In section 3.1, an introduction to feature technology is provided. Feature definitions are elaborated in section 3.2 as feature definitions have been and still are subject to change, which can cause a lot of confusion. Feature based design is elaborated in detail in section 3.3. Feature based manufacturing is discussed in section 3.4. Features as potential elements for the integration of design and manufacture are treated in section 3.5. Finally, a summary and conclusions are provided in section 3.6.
Feature technology, is expected to be able to provide for an adequate basis for the integration of design and the subsequent applications such as engineering analysis, process planning, machining and inspection. Over the last few years, a vast number of papers and other publications on feature technology has come forward. The most important aspects dealt with in these papers are summarized in the remainder of this chapter and in Appendix A. The focus is on feature based design, feature based manufacturing and CAD/CAPP integration. Other overviews of feature technology are provided in [Case 93], [Catania 91], [Roller 89], [Ruf 91], [Salomons 93a], [Shah 91c], and [Wierda 91]. Currently, three main views are discerned on how to obtain application features, such as manufacturing features, analysis and inspection features, from a product model (e.g. [Joshi 90], [Shah 90b] and [Houten 91]):
Presently, the first two views - feature recognition and design by features - prevail. In [Husbands 91] the belief is advocated that feature recognition and feature based design alone are not sufficient to fulfil the requirements of CAD/CAPP integration. It is presently believed that future CAD/CAPP systems should support both feature recognition and design by features (e.g. [Houten 91], [CAM-I 90], [Wing#rd 91], [Wang 91a] and [De Martino 94a,b]). M#ntyl# and Laakko presented a hybrid feature recognition/feature based design system [Laakko 93]. This system is based on the belief that not all shapes can be represented by features. Thus, in addition to feature based modelling, conventional CAD techniques are required. This CAD-based geometry should then be subjected to feature recognition in order to arrive at a complete feature based model at the end of the design process. This seems like a contradiction: admitting the fact that not all shapes can be represented by features, but at the same time aiming at a complete feature based model. In [Sreevalsan 91] it is even advocated that all three methods should be integrated in a unified approach.
In comparison with the other two approaches, feature based design has the advantage of offering the possibility of capture and storage of relevant information during design which can be used for applications that take place after the design process has finished. Feature based design also offers the possibility for considering manufacturing and assembly aspects early in the design process. When using only feature recognition or interactive feature definition, this is not possible. Thus, feature based design seems a promising method supporting CAD/CAPP integration in a CE approach. However, feature based design systems not only need to communicate with feature based CAPP systems, but also need to receive feedback information from CAPP systems. These feedback information flows are not shown in figure 3.2, but are addressed later.
- A specific geometric configuration formed on the surface, edge or corner of a work piece [CAM-I 81]. A more recent definition of a process planning related form feature is: - Distinctive or characteristic part of a workpiece, defining a geometrical shape, which is either specific for a machining process or can be used for fixturing and/or measuring purposes [Erve 88]. Another more recent definition, also applicable to other domains than just process planning, is the definition of a form feature by Wing#rd: - A generic shape which carries some engineering meaning [Wing#rd 91].
A lot of confusion has arisen about feature definitions as features became also relevant to other application domains such as engineering design and analysis and because features did not necessarily relate to form. Therefore, the first two previously mentioned definitions are considered as definitions of manufacturing form features, or simply, manufacturing features and not as features in general. The reason for this is twofold: firstly to indicate that these features are form features and not more or less abstract features, secondly because these features are specifically related to manufacturing processes; they only have meaning for manufacturing and not necessarily for other applications.
As feature technology spread from process planning towards design, inspection and engineering analysis, feature definitions tended to become more general. Examples of these more general definitions of features are:
- Recurring patterns of information related to a part description [Shah 90b]. - A semantic grouping used to describe a part and its assembly. It groups in a relevant manner functional, design and manufacturing information [Giacometti 90a]. - A geometric form or entity whose presence or dimensions are required to perform at least one CIM function and whose availability as a primitive permits the design process to occur [Luby 86a]. - A carrier of product information which may aid design or communication between design and manufacturing, or between other engineering tasks [Shah 90c]. - Any entity used in reasoning of design, engineering and manufacturing [CAM-I 90]. - A region of interest [CAM-I 90].
A lot of different kinds of features have been proposed: functional features e.g. [Giacometti 90a,b], assembly features [Sodhi 91], mating features, physical features [Kiriyama 91] and even abstract features [Shah 91b]. Shah defines an abstract feature as follows:
- Entities that cannot be evaluated or physically realized until all variables have been specified or derived from the model [Shah 91b].
Abstract features can be used during the design process because a lot of feature information may not be known in detail before the end of the process. The research as described in [Salomons 92b, 93b] provides evidence that abstract features are indeed used in the cognitive processes of practising designers.
No matter how general the definition of the term feature, what seems to be taken for granted is that features finally are attached to some geometric shape. Shah defines what requirements a feature should at least fulfil [Shah 90c]:
- a physical constituent of a part. - be mappable to a generic shape. - have engineering significance. - have predictable properties.
In Appendix A.1 feature based design is elaborated from the viewpoints that have been introduced in chapter 1; the design process, the design object and the design knowledge. Some academic (feature based) design systems are also discussed in Appendix A.5. This subsection only briefly summarizes some of the most important issues related to feature based design.
The decisions that are made in design are not limited to single parts only as is often the case in process planning. The decisions in design are also related to assembly and functioning aspects. The first references on features in mechanical engineering design, e.g. [Pratt 84, 85] and [Luby 86a,b], were concerned with the modelling of single parts only. This is probably due to the fact that the features concept has been brought over from process planning where the decisions are indeed related to single parts. Although feature based modelling of single parts is a significant improvement when compared with traditional CAD systems, true computer support of designers still has not been achieved. As will be shown later on, this situation is now changing.
From a design point of view, feature based design has a much better potential for computer support of the design process than current non-feature based CAD systems do. Features are meaningful elements for designers and the use of them can speed up the design process as well as provide a means for standardization, thus reducing cost and time-to-market. Other advantages which can be expected from feature based design are improvement of the quality of design and a better interface with applications such as process planning and analysis. Features can contain manufacturing oriented information. Features, therefore, can be a basis of communication on manufacturability between CAD and CAPP.
Despite the promises of feature based design as mentioned above, it has not yet reached its expectations. According to Sreevalsan et al., the main reasons for this are [Sreevalsan 91]:
- there is no finite set of features in design. - data management problems are non-trivial. - the need for feature recognition does not go away as features are application-specific. - it is not clear whether designers actually design in terms of features or that features result from other considerations.
With regard to the first point, it can be said that it is generally true that a finite set of features is insufficient to create all possibly desired geometry. However, if a finite set of features is sufficient to generate a relatively high percentage of the desired geometry (say about 80%), the remaining geometry can be generated with traditional CAD operations, such as proposed in e.g. [Laakko 93]. Many industrial applications of feature technology seem to support this pragmatic feature approach.
The second point stated by Sreevalsan is also generally true; data management problems are non-trivial. For instance, if feature dimensions are modified, other but related constraints may also need to be re-evaluated. Intersections resulting from the dimension change have to be calculated and the new or affected features need to be validated. These issues are non-trivial. Moreover, data management problems are not restricted to features alone, but are also related to assembly models, the design history and a lot of other information. Especially in collaborative design, data management will be of increasing difficulty.
The third remark made by Sreevalsan is also generally true: the need for feature recognition will not vanish as features are application specific. This is shown more clearly in section 3.5 and Appendix A.4.5.
With regard to the fourth point mentioned above, it is important to note that our current understanding of design indeed supports the view that designers think in terms of features. Of course, other considerations such as function aspects, are considered in practice as well. Evidence of this understanding has been derived from protocol studies described in e.g. [Kuffner 91] and [McGinnis 91]. The results of two protocol studies undertaken by the author in order to investigate the use of features in design and manufacturing, also support this view [Salomons 93b].
- manufacturing features provide for a natural form of communication (process planners think in terms of holes, pockets etc.). - manufacturing features simplify process planning since there are only a finite number of ways to manufacture a feature.
Automation of process planning requires that product data, i.e. manufacturing features, are extractable from the product model automatically. However, CAD product representations in product modellers usually differ from the type of information required in CAPP (e.g. manufacturing features). Until now, feature recognition has been the most common approach to extract manufacturing features from CAD product models. In fact, this means inferring a lot of information from the CAD product model while most of this information already has been generated during the design process. This information is lost when the result of the design process is stored in the CAD model. Feature based design can (at least partly) help to overcome this problem.
Process planners prefer to have as much freedom as possible in making the process plan. Therefore, they might prefer to have some notes together with a sketch instead of a completely detailed drawing. In feature based CAPP systems, this would imply the use of abstract features and incomplete geometry as input. In practice, process planners often can give manufacturing oriented design advise if they are informed of the functions or the design intent. Enabling CAPP systems to include such a functionality, access to the design history and the assembly model is necessary.
Manufacturing form features are closely related to the machining processes and are instrumental for solving fixturing and measuring problems induced by the manufacturing process. The most important properties of manufacturing form features are:
- geometry: often a distinction is made between protrusion features, depression features and surfaces, as these features can closely be related to different machining processes. - entry/exit boundaries: These boundaries are important for giving tools entry access or exit capabilities. - depth boundary: which is important for the length of the tools. For example the length of a bore has to be at least as long as the depth of the hole. - external access direction: this indicates the possible directions from which features can be machined. - exit boundary status (through or not through): this is an important characteristic for fixturing operations. - dimensions (parameters of the feature): these are important for tool selection; pocket milling for example requires the mill radius to be equal to or smaller than the corner radii of the pocket. Extreme dimensions can make a part non manufacturable with the given tools and machines. - tolerances: these are important for finishing operations and fixturing. Tight tolerances will make the product expensive and hard to manufacture. - material (treatment); not every tool can be used to machine each type of material at the required tolerances.
Sometimes it is argued that functional information or design intent should also be conveyed as a relevant property of features for process planning. As an example, Dong et al. present a tribological surface that has been heat treated and precision finished [Dong 91]. This feature (surface) should not be used for either clamping or supporting as this could damage its function as a tribological surface.
Most of the research in feature based manufacturing has been focused on material removal operations such as milling, drilling and turning, e.g. [Erve 88], [Cutkosky 88], [Anderson 90a,b], [Houten 91]. Other manufacturing operations such as injection moulding and sheet metal manufacturing have been addressed to a lesser extent, e.g. [Irani 89] and [Vin 93a,b, 94a,b], [Kappert 93, 95]. The latter approaches can differ significantly from those used for material removal operations.
As has been discussed in the beginning of this chapter, there are three main ways of obtaining manufacturing form features from product models: automatic feature recognition, interactive feature definition/identification and design by features. The latter has been addressed in the preceding section. Due to space limitations the first two methods are summarized in Appendix A.2 and A.3.
Machine tool selection can be divided into two distinct steps: technical and economical machine tool selection [Houten 91]. Technical machine tool selection deals with the selection of machine tools which can be used to produce a part in a technical sense. In the economical machine tool selection procedure, it is determined which machine tool is the preferable one, with respect to availability and economy. Features carry important information for machine tool selection. However, it is hard to give some general guidelines as these are often very company specific.
In the design of jigs and fixtures, features also play an important role. In these cases, features are often faces (primitive features). The tolerances between or on these faces are the constraints. Because several different tolerance types are allowed, it is difficult to unambiguously interpret the tolerances in order to determine the best set up and the jigs and fixtures required. In order to overcome this problem, tolerance conversion has been proposed for prismatic parts [Boerma 90].
Manufacturing features (in the end) always relate to manufacturing processes. To map manufacturing features onto manufacturing processes, a variety of techniques can be used. Decision tables or rules can be used for this process [Houten 91]. This works rather well when the features are relatively simple. However, problems occur when features become more complicated, or when they intersect. In PART and PART-S, compound features are used when dealing with complex features. Compound features are composed of a number of (atomic) manufacturing features. In PART and PART-S machining methods cannot only be defined for atomic features but also for compounds. In PART and PART-S rules are used for knowledge representation and a backward search from the final part specification to the blank specification, using a tree search procedure, is applied.
After machining methods have been selected, cutting tools can be selected. In cutting tool selection, a database of available tools is required. This database has to be checked, in order to verify whether the tools necessary are present in the database. The rules that are used in tool selection specify attribute ranges instead of fixed values as this allows freedom in the selection of tools, supporting the idea of using the same tool for different operations as much as possible. Boogert carried out an extensive study on the application of tool management in PART [Boogert 94].
Features as used in design can differ significantly from those used in manufacturing. The problem arising from this is sometimes referred to as the multiple views problem towards design features and manufacturing features. Figure 3.3 gives a frequently quoted example of multiple views on design features and process planning features in the case of a prismatic component. In this example, the designer would prefer to design with protrusion features, the ribs, as these are functional to him. A process planner would look at the material to be removed, the depression features, are of most importance: in this case the slots and the step.
Another example of multiple views on features is provided in [Hummel 89a] in which design features, process planning features and inspection features are distinguished. The design features have been identified from studying a designer who was asked to model a single component with the aid of a feature based CAD system. The features offered by this system were geometric building blocks, not related to the application domain. Thus, the design features that were identified were only generic shape elements related to the modeller used and were suitable only to document the final geometry of the component. So, the design features as identified by Hummel et al. do not have a functional meaning or design intent; they could have been used by the traditional drawing office, but hardly by actual designers.
In [Salomons 93b] attempts have been made to identify the multiple views on design form features and manufacturing form features in a situation where no feature based CAD system was employed. Paper and pencil were used instead of a feature based CAD system as in the research by Hummel et al. [Hummel 89a]. In this research, multiple views between some design form features and some manufacturing form features could be identified. However, they were not as different that they could not be combined if necessary.
In unifying design features and manufacturing features, powerful feature definition facilities are required as part of a feedforward approach to CAD/CAPP integration. Relatively little work has been performed with regard to feature definition as an important factor in CAD-CAPP integration. This "non-classical" approach is detailed further in section 3.5.3 and Appendix A.4.5. In the following, the classical approach towards multiple views on features is elaborated, starting from the assumption that different feature types are necessary in different applications and that some kind of feature mapping or transformation is necessary.
The classical multiple views problem can be tackled in a number of ways. One possible solution is to use the same (manufacturing) features in both design and manufacturing. Another solution is to use features in design that differ from manufacturing features and to use some way of feature conversion or mapping in order to arrive at a product model in terms of manufacturing features. These conversion techniques are research issues and are elaborated in Appendix A.4.5. Feature recognition can be seen as a way of feature mapping [Kappert 93, 95]. In this way, the feature information that was generated during design is completely discarded and manufacturing related feature information is inferred from the solid model of the part. In the following two sections the two classical views on the use of features for the integration of design and manufacturing are elaborated. Then, a closer view on the integration problems is provided.
By allowing abstract features in design which finally result in manufacturing features, a more natural design process and a better link with CAPP could be achieved. Abstract features enable a better link of features with function. Abstract features are not necessarily volumes but they can consist of functional elements. It must be noted, that functions are related here to the idea of working principle dependent functions as presented in chapter 1. Besides working principle dependent functions, a higher level function representation, not necessarily directly or indirectly related to features, seems an interesting research topic for the future. For this, techniques like bondgraph representation may be used. Bondgraphs allow the modelling and simulation of electro-mechanical systems. Bondgraphs can be used to model the dynamic and kinematic behavior of mechanical systems. Examples of bondgraphs applied in modelling systems are the Schemebuilder [Bracewell 93] and MAX [Vries 94a] systems. These systems amongst others allow for automatic generation of design alternatives based on a bondgraph description of the required functionality. The combination of these techniques with catalogs of working principles coupled with the use of (abstract) features could possibly allow for a mapping of function onto (preliminary) form. This thesis, however, does not elaborate further on higher level function representation.
The abstract feature concept allows a design by least commitment approach to be incorporated. In this case, designers can leave functionally irrelevant parts of the design unspecified and let the process planning system fill in the details.
Assembly aspects must be taken into account when willing to integrate design and manufacturing as features and (functional) tolerances often result from assembly aspects. Thus, the reasons behind previous design decisions, such as functions and tolerances, can better be inferred from assemblies than from single components.
A tool which may be of help when manufacturing problems occur is a tool which is able to capture and play-back the design intent or design history of a design object. Initial studies in this field have been reported in [Ullman 91] and [Aasland 93].
In the case where neither feature definition, nor the use of abstract features nor design history tools can help to sort out manufacturing problems, ordinary communication or negotiation can. Present computer tools begin to support this paradigm by Computer Supported Cooperative Work (CSCW) tools. These tools are not only of potential use in cooperating between design and process planning, but also in a computer based approach to CE as a whole.
A number of feature transformation or mapping techniques have been proposed to solve the multiple views problem. Most of these techniques seem to be inspired from the domain of modelling prismatic components for machining. However, these techniques seem to be focused on single components: they are based on the idea that the design features of a component should be transformed into the manufacturing features of the same component (Appendix A.4.5). However, in this research it has been overlooked that in most cases the design features are not application specific in the first place and that designers, if not forced to use a particular CAD system, would like to use more application specific features. A protocol study has indicated that consensus on feature names and feature attribute names could alleviate most of the multiple views problems [Salomons 93b]. If this proves to be true, flexible and interactive feature definition capabilities are required. Therefore, feature definition becomes an issue of importance. In feature definition, one can first of all try to combine multiple views on features into one feature definition on which consensus can be agreed upon. If no consensus can be obtained, mapping rules have to be defined. If feature mapping is not as straightforward as feature recognition, hints may be defined in order to facilitate feature recognition. Both the ease of feature definition and the ease of defining mapping rules depend on the way features are represented (Appendix A.4.1 and A.4.2).
Apart from the transfer of nominal part geometry, tolerances are of great importance in a CAD/CAPP link. In order to prevent the specification of too many and too tight tolerance constraints, functional tolerancing should be supported. Functional tolerancing cannot be dissociated from the assembly model. In most feature based CAD approaches, assembly concerns, which are often important during design, are not or insufficiently taken into account. As tolerances and features often are the result of assembly considerations, it is important to take assembly aspects into account.
Design by least commitment could be supported when incomplete geometry (abstract features) can be used. Another issue of importance is that of design histories: these could be of great help when problems with regard to manufacturability occur. Also, computer supported cooperative work seems an interesting opportunity for improving the CAD-CAPP link as well as a means of fulfilling some of the needs of CE.
Thus, some of the most important issues in bridging the gap between feature based design and feature based process planning are the way of representing features and the way of defining them as well as the possible use of abstractions in a design by least commitment approach. Also, functional tolerancing and a design history tool should be offered in addition to facilitate manufacturability and thus improve the link with CAPP. This is shown schematically in figure 3.4, which can be seen as a more detailed version of figure 2.5.
Note that the arrows in figure 3.4 which do not actually end in a function block but at the border of either CAD or CAPP functionality (dashed), are specific information flows. These flows provide information generated at "the other side of the wall" on the request of either the CAD or CAPP system end user. For this information, apart from presentation functionality, no additional functionality needs to be present to process this information automatically; this is done by the users who requested for it. Note also that feature identification in figure 3.4 is generic; via feature definition, it has a link with feature recognition. Thus, feature identification can be seen as a possible mode of feature definition, especially suitable for CAPP; see also chapter 6 and 7.