Fig. 1: The engine head.
27th ISATA - International Symposium on Automotive
Technology & Automation
Aachen, Germany - October 31 - November 4, 1994), pp. 51-57, ISBN 0-947719-64-4
F. Folini*, M. Galloni*, U. Cugini*,
and G. Bocchi+
(*) Department of Industrial Engineering, University of Parma
(+) B.R.D. SRL, Vicofertile, Parma Italy
Today CAD systems are mainly conceived as tools for the production of detailed
and fully defined geometric models. The technological improvement of such
systems can enhance dramatically the time required to obtain a stable model of
a new product. Besides, the introduction of rapid prototyping techniques,
offers a new perspective in their usage and sets new requirements: they have to
provide the designer with complete support for implementing feedbacks coming
from the analysis of the prototype. In our experience, these feedbacks are
often indications for refining and optimizing the part and for catching minor
errors and do not affect the logical and functional structure of the part.
Therefore, the designer should be able to change the shape of the part, add new
features and modify geometrical dimensions without losing the model integrity
and significance; parametric and variational CAD systems offer this kind of
functionalities.
In order to verify the practical implications of coupling prototyping
techniques with parametric design tools, we chose a significant design problem:
the design of a motorbike engine with six cylinders and five valves each
cylinder, already sketched in the main structure, but incomplete in a lot of
characteristics and dimensions.
First, we defined a logical structure of the specific design process, the
freedom degrees of the part and the full set of functional and geometrical
relations. Then, on the basis of this analysis, we analyzed the CAD systems
commercially availables and chose Eureka, a new parametric and variational CAD
system realized by Cad.Lab. With respect to other systems, Eureka provides the
tools for representing the wider set of relations and offers good
functionalities and performance with direct support from the original software
company.
Given that Eureka, as other similar CAD systems, is intrinsically unable to
represent all the constraints and relations identified in the analysis phase,
we made a pondered selection and simplification of the original set. Then, we
defined a solid model containing a subset of the selected relations and
controlling parameters and imposed the remaining relations and constraints. The
result was a geometric model enriched with non-geometric information and
flexible enough to support deep dimensional and morphological modifications.
Starting from an initial configuration, we produced the first tentative real
prototype by means of a specialized service supplier, employing
photopolymerization techniques. Using the prototype and the drawings obtained
from the CAD model, we checked the results with the designer and the
technological and production experts. The meeting produced a lot of discussion
based on the prototype with minor interest on the traditional drawings. We
collected and coordinated dimensional and morphological changes proposed by the
experts and implemented them into the CAD model.
The use of a parametric and variational system was fundamental in this phase
because the changes could be done interactively with an immediate graphical
feedback. The changes required by experts where fully supported by the CAD
model we realized because, as expected, they do not have any impact on the
logical structure of the design but only in non functional detail and minor
aspects. This experience showed also the limits of these CAD systems; they are
often unable to represent non trivial design rules, as for gears and helical
springs dimensioning, and to map values computed from constraints evaluation
within availability catalogs, as for bearings.
Feedbacks from the experts showed that the immediate availability of a real
prototype is effective in reducing the lead time and improving the part
quality.
The impact of rapid prototyping in the design process and methodologies and the practical implications of coupling prototyping techniques with parametric design tools, can be pointed out only experimenting a complete design process on a significant test case. Therefore, we looked for a mechanical part or component satisfying the following criteria:
Components from three different industrial areas, were analyzed: packaging
machines, pumps and motors. After the analysis of different alternatives, we
decided to work on a mechanical component of a motorbike engine; in the engine,
the head (Fig. 1) shows the mechanical and morphological characteristics
we are looking for, and satisfies all the criteria pointed out.
We selected a four strokes competition engine having the following
characteristics: six cylinders of five valves each, "V" architecture of 90°,
water cooled, total displacement: 750 cm3, bore: 65 mm, stroke: 37.5 mm,
maximum rpm: 14000, compression ratio 12:1.
Fig. 1: The engine head.
It was the result of a previous experimental work made by the designer G.
Bocchi, stopped at the conceptual design phase, [1].
Our study started from documents and data produced in the conceptual design
phase; the information was mainly available in form of sketches with a lot of
not fully defined details. We organized a series of meetings, involving the
designer and CAD experts, for acquiring the basic information on the engine;
according to the designer indications, we formalized the global requirements of
the design, the main design rules and an operative procedure for completing the
design.
The design requirements were at high level and in implicit form (e.g., to
maximize the specific power within the total displacement); we tried to
explicit the requirements in order, both to get clearer targets for the
parametric model definition and to define a set of verifications and analysis
on the model and on data obtained from it. The main requirements of the head
design are:
The next activity was the acquisition and description of design rules, i.e., rules used by the designer for choosing among several alternatives and for positioning, connecting and dimensioning functional elements of the engine head. We tried to explicit these rules in terms of constraints, equations, geometrical relations, conditional statements, and so on. This phase was fundamental in order to obtain, in the following phases, a model satisfying the designer expectations. We experimented that only a small part of the designer know-how and engineering rules can be mapped in an explicit form useful for being inserted in a computational model, [2, 3]. For example, the optimization of inlet and exhausts ports shape involves so many and so complex rules, methods and experiences that could be inserted only partially in the model. Furthermore, talking with the designer, we found out the design steps logical sequence and the priority associated to each target; this helped us to plan the work and to join results from each sub problem.
Within the previous phase, we chose a meaningful test case and collect unstructured knowledge on the engine head design problem. In order to select a parametric CAD system, we analyzed some commercial systems and a research prototype and finally decided to implement the model in the Eureka system, a new parametric and variational system produced by Cad.Lab SpA (Casalecchio di Reno, Italy). The Eureka features relevant for our work are, [4]:
Eureka system uses variational techniques to solve constrained 2D profiles and parametric techniques to represent 3D models. This mixture of parametric and variational techniques is usual in commercial CAD systems; it seems to be the best compromise between the flexibility of the numerically unstable and slow variational approach and the stiffness and performances of parametric approach [5, 6, 7]. The Eureka standard work procedure starts with the definition of a 2D profile, followed by the imposition of a set of constraints and the definition of simple relations among parameters. The behavior of the profile is then tested by changing the value of some driving parameter and analyzing the resulting geometry. When the behavior has been verified, the profile is ready to be extruded along a vector or a spine. In Eureka system, various profiles can be created in different work planes and connected by means of positional constraints; furthermore, the parameters of each profile can be related, via algebrical relations, to parameters of others profiles (e.g., the radius of this fillet is two times the radius of this hole).
Fig. 2: Combustion chamber modeling.
We chose to work initially on a single module of the head, corresponding to a
single cylinder and then to combine single modules in order to obtain a
complete engine head, [8]. According to the Eureka standard procedure, we
started defining a group of key profiles of the engine head. The first profile
was the triangular combustion chamber cross-section, important for the
definition of valve angle. In order to generate the combustion chamber, the
profile extrusion was intersected with a cylinder so providing the required
geometrical references for positioning the valves (Fig. 2). Working
on the combustion chamber volume, we studied both the four valves and the five
valves solutions. Using advanced inquiring functionalities we were able to
verify that, under the design requirements, the five valves solution is the
most convenient; in fact the valves inertia reaches the optimal values even if
the useful area decrease.
Then we attached the spines, a set of spatial curves, on the combustion chamber
and used them to extrude a circular cross-section. The obtained solids,
completed with the spark plug housing, refined with adequate fillets and
connected to the combustion chamber, previously defined, represents the empty
volume of the engine head. Offsetting the surfaces of resulting solid we
generated the central solid structure of the head with complete inlet and
outlet manifold, shown in Fig. 3.
Fig. 3: Inlet and outlet manifold modeling.
The next activity was the definition of a new 2D profile: a cross-section of the engine head external box, shown in Fig. 4. In this activity we experimented the limits of variational techniques: the definition of a complete and not contradictory set of constraints requires a lot of work; the users spent a lot of resources facing the system idiosyncrasies instead than solving the problem. The external box was connected with combustion chamber, inlet and outlet manifold thus obtaining the basic module of head. The following activities were the insertion on the module of minor details such as holes to fix the cam shafts and the definition of relations among parameters of different sub-parts. At the end we verify the behavior of the global model shown in Fig. 5.
Fig. 4: Modeling the profile of the engine head external box.
We were not able to reach the expected flexibility on the final model: the main problems did not depend on dimensional and morphological variability but on the significance of the results. For example, we were unable to set a parameter that switches completely and correctly between four and five valves configurations. On other targets, not so ambitiouses, the model behavior was satisfactory.
Fig. 5: The engine head module model.
The use of rapid prototyping techniques introduces an interesting question:
which is the best status where to stop the design process and to produce a
prototype? In the engine head design we stopped just before the insertions of
minor details on the model that could limit the flexibility of the model and
preclude some design alternative. At this point the designer already made all
the important decisions on the engine head but he needed a validation
mechanism.
The generation of a computer file containing the model description for the
rapid prototyping services required a single command on the CAD system. The
data format used by the CAD system was the SLA, the standard format used for
data exchange in stereo lithography systems; using SLA a solid model is
described by means of triangulated surfaces represented in binary or ASCII
formats. The first tentative for obtaining a prototype failed because the
Eureka triangulation algorithm left some hole along the boundary between NURBS
patches of the model. The software company, involved in this problem, quickly
fix the algorithm and we could produce a completely close triangulation of the
part.
Fig. 6: The prototype.
In a few days, we obtained the physical prototype (Fig. 6) by means of a local rapid prototyping service. When the prototype becomes available, we organized a meeting with designer and CAD experts providing also drawings plotted from the CAD model. The main interest was on the characteristics of the prototype and with minor use of drawings. The meeting confirmed the main design choices shown by the prototype and suggested minor dimensional and morphological changes. We collected the indications and directly implemented them on the CAD model modifying the corresponding parameters. We observed that:
Peoples involved in the design, agree that in a more favorable situation, with not so expensive rapid prototyping tools, the designer should analyze more than one prototype corresponding to different design alternatives, e.g. four and five valves solutions.
The increasing diffusion of rapid prototyping techniques can really change and enhance the current design process; the availability of the physical prototype, reproducing physically the model shape, can not be effective without adequate software tools. Our experience demonstrated that parametric techniques can provide today the required modification mechanisms. We also pointed out the limits of these systems but the global valuation it is very positive.
We wish to tank the CEDI Laboratorio CAD Avanzato at the Università degli Studi di Parma for providing computer resources and technical support and Cad.Lab SpA for providing the software tools and founding the prototype.