IEEE Circuits and Systems Magazine - Q1 2023 - 38

completely new simulator, solution procedure included,
for every new case.
As doing so would be practically infeasible, the natural
way to go is not to provide standard " monolithic " cooling
simulators to configure, but instead to offer libraries of
their components (such as pumps, valves, Peltier elements
and so on) for the analysts to assemble their own cooling
system model for the case at hand. Quite intuitively-
as will be shown here as well-assembling a model this
way cannot be done with zero knowledge of the involved
physics. However, the required knowledge is enormously
smaller than that needed for building a new simulator from
scratch, or even for modifying the core of an existing one.
In particular, and fortunately, EBM makes it possible
to cope with the requirement just sketched with
an amount of modeling culture accessible to many, and
most important, with no need for the analyst to learn
about the numerical model solution; we here contribute
to putting the reader in the position of exploiting EBM
for the above purpose.
II. Synopsis
As said, modern IC cooling systems give rise to multiphysics
and inherently cyber-physical systems. To address
such systems, it is necessary to touch several subjects.
Given the wide audience we are addressing, each reader
most likely has a differently distributed knowledge of the
complete panorama. Achieving a consistent and streamlined
presentation, suitable for all, is thus a challenging
task. We provide in this section a minimal synopsis of the
article, showing what experts in one or another subject
could quite safely skip, at least on a first reading.
■ Section III provides a minimum of general modeling
background, introducing in particular the idea
of declarative-as opposed to imperative-model,
and sets the required terminology. We think
everybody should read this, at least to prevent
misunderstandings.
■ Section IV introduces EBM by example, to make
the reader capable of mastering the basics enough
to follow the rest of the treatise; references are
provided for a further, more structured study. Experts
of EBM can safely skip this section.
■ Section V comparatively discusses declarative
and imperative modeling, thereby providing evidence
that in MPSoC cooling the two approaches
need combining. Few ideas here would look new
to EBM-proficient people. However, we consider
it useful not to skip the section, at least for the
MPSoC-related considerations.
■ Section VI reviews the encountered physical phenomena,
and how to describe them as differential
38
IEEE CIRCUITS AND SYSTEMS MAGAZINE
equations. Experts of
first-principle modeling-
basically for thermohydraulic systems-could in
principle go directly to the next section. Nevertheless,
in this section we are also introducing some
domain-specific approximations, and these may
require one to use the section as reference for the
following ones.
■ Section VII explains how to describe the thermal
phenomena that occur in MPSoC cooling systems
using EBM tools. It also discusses how to handle
spatial discretisations in an optimized way, a matter
that could be of interest also to those already
proficient in EBM modeling.
■ Section VIII describes how Modelica models of
cooling systems, typically built with the said library,
can be connected to 3D-ICE-and potentially,
to any external simulation code that can be
endowed with an FMI interface compatible with
the shown one.
■ Section IX illustrates some examples of cooling
circuits built with the library, also covering the
interfacing with 3D-ICE. We suggest the interested
reader to employ the presented models as a basis
for his/her own first ones.
■ Section X concludes the article with some brief
considerations and sketches out future developments
and research activities.
We wish that the community will contribute to the
development process we started, and consequently that
the Modelica library we are presenting will foster the
growth of an ecosystem for sharing modeling experiences
and results.
For this reason, as well as to allow the reader to reproduce
all of our experiments and to build on our models,
we are releasing the library-both for standalone
use and connectable to 3D-ICE with the provided interface-as
free software, under a 3-clause BSD license. On
the same front, though the library should work with any
Modelica tool, we developed and tested it with the free
and open source translator OpenModelica [14], available
at https://www.openmodelica.org, so that no barriers
exist to using and sharing.
The library and the associated material can be
downloaded at https://github.com/looms-polimi/
computer_cooling/.1
III. Modeling Background
The word " model " has a wide range of meanings, and
more must be added if we also embrace physical or
1The library is continuously developed; we suggest the interested reader
to clone the repository via git and check/download updates with git
pull.
FIRST QUARTER 2023
https://www.openmodelica.org https://github.com/looms-polimi/computer_cooling/ https://github.com/looms-polimi/computer_cooling/
IEEE Circuits and Systems Magazine - Q1 2023

Table of Contents for the Digital Edition of IEEE Circuits and Systems Magazine - Q1 2023
