I have employed in COBIK (Centre of Excellence
for Biosensors, Instrumentation and Process Control) in September
2010. My main responsibilities at COBIK were
development of optimization and modelling software, leading software
development within the Laboratory
for Advanced Materials Systems (former Laboratory for Control
Systems) and Laboratory
for Multiphase Processes of the University of Nova Gorica,
supervising researchers and developers and managing related
projects. The accomplished objectives were to connect
development of numerical software in both laboratories and to
establish collaboration with industrial partners that will enable
quick transfer of knowledge to industry and its commercial
exploitation, for which I have prepared plans and worked on their
realization.
With valuable input of co-workers, I have prepared a business
plan for a spin-off enterprise of COBIK that would take over
commercialization of the developed knowledge, which elaborated on
the necessary investment, founding partners, intellectual property
ownership, initial generation of startup's revenue and evolving
business models, and fundraising at later stages. The spin-off was
not launched due to conflicting interests of principal stakeholders
that were impossible to match, but we obtained valuable experience
on commercialization opportunities, specifics and obstacles for this
kind of technology, about academic-industrial collaboration,
specifics of operation on both sides, and alignment of their
interests necessary to carry out successful projects. Within the
scope of COBIK, we concentrated on supporting emerging materials
industries (especially carbon nanomaterials like fullerenes and
CNTs) and steel production (as traditional industry with plenty
of niche opportunities where new technologies can be
introduced), although plans were incorporated to broaden the future
scope. We created complex numerical models that were successfully
used in steel production industry. Our accomplishments were based on
development of substantial amount of new modelling software and
algorithms, efficient integration of existing knowledge and tools
with implementation of well targeted improvements to achieve
specific objectives, and effort invested in coordination of
multidisciplinary teams. Many of these would not be possible without
careful planning based on more than twenty years previous experience
in R&D, academic-industrial collaboration and work with hi tech
industry. Yet there were many new lessons learned, in particular
about agility needed in turbulent conditions with rapidly changing
context and complex interplay of conflicting interests of major
stakeholders.
Documentation
A number of documents were made available about the ideas,
conducted research, developed software and industrial applications.
Some of these documents were entered into the COBISS
database (see this
page for a partial list) with links to documents
stored on Dropbox. The current web page contains links to
selected documents. In addition, you can find links to reports and
other documents prepared by Igor
Grešovnik and Tadej
Kodelja (whose work focused on artificial neural
networks-based modelling).
Organization of Code Development for
Collaboration with Commercial Partners. Treats
organization of software development within academic environment,
which enables collaboration with commercial corporations and direct
transfer of knowledge through development of customized
applications. Proposed organization tries to satisfy all the needs
of research institutions, wile enabling development of
industry-standard commercial software on top of the same code base
in such a way that conditions imposed by commercial customers can be
met completely, including protection of intellectual property rights
and trade secrets.
Modeling of Continuous Casting of
Steel with Artificial Neural Networks (2012). This
technical report makes a brief introduction of artificial neural
networks (ANN) and their use in modeling and approximation, and
demonstrates application of neural networks to modeling and
optimization of the steel production process. By Tadej Kodelja and
Igor Grešovnik, 2012.
Report
on Modelnig the Complete Process Path by an Artificial Neural
Network (abstract).
This detailed study describes an attempt of modeling of the complete
chain of processes in a steelwork that includes
steel making, continuous casting of steel, hydrogen
removal, reheating, multiple stage rolling, and cooling on the
cooling bed.
Model is based on about 2000 data sets comprising of
detailed records of chemical composition, processing parameters, and
obtained final properties of the produced steel. Due to data
deficiency, the model can not accurately recapitulate the reality,
and it has been only used to demonstration of concepts and
capabilities of the developed software for neural network-based
modeling. The created model was later used in research (including a
number of experimental studies) on feasibility of various modelling
strategies in steel production, aiding to successful development of
more accurate models that were used in practice. The report features
detailed study of the model with numerous graphical illustration of
the model-based responses, many of which can be easily qualified by
experts as incorrect. By Tadej Kodelja and Igor Grešovnik, 2013.
Presentations
Below are some presentations that I have shown in order to
promote my ideas of software development.
Software
Development in the Laboratory. This presentation was perepared
for staff members of both laboratories in order to explain how
software development will be organized.
A
Software Framework for Optimization of Process Parameters in
Material Production, Applied Mechanics and Materials
(Volumes 101 - 102 - Advances in Engineering Design and
Optimization, by Di Zheng, Yiqiang Wang, Yi-Min Deng, Aibing Yu
and Weihua Li (editors)), pp. 838-841, original available at scientific.net,
by Igor Grešovnik, Tadej Kodelja, Robert Vertnik, and Božidar
Šarler, 2012.