C.S.E. - A BRIEF HISTORY OF THE PROJECT
From the end of 1990s it became evident that
Sargon required a
post-processor dedicated to joints. However, an investigation of the problem showed how
difficult it was to address the subject; on the other hand it was rather clear
that the solution suggested by some of the program’s users, that is, putting in
place specific routines for the calculation of typical joints, would not result
in a product that was adequate to meet the actual needs of the specialists, the
practitioners who comprise the principal market for both Sargon and CSE.
Given that the problem was not only difficult but also exceptionally interesting, it
was decided to address it in the most general way possible, within a broad class
among all possible classes, aware that the work on the project would probably
take several years. Among other things, the requirements of other work
commitments made it impossible to carry out this project uninterruptedly, nor
was funding from other institutions or associations in fields related to such
problems forthcoming: generally speaking, either the products that are sold in
Italy are created by specialists in other countries (any nation or country, no
matter how far away or exotic “sounds” better in our technical and scientific
arenas, in spite of the glorious Italian tradition that includes Galileo,
Castigliano, Ricci and Piola among others, and the fact that there are excellent
Italian scholars working in the field today), or the projects that do manage to
find funding prove to be futile, resulting in the waste of public funds (there
are innumerable examples). Research in the true sense of the term, which
requires years before producing results and innovations, is extremely difficult
to carry out in Italy, even more so if the researcher has no “patron saint”, and
thus requires a degree of courage and perseverance that few can maintain. The
CSE project was carried out in intervals of one or two months a year over the
course of about ten years.
Notes on the research in the analytics of jnodes.
The
first attempts to define the problem date back to 1999. A specific terminology
was introduced that included new terms such as “unito” ("joined"), “unitore”
("joiner" or "connector"), “tramite” ("through"), “matrice di trasferimento” ("transfer
matrix"), “jnodo” ("jnode"), “renodo” ("renode"). This was also the period of the
first thoughts about how to describe the joint in the most general way possible.
Arduous problems soon arose that could be initially described but not resolved.
In the meantime (2000, 2001) work began on writing the general parts of CSE,
those which in any case had to be addressed independent of problems related to
calculation: the graphic representation in 3D, hidden surfaces, the cataloguing
of the nodes. In rendering hidden surfaces, CSE does not make use of raster
algorithms, but rather of vector algorithms: this makes it possible to create
borders that much more easily readable than those produced in many other
programs that use pre-set routines using raster algorithms. This results in
images that are clear and readable, much more so that those of other programs
that use shadows and various hues of colours to represent the shape of the
objects.
Chronologically, the first specific problem addressed was that of how
to convert a FEM model into a model of the members. It became clear (this was in
1999-2000) that additional information had to be included in the FEM model, and
the program was adjusted so that it was able to extract a model of the members
from a FEM model.
The next problem addressed was the thorny one of how to
recognise equal jnodes. It was clear that there had to be a way to distinguish
jnodes from renodes. The routine for the recognition of equal jnodes was the
first extremely important success in the development of the program. Even at
that early phase of development there was no other program that could accomplish
anything similar.
This goal was reached thanks to jnode analytics, a study
which explained the logic of the most frequent connections. So CSE was able to
recognize them automatically. Today CSE equal-jnodes recognition is a single
command, and it lasts some seconds, but years were necessary to completely
deepen the issue.
This brought the work to the early years of the 2000s, and the
program was marketed with these somewhat limited (but innovative and useful)
capabilities of analysis. At the time it was generally believed that the
calculation of joints could only have been accomplished by means of
“instructions” given to a program by the user, given the lack of general rules
for calculation. Therefore, the program began to be equipped with a
substantial set of functions aimed at encompassing the rules of calculation.
Thus appeared the pre-defined variables associated with the components (e.g.,
m2.tw, P1.A, etc.), the possibility of adding new variables (e.g.,
sigmaN=m1.N/m1.A), and the addition of conditions, or relations (generally
inequalities) expressed as a function of the pre-defined variables and those
introduced by the user, capable of describing the checks to be performed (e.g.,
m2.N < m2.fu* (m2.A - B1.n * B1.dh * m1.tw)/gammaM2).
Meanwhile, much effort had
been dedicated to the construction of the renode, that is, to all those commands
intended to ensure that the user was able to work in an environment that was
perfectly 3D when adding components and processing them. You might say that the
aim was to create a virtual LEGO™. This part of the project was quite complex
and took a long time to develop. It required creating commands for positioning
and manipulating the objects, not only by means of rigid-body transformations
(rotations, translations) but also by means of additional and indispensable
processing. This environment is typical of programs that allow 3D modelling,
whose capabilities, however, are usually “limited” to drawing or “rendering” the
objects once they are created. In the environment of the CSE project, such
functions could in some way be considered preliminary, since the fundamental
object was calculating the joints and not drawing them. The problem was
to
calculate, and to check.
Notes on the topology of the elementary renodes
Work on
the problem slowed from 2005 to 2007, in part because large projects and
commissions made it very difficult to work on research. While it was already
possible to recognise equal (j)nodes, construct the scene by positioning the
components, recognise the joints solely on the basis of their geometric
positioning, and add many other useful functions, the problem of calculation was
still unresolved. Could it be true that the user had to explain to the program
how the joint was to be calculated? Was there no other way? The idea (which
remains such today) is that the user could explain to the program what checks to
make, and that the program would then apply them automatically, memorising the
typologies of nodes in order to call them up at a later moment. Although general
and useful, this approach still required a certain degree of effort on the part
of the user: we would have to specify typical nodes and then the user could go
forward from there.
Between the end of 2007 and the spring of 2008, a window of
several months’ time dedicated to in-depth study finally made it possible to
arrive at a solution to the main problem, that is, the automatic calculation of
freestanding joints. Concepts such as hyperconnectivity were defined, and from
there it was finally possible to understand how to describe the problem of
calculation in a unitary way by means of a model with finite elements that was
clean, reliable, balanced and coherent. In the space of a month the program was
adjusted so that it could calculate, on its own, using hypotheses that were
broadly general and normally verifiable, with no need of instructions on the
part of the user, the stresses acting on the joiners, as well as
bearing checks, checks that were written and tested during the
spring and summer of 2008. A book containing information about the program and a
manual for validation to demonstrate both the validity and the reliability of
the calculation were written. The class of the problem was notably broadened
through the introduction of anchors, friction bolts and bolts only in shear.
There still remained problems to address and functions to add, but the main road
had been paved.
Knowing the stresses on the connectors opens the way for
checking the throughs and the
members. The problem was addressed starting in
autumn 2008, by first introducing the bolt layout with bearing
surface (e.g. base-plate), and then the checking of the net-sections of the
members, and finally
the simplified checking of the throughs. At the beginning of 2009 the
fundamental characteristic was added, which consisted in the automatic creation
of the fem models of the components, subject to
forces that were known
because they were calculated during the first, fundamental analysis. This opens a completely new vista: the automatic creation of
FEM models (throughs, members, stiffened or not) leads to checks that are much more rigorous of both the
throughs and of the members
themselves than those performed manually.
During
spring 2009 the hypertext help manual and the guide were updated and it was
decided that, although the program was still in full-blown development, it was
now time to make it available to the community of experts, not least because
this was a valuable source of funding for the project, and one that derived from
the project itself.
As of May 2009, the road ahead is clear, and although it is
certainly complex, and it will require more time to complete, there don’t appear
to be any more dilemmas such as those that were faced in past years. All of the
steps still to be taken are understood, partly because the functions that CSE is
still lacking (specifying the parameters of typical joints, specifying the
joints, further amplification of the range of applications of the program in
order to better manage certain problems such as complete penetration welding or
hollow sections), these are functions that appear not to present any particular
conceptual problems: the program will eventually arrive where other programs
arrive today (at the steel structure drawings or at the calculation of only
typical joints) but from a completely different point of departure, since by now
the problem of calculating the joints of steel structures has been solved in a
way that is very general and reliable from an engineering standpoint.
Milan, 17
May 2009 – some ten years after the CSE project began
home
