Special issue on object-oriented modelling and simulation
The field of object-oriented modelling and simulation (OOMS), combining equation-based a-causal modelling with object-oriented concepts borrowed from computer science, has been evolving for two decades and is now well- established as the most appropriate framework for the dynamical modelling of complex engineering systems at the system level.
The success of OOMS builds mainly on two complementary features. On the one hand, the equation-based approach provides unparalleled flexibility, customizability and development ease when modelling innovative systems (or parts thereof), no matter which domain they belong to. On the other hand, the object-oriented paradigm strongly promotes the reuse of modelling knowledge, thus allowing us to focus the model development effort on the innovative components and processes, while leveraging extensively on libraries of well-tested models for more conventional parts of the system. The combination of these two factors is particularly attractive in the context of academic and industrial research, even more so when dealing with heterogeneous, multi-domain systems.
Software tools that handle object-oriented models are increasingly available and have now reached maturity, as far as the generation of simulation code is concerned. Much work remains to be done, both at the theoretical and at the implementation level, to support the many other possible usages of the mathematical models that go beyond the simulation of specific transients.
First, improved high-level debugging and error reporting is required. For example, the causes of numerical errors should be traced back to the original object-oriented model in a comprehensible and user-friendly way, so that people who do not have a Ph.D. in modelling and simulation techniques can also deal with such situations, which arise all of the time during everyday practice. In parallel to that, more robust techniques are required for the initialization of object-oriented models, which is currently a weak point in object-oriented tools and techniques in many application contexts.
Currently available tools already allow us to use experimental data to tune some uncertain parameters in the model, by optimization techniques. However, better algorithms and tools are needed to support the end user understanding of to what extent the result of this activity is meaningful, for example, by assessing the identifiability of the uncertain parameters, and also to help him or her in the design of experiments that improve the quality of the parameter estimation.
Object-oriented models are represented in a completely symbolic way: this allows us to perform a wide range of model transformations and model analysis activities, which go beyond the simulation of specific transients. For example, model order reduction (MOR) can be extremely useful for a range of applications, particularly in control system design. Also the transformation of the models into specific control-oriented formalism (symbolic transfer functions, linear parameter-varying systems, linear fractional transformations, etc.) could be extremely useful for control-related applications.
Last, but not least, object-oriented modelling tools should be integrated with tools for object-oriented system requirement specification (e.g. using the UML language), allowing us to perform the automated checking of the conformance to the specification, and with the ultimate goal of automatic generation of systems conforming to the specifications.
This special issue on OOMS collects seven papers. The first four papers are somewhat classical, the first two describing Modelica libraries for specific applications and the second two describing models for control system assessment by simulation. The last three papers try to go beyond modelling for simulation, suggesting the adoption of object-oriented models for knowledge-based engineering design and plant diagnosis and surveying non-linear MOR techniques applicable to object-oriented models for dimensioning and direct control design.
The first paper, ‘Dynamic modelling of PEM fuel cells using the FuelCellLib Modelica library’, by M. A. Rubio, A. Urquia and S. Dormido, illustrates the Modelica library FuelCellLib, for dynamic modelling of proton exchange membrane fuel cells (PEMFC). The library allows for a very detailed modelling of PEMFC, considering several physical–chemical phenomena taking place in the membrane (transport of water in liquid and steam phase), in the catalytic layer of the cathode (transport of water in liquid and steam phase, transport of oxygen in steam phase, proton and electron conduction and electro-catalytic reaction) and in the diffusion layer of the cathode (transport of water in liquid and steam phase, transport of oxygen in steam phase and electron conduction). A finite volume method has been applied to discretize the PDE in the spatial coordinate perpendicular to the layers. Models have been validated with respect to results reported in the literature and with respect to the experimental data. In the latter case, the GAPILib Modelica library has been used to estimate the values of the model parameters based on the experimental data, by applying genetic algorithm techniques.
With the aim of achieving a good readability and usability, a new Modelica library for thermodynamic systems (TIL) has been developed and described in the paper ‘Modelling of heat pumps with an object-oriented model library for thermodynamic systems’, by M. Gräber, K. Kosowski, C. Richter and W. Tegethoff. Readability is actually pursued by limiting inheritance as much as possible, as already suggested by other authors but, on the contrary, some violations of a fully modular approach are introduced. For example, fluid properties are distinguished into five different types and component models are explicitly written for one fluid type. Moreover, the inner/outer connection of components is extensively used, even to propagate, through a centralized System Information Manager, the derivative of pressure, computed by a non-physical model (in this way numerical efficiency should be gained, because pressure is no longer a state variable in each control volume). The library has been designed mainly for refrigeration, air conditioning and heat pump systems and has actually been used in this article for the modelling and simulation of a novel CO2 heat pump system for domestic hot water supply; simulations have been compared with test stand measurements.
The paper by L. J. Yebra, M. Berenguel, J. Bonilla, L. Roca, S. Dormido and E. Zarza, ‘Object oriented modelling and simulation of ACUREX solar thermal power plant’, describes the current status of a research focused on modelling and simulation of a Parabolic Through-Collector Solar Power Plant, installed at the ACUREX facility at the Plataforma Solar de Almería. The model will be used in the design of hybrid model predictive control and intelligent control schemes to optimize plant performance, even under start-up and shutdown manoeuvres. The modelling approach, while being essentially based on the ThermoFluid library, largely exploits the object-oriented features of the Modelica language, such as polymorphism, inheritance and class parameterization. Also, the use of the Modelica StateGraph library for discrete and reactive system modelling needs to be mentioned, for the implementation of a hybrid automatic operation and control system, including states and transitions for the whole daily operating range.
With the aim of studying the temperature control of the starch mashing phase, which is crucial for the quality of beer, a literature mashing model is first extended with energy balances in the paper ‘Object-oriented modelling of starch mashing for simulation-based control studies’, by A. Leva, F. Donida and M. Maggio, and then coupled to a control representation of scalable detail, including a quasi replica of the control code. A Modelica hybrid model is therefore obtained, suitable for both control strategy and process instrumentation assessment. In particular, the flexibility of the object-oriented approach allowed them to easily obtain different models of the mash/vessel/heating compound, with various topologies and detail levels, with a significant code reuse.
The declarative definition of models, their port-based interconnections, as well as the object-oriented features of the Modelica language, in particular the reusability of components obtained by library development, not only makes it easier to develop models of complex system but may also establish the basis of a knowledge-based engineering design. This concept is exploited by M. J. Foeken and M. Voskuijl in the development of a simulation model generation method, as a part of an automated control software development framework, described in the paper ‘Knowledge-based simulation model generation for control law design applied to a quadrotor UAV’. To support the knowledge base, a language ontology for Modelica models has been created and then concepts and relation have been restricted to ensure structural correctness. The method is applied to the generation of the model quadrotor UAV, a complex and multi-disciplinary mechatronic system, which well represents the need of an integrated system and control design.
A ‘New method to assess tube support plate clogging phenomena in steam generators of nuclear power plants’ is described in a paper by D. Bouskela, V. Chip, B. El Hefni, J. M. Favennec, M. Midou and J. Ninet. The method is based on a 1D Modelica model of the steam generator and its control system and yields a global estimation of the tube support plate clogging ratio, on the contrary of usual methods, which give a local estimation of the clogging ratio based on local measurements or visual inspections during the yearly outage of the plant for refuelling and maintenance. The model is used to compute the response curves of the steam generator to a particular transient (routinely performed in normal operation, i.e. during the periodic testing of control rods), to be compared with real response curves measured on-site; from this comparison an estimate of the clogging ratio is obtained. It must be pointed out that a dynamic analysis is needed because the clogging ratio is reflected by a delayed response of the steam generator behaviour to a decrease in the primary thermal power. More in general, tube support plate clogging may reduce the cooling efficiency and modify the dynamic behaviour of the steam generator, leading to possible safety issues.
Object-oriented concepts and tools may speed up to a great extent the process of an accurate analysis by simulation of complex multi-domain technological systems but, on the contrary, for sizing and, above all, for control system design, the need for tools capable of extracting from the overall dynamic model reduced models, representing the dominant behaviour, has emerged. In fact, so far the use of such models to support control-related activities has been mostly limited to the verification of the control system performance in closed loop by simulation, rather than giving direct support to controller design. A wide array of literature in the field of MOR methods for linear systems has been produced in recent years but the results cannot be directly applied in general to models of real technological systems, which are usually nonlinear. In the paper ‘Model order reduction for object-oriented models: a control systems perspective’, by F. Donida, F. Casella and G. Ferretti, recent results in non-linear MOR, mainly originating from research in computer-aided analysis and design of electronic circuits, are reviewed and put in a control design-oriented perspective, pointing out interesting future research directions.
Guest Editors
Gianni Ferretti
Francesco Casella
Politecnico di Milano,
Dipartimento e Elettromica e Informazione
Piazza Lconards
Da Vinci 32, 20133, Milan Italy