The Physical Systems Modeling Laboratory (PSML)

For more than 30 years, engineers have been describing physical systems through either linear or nonlinear state-space models. The modeling and simulation tools developed during this era, such as ACSL reflect this emphasis on causal models.

Physics, however, is essentially acausal (Cellier et al., 1995). It is one of the most flagrant myths of engineering that algebraic loops and structural singularities in models result from neglected fast dynamics. This myth has its roots in the engineers' infatuation with state-space models. Since state-space models are what we know, the physical realities are expected to accommodate our needs.

Unfortunately, physics doesn't comply. There is no physical experiment in the world that could distinguish whether I drove my car into a tree, or whether it was the tree that drove itself into my car.

The Physical Systems Modeling Laboratory concerns itself with the design of modeling methodologies and tools that reflect more closely the physical realities than the traditional state-space descriptions. The development of the object-oriented modeling methodology has become the central aspect of this research activity. The bond graphs, a dialect of object-oriented modeling, turned out to be excellently suited for this endeavor.

This research activity is a collaborative effort involving a number of partners from different countries. Dynasim AB, a Swedish company located in Lund, is responsible for the design, development, and implementation of the object-oriented modeling tool Dymola, as well as its two graphical front and rear engines Dymodraw and Dymoview. The Deutsche Luft- und Raumfahrt GmbH (DLR) is responsible for the design, development, and implementation of the associated simulation engine Dymosim. This collaboration has already led to a large number of publications, including 1 book, 3 chapters in books, 6 journal articles, 5 keynote addresses at conferences, 15 contributed conference papers, and 11 MS theses. Recently, a consortium of research institutes was created to further the advancement of this technology including, beside from the three partners mentioned earlier, also the Lund Institute of Technology (Sweden), the developer of another object-oriented modeling tool, Omola. One of the aims of this newly founded research consortium is to jointly seek sources of funding for future activities in this exciting research area. Recently, one researcher from the Universitat Politècnica de Catalunya (Spain) and another from the Instituto de Investigaciones Eléctricas (Mexico) joined this research effort. We also profitted a lot from a student exchange program with the Universität Stuttgart (Germany), through which we receive every year in the order of 8 excellent exchange students, several of whom have participated in PSML-related activities over the years. Two of them are currently working on PSML-related research topics.

Current Research Projects

Energy Balance Model for Biosphere 2

Biosphere 2 is a closed ecology project located some 35 miles north of Tucson. With its 1800 sensors being recorded once every 15 minutes (on the average), Biosphere 2 is the largest closed ecology system ever built by mankind. Its aim is to study and understand the dynamic relationships between living organisms and their environment. In this light, Biosphere 2 comprises a large-scale modeling effort.

In order to maintain the physical conditions necessary for life to be sustained (temperature, pressure, humidity, CO2 contents in the air, to mention just the most important ones), Biosphere 2 contains a very complex air circulation system. In this light, Biosphere 2 is one of the most complex technological systems ever built by mankind.

Accordingly, Biosphere 2 represents extraordinary challenges to both science and engineering. With its new management (Biosphere 2 has recently become a National Laboratory managed by Columbia University) came a somewhat modified charter. Had Biosphere 2 been viewed previously as an experimental setup for the study of closed ecologies, it is now seen more as an instrument to further the understanding of the relationships between living organisms and their environment, an instrument that shall ultimately help us navigate our spaceship Earth from an operational mode of constant expansion and exploitation into one of equilibrium and sustainability.

With this modified charter came a different way of using the instrument. Had before Biosphere 2 been used as one experimental setup, it has now become a laboratory, in which multiple experiments can be performed simultaneously and in parallel. This creates new challenges for engineering. Before, it had sufficed to keep the physical parameters of Biosphere 2 within acceptable ranges. Now that experimentation has assumed a more active role, it is important to understand beforehand and be able to predict what effects engineering decisions once made will have on the physical parameters of Biosphere 2. By subdividing Biosphere 2 into smaller units for parallel experimentation, the independent control of each of these entities has become much more difficult.

This is where our involvement begins. We wish to model the physical properties of Biosphere 2 such that controlled experiments can be simulated beforehand with high accuracy and reliability. To this end, we are creating a dynamic energy balance model based on bond graphs.

The faculty involved in this project includes François Cellier from the University of Arizona and Angela Nebot from the Universitat Politècnica de Catalunya in Spain. Another researcher involved in this project is Francisco Mugica from the Instituto de Investigaciones Eléctricas in Mexico. Students currently involved in this project include Steve Pitts, a Ph.D. student of the University of Arizona, formerly working at Biosphere 2, and Julia Miersch, an exchange student from the Universität Stuttgart in Germany.

Multienergy Systems Modeling

This project is thematically related to the one described above, yet takes a somewhat broader and less application-specific perspective. The problem is related to modeling energy conversion between the thermal and non-thermal domains. Thermodynamics are well understood only in the vicinity of flow equilibrium. The models currently used to describe the dynamic exchange of energy between thermal and non-thermal domains farther away from equilibrium are crude, inaccurate, and often experimental, i.e., not reduced to the foundations of physics.

This is a long-term effort, for which we have received funding over the years from various sources, but mostly from NASA through our Space Engineering Research Center (SERC). In this context, we have been working on a prototype of an oxygen production facility for planet Mars. In such an effort, it is essential to have a rather precise idea about the overall energy consumption of the facility through a detailed model of all sorts of energy conversion phenomena taking place, and of heat losses occurring in the process. It is expected that a consequent application of bond graph modeling techniques can help us in this effort. Right now we are studying the detailed energy conversion mechanisms taking place in a pressure cooker (a simplified version of energy conversion phenomena as they occur in many physical devices, such as dearators and distillation columns). Involved in this effort is currently only one student, Robert Hamilton.

Electronic Circuit Modeling

This effort relates partly to the creation of an alternative to Spice for circuit modeling. The reasons for doing this are manifold. (i) Spice is not well suited for dealing with discontinuities (electronic switches). Evidently, it is possible to model the dynamics of the switching itself, but this makes the simulations run slowly, usually without providing additional insight. (ii) Spice is unnecessarily restricting in allowing the definition of new elements. (iii) It is often desirable to perform mixed electronic and other simulations (e.g. mechatronics). Spice doesn't lend itself well to such types of simulations. The object-oriented modeling paradigm addresses all of the above problems elegantly and efficiently.

A second aspect dealt with in our research are the thermodynamics of electronic circuit components. We are looking in detail at the mechanisms of how electronic components heat up as a consequence of current flowing through them, and what the thermal effects on the electrical characteristics of these devices are.

At the current time, there are two students working on this project. Farah Bates is looking at object-oriented MOSFET models, whereas Michael Schweisguth is working on a bond graph model of the thermodynamics of bipolar junction transistors.

Modeling of Multibody System Dynamics

This research deals with object-oriented descriptions of mechanical devices for simulation and control. A research project that is only just starting deals with improvements of the control performance of a mechanically suspended and gyroscopically stabilized movie camera for shooting from a helicopter. This project is being performed in collaboration with Dr. Hal Tharp from the Electrical Engineering Department and some folks from the Optical Sciences Center. This work is being performed on request by a company in California.

Recently we completed a study on vehicular dynamics requested by Ericsson Infocom from Sweden. This study was performed in collaboration with Dr. Parviz Nikravesh from the Aeronautical and Mechanical Engineering Department. The actual work was done by by Peter Schröder, another of our exchange students from the Universität Stuttgart in Germany. The results of that study were presented in ( Schröder, 1995).