upon a special decomposition of the shape to be cast into elements. Whereas other approaches to design support tools, e.g. [NAT89], have allowed an arbitrary decomposition, followed by a numerical estimation of the element moduli, the foundry engineer uses familiar elements with well defined moduli. Traditionally, methoding manuals such as [WLO 67] classify elements under this scheme, and provide empirical tables and modulus calculations for each classification. Hence, on grounds of user acceptance alone, there are advantages in the construction of a modulus model which emulates the expert process. However, there are other advantages to be gained from this approach, in view of the fact that much other experiential knowledge of castability has grown up around such decompositions. For example, the concept of effective modulus, which differs from the vol/cooling area by the inclusion of corrections based on pragmatic considerations such as radiation from nearby surfaces, can be keyed to the element classes. Also, knowledge of re-design is often keyed to this element classification. For example, experience shows that Y junctions often give rise to problems which may be avoided by re-design as a T together with a bend.
This article describes the principles on which this decomposition into elements is based, and its realisation as a prototype object-oriented support system at the University of Greenwich. The uses of the system as a design support system are described, as a modulus model, a design advice system, and a case based reasoning tool.
2. Principles of shape decomposition.
Feature-based computer-aided design relies upon the existence of a library of elements which may be assembled and merged together to form a desired total shape. The method engineer must then consider the shape, and decide on its potential castability. The most important technique used for determining castability is the calculation of the cooling modulus of the various elements of the shape, so that their freezing order may be determined. Ideally, each element selected by the user during feature based design of a shape should trigger an appropriate modulus calculation so that the modulus of each part is determined as the user builds the system. A system like this will give the user immediate feedback on the castability of the new shape.
However, the library of elements for such a system cannot consist of arbitrary geometrical shapes, for a number of reasons.
First, for an arbitrary element the modulus calculation depends upon the type, dimensions and orientation of other shapes joined to it. This is because there are significant second order effects due to heat radiation from nearby elements. For example, consider two plates joined in a T, as in Figure 2.
There is in fact slower cooling at the joint due to the radiation shown. The joint will be a hot spot, which will freeze in isolation in the casting process. This is essential information to the methoding engineer, who will take suitable remedial action - placing a chill or feeding reservoir there, or possibly modifying the shape of the joint. If we simply calculate the modulus of each plate separately, we get no indication of this important effect.
In fact, a skilled methoding engineer will not decompose the T shape in this way. Rather, he will adopt a breakdown as shown in Figure 3. The essential difference here is that there are no significant second order effects to be taken into account. The modulus of the four separate parts may be calculated from simple standard rules. Of course, these rules must take account that each element is joined to others at certain places, but the essential thing is that it doesn't matter what other compelementapes it is joined to, merely the fact that it is joined at various constrained places. Hence the engineer is able to calculate the modulus of each element according to his knowledge of the element. This is traditionally done by a diverse set of rules-of-thumb, tables, graphs and formulae (qv Wlodawer [WLO 67]).
A second constraint on the library of elements is determined by the engineer's knowledge of the casting process. For example, the T joint shown in Figure 3 is not physically achievable, nor indeed desirable. The mould itself cannot achieve the sharp join between the plates which is shown in the central element, and if it could be achieved it would form a point of weakness in the casting. The engineer would naturally think in terms of adding fillets to smooth the discontinuity, and avoid tearing. This is best captured by removing such 'bad' elements from the database. Instead, experience shows that the central element shown in Figure 4, which includes fillets, is more practical.
The basic knowledge for such a CAD system is readily available in the large quantity of published material on element design, representing the collected experience of casting engineers over many years; see for example [CD62]. Such knowledge is not made available in current CAD systems; the novelty of the approach described here is that it incorporates this expertise at the basic element level; in effect, we are proposing a "shape grammar" which precludes the possibility of the design of an inherently uncastable shape.
3. The prototype system
A prototype system based on a PC based CAD package has been constructed according to the principles laid out above. Figure 5 shows the view presented to the designer of a section of a solid shape - in this case a flanged hansing cover - to be cast. Shapes are constructed largely from a library of basic elements. At any stage in the design, modulus calculations can be triggered and the freezing order of the elements displayed using a colour coding. In figure 5 colours are represented by grey scales, the lighter the shade the smaller the modulus. For this example, we see that the order of freezing is: .. In this way the engineer can obtain rapid feedback on potential problem areas that can be tackled either by redefining the basic geometry, or by a suitable orientation of the mould and its gating and running system, and the location of suitable feeders and/or chills.
For this prototype system, a limited library of basic elements - bars (or plates), L junctions, T junctions and X junctions - has been built. In principle, other less common elements could be created (and saved) by the engineer; for these the geometric modulus could be used. Two-dimensional shapes may be constructed from these elements, together with corresponding three dimensional extrusions and shapes with rotational symmetry, such as cylinders, wheels, flanges, etc. The basic elements have a uniform representation, namely
Element(Identifier,Shape_Id, Type, Orientation, Geometry,
Identifier is a symbol uniquely identifying a given element.
Shape_id is a symbol uniquely identifying the whole shape
Type is one of: Bar, Ljunc, Tjunc, Xjunc.
Orientation is a left/right/up/down convention needed for L and T junctions.
Geometry is a list of parameters specifying the element geometry. As examples: for a bar, [Thickness, Length]; for an L or T junction: [Base thickness, Arm thickness, Fillet radius].
Connections is an ordered list of objects attached at specified connection points ( e.g. the two ends of a bar, the three ends of a T etc.). These lists completely specify the element network.
Status is a list of symbolic problem attributes for an object, e.g. fillet too small, fillet too large, thin arm, thin base, feeding distance exceeded, very short bar. In fact, these attributes are redundant in the sense that they may be calculated by object methods from other attributes.
The prototype has been evaluated as a support tool providing three main functions:
1) as a modulus tool, which shows graphically a traditional decomposition of a shape, together with the freezing order of the elements. For this evaluation, a number of standard case studies from the literature [WLO 67, MOD 80] were used to test the system.
An example of the output is shown in figure 5. The system provides a coloured display of the elements, with a range of shades from red to blue representing high to low modulus. In figure 5 the modulus is shown as grey scales, dark representing high modulus. From this display, the casting designer is able to identify the direction of solidification of the hansing cover, locate "hot spots" - zones of delayed freezing - and make decisions on moulding direction, the placement of feeding reservoirs and the use of metal and mouldable chills for artificial end zone generation and possible section modulus reduction.
2) as an advice system which provides advice on possible casting problems, and re-design advice. In order to evaluate this function, a number of examples were constructed which were known to give poor castings without modifications. An example is shown in figure 6. This shows a wheel with a hub and rim connected by a thin plate. However, experience shows that the thin plate cannot be fed satisfactorily from feeders at the rim and hub. Freezing will always occur at random over the thin plate, leading to porosity problems, and the casting will be faulty.
Advice for this example would be:
• Consider re-design by thickening the plate or tapering from hub to rim. ( Often the client has no objection to modifications which improve castability. )
• Place chills in centre of plate to reduce the effective feeding distance from hub to rim.
• Consult Case Base to examine other solutions.
This advice is triggered by patterns in the element objects under use. In this example, the key patterns are 1:
a) a T junction with one arm much thinner than the other.
b) a bar (plate) has a linear dimension greater than it's so-called `feeding distance' ( the distance over which it can be fed from either end)
These patterns are represented in the system as statuses for the element objects. For example, a) and b) are represented by:
The statuses are generated by element object methods. The rule system which triggers advice operates on the element statuses and element connectivity, e.g:
3) as a case-based reasoning system, which retrieves previously stored cases which might be appropriate to the current design. The important factors for the design engineer are related to:
Overall shape; The engineer would first need to know whether a shape like this has been cast before. The most important attribute for this kind of retrieval is the general classification of the whole shape according to existing practice. E.g. wheel/car wheel/racing car wheel. Of lesser importance is: material,size,quality.
Element problems; The engineer might want to know of all previous castings where a given problem has been solved. Problems are connected to element states and connectivity, as the above example of the plate exceeding its feeding distance connected to a thin arm of a junction.
The prototype approach to case based reasoning is via a database of cases gleaned from standard literature [WLO67] and from expert knowledge, which are constructed from the basic element library. The engineer is allowed to browse the database, using queries relating to both whole shape and to problems wih individual elements. For example, the engineer can specify:
"show me all racing car wheels with the thin plate problem"., by means of the database query:
Shape ( ?s, racing_car_wheel),
The design support tool described in this paper attempts to encapsulate experiential knowledge of a very diverse nature. An experienced engineer is able to call on empirical knowledge of elements, and of techniques to evaluate cooling properties; on previous experience of the castability of whole shapes and of element configurations within whole shapes. In addition the engineer can analyse a 3D shape into constituent elements in a significant manner.
The paper proposes an object oriented design by feature approach which is capable of representing this knowledge. A prototype has been constructed and used to test out the three main functions of the system: a graphic modulus model tool, an advice giving expert system, and a database used for case based reasoning. The prototype has proved adequate functionally with respect to a limited subset of shapes.
Future work is planned to enlarge the library of element types, automatically identify directional solidification, advise on mould orientation and the placement of feeders and chills and to extend the capability of the system to allow for plastic deformation of elements.
[ADA 92] Adams D., Butlin G., Higginbotham G., Katgerman L., Hills, A. W. D., Charles J. A., September 1992, Modelling in casting development in the UK, Metals and Materials, pp. 496-500.
[BAR 89] Barletta R., Hennessy D, 1989, Case Adaptation in Autoclave Layout Design, in Proceedings of the 1989 DARPA Case-Based Reasoning Workshop, pp. 203-207.
[BAR 91] Barletta R., August 1991. An introduction to case-based reasoning, AI Expert, pp 42-49.
[CAM 91] Campbell J., 1991, Castings, Butterworth-Heinemann.
[CAS 93] Case K., Gao J., 1993, Feature technology: an overview, Int. J. Computer Integrated Manufacturing, vol 6 nos 1 & 2, pp 2-12.
[CHU 90] Chung J. C. H., Patel D. R., Cook R. L., Simmons M. K., 1990, Feature based modeling for mechanical design, Computers and Graphics, vol 14 no 2. pp 189-199.
[CRO 87] Cross M., Ashton M. C., Steiner J., 1987, Computer aided engineering technology for the casting process, Report to the DTI.
[CRO 89] Cross M., 1989, Solidification - An Overview, IBF Conference on Light Alloy Casting.
[FIR 87] Firth P., Nealon R., 1987, Process Planning: An Expert Systems Perspective, in Expert Systems - Theory & Applications, IASTED International Conference, Geneva, Switzerland pp.170-174.
[HIL 91] Hill J. L., Berry J. T., 1991, Geometric Feature Extraction for Knowledge based design of rigging systems for light alloy castings, in Modelling of casting, welding and solidification processes. Rappaz, Ozgu, Mahin ( Eds ), The Minerals, Metals and Materials Society.
[LUB 88] Luby S. C., Dixon J. R., Simmons M. K., 1988, Designing with features: Creating and using a features database for evaluation of manufacturability for casting, ASME Computer Review, pp 285-292.
[MOT 93] Mott S., 1993, Case-Based Reasoning: Market, Applications, and Fit With Other Technologies Expert Systems With Applications, Vol. 6. pp 97-104.
[NAT 89] Natarajan R., Chu C. N., Kashyap R. L., 1989, An integrated environment for intelligent design of castings, in Expert Systems Applications in Materials Processing and Manufacturing, Ed. Demeri M Y, The Minerals, Metals and Materials Society.
[O'C 92] O'Connor L. J., Lan M. S., Partridge D. R., Lee J. M. F., 1992, A Case-Based Reasoning Approach to Automated Weld-Process Design, Applied Artificial Intelligence vol. 6 pp 315-330.
[RIE 89] Riesbeck C. K., Schank R. C., 1989, Inside Case-Based Reasoning, Lawrence Erlbaum Associates, New York.
[SIL 91] Sillen R., Using Artificial Intelligence in the Foundry, Modern Casting, December 1991.
[SIR93] Sirilertworakul, N., Webster, P. D., Dean, T. A., A Knowledge Base for Alloy and Process Selection for Casting, Int. J. Mach. Tools Manufact. Vol 33 no. 3, pp 401-416, 1993
[WLO 67] Wlodawer R., 1967, Directional Solidification of Steel Castings, Pergamon.
[WOO 90] Woodward J. A. J., Corbett J., 1990, An Expert System to assist the design for manufacture of die cast components, Engineering Designer.
[MoD 80] Design and Manufacture of Nickel-Aluminium-Bronze Castings, 1980, MoD Ship Dept, 2nd Edition.
[CD 62] The Casting Design Handbook, 1962, American Society for Metals.
[CRU] CRUSADER software, SCRATA, Sheffield.
[SOL] SOLSTAR, FOSECO Ltd, Tamworth
[FEE] FEEDERCALC, FOSECO Ltd, Tamworth
1 The patterns are expressed in terms of 2D equivalents of the 3D objects with rotational symmetry. In fact, the results apply equally to the 2D shape or the 3D shape generated by rotation.
|Design Methods and Analysis of Algorithms 4 csm102 Object Oriented Programming through java||Applying uml and Patterns: An Introduction to Object-Oriented Analysis and Design and Iterative Development (3rd Edition)|
|Object-Oriented Object-Oriented Languages||This paper introduces the basic concepts of Agile Database Techniques, and effective ways to go about the Data-Oriented aspects of Object Oriented software|
|Object-Oriented Metrics: an Annotated Bibliography||Бюллетень новых поступлений за декабрь 2011г|
Питер, 2012. 366 с ил.; 24 см. (Библиотека программиста). Библиогр.: с. 353-358. Алф указ.: с. 359-366. Загл и авт ориг.: Design...
|An Introduction and Brief Examination of Object-Oriented Data Modeling||The Domain Analysis Integrated in an Object Oriented Development Methodology|
|Hypergraph-based Object-oriented Model for gis application||Evaluating the impact of different types of inheritance on the object oriented software metrics|