Civil Engineering Reference
In-Depth Information
phenomenon as a defect, while others may not because they do not have the same focus. Moreover, the
other usual shortcomings of expert systems are also faced in metal forming : the difficulty to verify the
expertise and to check the integrity of different sources of expertise, in order to avoid antinomies, and
the fact that expert systems cannot give an answer if the expert does not know it. Therefore, they allow
non expert people to study simple designs without requiring the presence of an expert, but they cannot
increase the knowledge.
In order to eliminate some of these shortcomings, the Case Based Reasoning techniques have been
developed, in particular when the rules are not easy to define. They are based on the analogies between
the different problems. First, they provide another software environment to store all the experimental
design results, so allowing, as expert systems, to keep the know-how of a company. Then, when a new
design has to be defined, the system computes the distance between the new problem and the already
solved problems stored in the data base. This way, it can suggest one or several solutions close to the user's
problem. This is just a first solution as well as indications which will eventually have to be improved.
However, if the design problem does not belong to one of the family of the database, the system will require
a further help from the expert. This approach has been successfully applied to forming processes [67].
However, on one hand, it is limited to a specific process, which means that there are different softwares
for different families of parts, and on the other hand, it does not actually solve the design problem.
Consequently, these two approaches can essentially be used to reduce the number of initial trials, but
the suggested design still has to be validated by calculations or experiences. However, they do not really
increase the knowledge about the design process and the design techniques. The other approaches, which
are described in the next sections, seem more promising.
The numerical simulation can efficiently be used to solve the various problems which have just been
raised: gathering the human expertise, standardizing the pieces of information and results, and verifying
the outputs of the artificial intelligence softwares. Actually, it can easily train the system. However, in this
case, the neural network, genetic algorithm approaches seem better adapted. Moreover, using the bio-
logical system analogy, they make it possible to suggest a design for a rather new problem. The optimi-
zation methods provide another type of design strategy which can also suggest innovative designs.
Backward Tracing Method
In the early 1980s, a pioneer work was carried out on the basis of the finite element simulation of the
forming processes [57]. In the frame of numerical based approaches, it is very specific as it is a constructive
method for preform design in forging or other related processes. This way, it can provide new and original
designs. The basic idea is to trace back the deformation path. The procedure starts from the final desired
part and computes a backward deformation path until the beginning of the process. It so proposes a
preform which will make it possible to obtain the final part. Thus, this method consists in winding up
time, which is a very complex problem. At each time increment,
t
t
, upon the
t
t
configuration
(where
represents the domain of the workpiece), the algorithm computes the
t
previous configu-
ration, the corresponding velocity field
V
t
, and contact conditions such that they would provide the
configuration if the direct calculation was carried out. However, it is clear that the backward
deformation path is not unique and that the contact phenomenon is irreversible. Several backward
algorithms have been investigated for releasing the contact nodes, either using a prescribed sequence or
a specified criterion, or by recording the boundary changes during the forward simulation. So, the method
significantly relies on the selected strategy, which makes it quite difficult to produce a general and robust
software.
This method has been successfully applied to preform design for several forming 2D and 3D processes,
such as plane strain rolling, [43], three-dimensional rolling [44], and axisymmetrical forging [41, 68]
When the material has a complex behavior, for example when it exhibits a dependency on the strains
and temperature, this approach faces some additional difficulties regarding the irreversibility of the
deformation path. However, some applications have been carried out in these cases.
t
t
