First Workshop on Industrial Automation Tool Integration for Engineering Project Automation Integrated Design of Simulation Models for Passive Houses Petr Novák∗ ,† Radek Šindelář∗ ∗ Christian Doppler Laboratory for Software Engineering Integration for Flexible Automation Systems, Vienna University of Technology, A-1040 Vienna, Austria {novak,sindelar}@ifs.tuwien.ac.at † Department of Cybernetics, Faculty of Electrical Engineering, Czech Technical University, Prague, Czech Republic Abstract—Modern automation systems require both design- Simulation Floor plan Library time and runtime integration of diverse engineering tools. Tra- HouseBuilder Bldsimlib Simulation library file ditional integration approaches are based on repeating manual Config.xml Simulation work, being time-consuming and error-prone. In this paper, file library file applications of semantic integration, dealing with meaning House Builder of objects and their interfaces, is explained and shown on Config File Expert Assisted a real industrial use-case. Simulations are useful tools for Simulation Tool Parser process optimization or performance testing and the presented Plant ontology Simulation methodology makes their design for particular industrial plants individuals library features flexible. The use-case shows that the design of simulation Simulation models for passive houses can be user-friendly and feasible even Plant Ontology Library for non-experts as it is based on a graphical tool that enables Ontology to draw a passive house floor plan. Since neither this tool nor OWL OWL a universal simulation library, comprising atomic simulation Ontology Ontology blocks, were intended for simulation purposes, the presented methodology is a typical example of tool integration having Semantic Engine heterogeneous data models. The goal of this paper is to propose an ontology-based Simulation formalization of knowledge representing structures of real model file industrial plants and simulation models. The paper also intro- Simulation duces the design of simulation models for passive houses from Model other engineering sources, which can be used by non-experts for simulation modeling. The practical usage is restricted by the fact that simulation parameters must be entered manually. Figure 1. Workflow presented in this paper. The main contributions of the paper are the proposed structure of an automation ontology and a workflow of simulation model design that is not common in engineering disciplines. Keywords-Semantic integration, simulation model, passive simulations. The presented approach is based on Semantic house, ontology, automation system design phase. Web technologies. Knowledge about the tools under inte- gration is stored in ontologies. Ontology-based querying and I. I NTRODUCTION reasoning techniques are used to retrieve the information and Simulation models emerged as a very efficient way to op- derive new pieces of engineering knowledge. The proposed timize process operation. They can be used for performance solution realizes an ontology-based middle-ware layer be- testing of control algorithms or the whole industrial systems tween the tools. A use-case project, dealing with a design under both normal and extreme conditions, as well as for of simulation models for passive houses, is described in many other engineering tasks. Nevertheless, several issues this paper. It motivates the research, the examples from this dealing with simulation models have not been satisfactorily domain are given, and in the final part, this use-case project solved. The integration of simulation models and the coop- is discussed and evaluated. eration with other engineering tools still remain problems The goal of this paper is to present a formalization of real as well as the high requirements on engineering knowledge plant data in a machine-understandable way and to show and skills to create and configure them. Therefore, simu- the benefits of semantic integration of heterogeneous data lation models are usually designed and performed only by models on a passive house model use-case. simulation experts. The workflow of the presented use-case project is depicted This paper contributes to improve the simulation model in Fig. 1. The entries for the presented methodology are design phase, and it shows how engineering sources can be a representation of a graphical floor plan of a particular used to enhance simulation model usage for non-experts in passive house created in “House Builder WPF” software 13 First Workshop on Industrial Automation Tool Integration for Engineering Project Automation and a universal simulation library “Bldsimlib” 1 comprising The usage of ontologies in modeling and simulations is generic simulation blocks for passive house simulation [1]. introduced e.g. in [6]. In [7] the Ontology-driven Simulation Both these entries are suitable examples of engineering Tool (ODS) is described. The approach is based on two sources having heterogeneous data models that must be ontologies: a domain ontology categorizing a knowledge integrated involving meaning of data and interfaces. Floor including a problem vocabulary and its concepts are mapped plan of House Builder WPF software originally solves for onto a modeling ontology being used for the simulation visualizing passive house runtime data in a tool House model description. Our approach distinguishes between plant Viewer WPF, but in our approach, we also use the same domain and simulation knowledge in a similar way. file without any modification for defining so-called plant An ontology-driven simulation model design is presented ontology individuals, i.e. ontology-based representation of in [8]. The paper is focused on generating MATLAB- the passive house structure. Simulation blocks and their Simulink blocks and defining them via DAVE-ML according features are formalized in a so-called simulation ontology. to the domain ontology. Connection of these blocks is Consequently, plant and simulation ontologies are used to done manually, thus this approach is complementary to the semantically create a simulation model for a passive house. methodology explained in this paper. The remainder of this paper is structured as follows: the The methodology presented in this paper uses so-called second section summarizes a related work. The third section power bonds and signal bonds to classify a type of device formulates research issues that are addressed in the fourth interconnection. These terms originate from a bond graph section, summarizing a methodology for formalization of theory, that is introduced e.g. in [9]. Power bonds are plant, simulation and other engineering knowledge in gen- common in real physical systems where the flow of energy eral, and in the fifth section, describing the use-case project defines power transmissions, whereas signal bonds refer to and its results. The sixth section concludes and proposes interconnections where energetic interactions can be ignored. further work topics. For example, there is usually assumed that sensors have no impact on measured variables, thus this kind of relationship II. R ELATED W ORK is called signal, whereas e.g. tanks and outlet pipe interact Semantic integration is a perspective way to integrate by power bonds. diverse systems and tools, which is based on data meaning. Semantic level stands on top of a technical integration level, that is concerned with data transfers. Further explanation can III. R ESEARCH I SSUES be found e.g. in [2]. Although the semantic integration can be implemented in many ways, the wide-spread approach is The problems, which are discussed in this paper, can be based on Semantic Web technologies, especially representa- summarized into following research issues. tion of knowledge in ontologies. RI-1: Formalization of plant and simulation knowl- The term ontology, originating from philosophy, is in edge in general. As real plants have diverse structures and engineering defined in many ways [3]. One of the most devices, there is a need for formalizing their description. cited definitions is by Gruber: ”An ontology is an explicit Such a formalization is useful for reusability, flexibility in specification of a conceptualization” [4]. In the presented terms of process redesign, and automatic or semi-automatic approach, ontologies are represented in OWL DL2 format methodologies for supporting both design and run-time that provides a suitable compromise between expressive phases of the automation system lifecycle. power and performance of reasoning. We use ontology RI-2: Applications of the formalization for design querying language SPARQL3 and ontologies are managed of simulation model for a particular passive house. from Java code via framework ARQ4 , providing query The particular use-case project integrates the two stand- engine on top of Jena API5 . alone engineering tools. A universal library Bldsimlib is Although for simulation integration general-purpose tech- implemented in MATLAB-Simulink6 and it comprises so- niques such as DCOM, CORBA, J233 could be used [5], called generic simulation blocks. They approximate building there exist frameworks including standard vocabulary for elements, such as windows, walls, doors, or rooms. Usually, simulation integration, such as DIS, SEDRIS or HLA [6]. only simulation experts are able to create and perform the Especially HLA framework is widely cited, but it does not simulation model. We try to overcome this shortcoming by support integration on semantic level. Therefore, we focused integrating the second tool, the graphical House Builder on ontology-based approaches. WPF application. The XML config file, being its output, is used to recognize a structure of the house. It is stored in 1 Acronym of “Building Simulation Library”. a machine-understandable form and consequently, a simula- 2 http://www.w3.org/TR/owl-features/ 3 http://www.w3.org/TR/rdf-sparql-query/ tion model is generated semi-automatically. 4 http://jena.sourceforge.net/ARQ/ 5 http://jena.sourceforge.net/ontology/ 6 http://www.mathworks.com/products/ 14 First Workshop on Industrial Automation Tool Integration for Engineering Project Automation Domain of real industrial plants controls Device “A” Domain of control Device “B” and measuring systems signal_bond measures Room Wall Room Device “C” power_bond room1 wall1 room2 simulates affects Domain of Figure 3. Comparison of interconnections in terms of plant description Domain of simulations and signal-oriented simulators. disturbances Figure 2. Formalization of problem domains and relationships of their elements. of the bonds is realized while assembling the simulation model in Java code, performing series of SPARQL queries. As we defined several domain ontologies that describe the IV. F ORMALIZATION OF P LANT, S IMULATION , AND real plant and simulation, the relationship between these on- OTHER E NGINEERING K NOWLEDGE tologies are realized via relations summarized in the previous paragraph. All of the ontologies and their relationships build Our approach is based on explicit distinction between a so-called automation ontology, being depicted in Fig. 4. plant, simulation, and other engineering knowledge. Plant The rectangular blocks represent ontology classes, whereas knowledge is related to existing devices and elements, the rounded blocks are ontology individuals. For better read- whereas simulation knowledge comprises features of avail- ability, some individuals are omitted, but the fundamental able simulation blocks, e.g., their interfaces. We store plant ideas are covered by this figure. We can see the classes filled knowledge in a so-called plant ontology, and simulation in blue color that represent an upper layer, shared within knowledge in a so-called simulation ontology. Further, we diverse automation systems. The plant ontology comprises define supplemental ontologies such as a signal ontology. real industrial plant devices, which are categorized into Specification of fundamental domains and their relation- five classes. “Actuator” class involves controllable devices, ships is introduced in Fig. 2. The upper left set represents “Passive Element” represents uncontrollable devices. “Dis- a real plant domain. The figure depicts that real plants turbance” affects real plant, it defines boundary conditions comprise devices, that are connected by two basic domain- and it is often non-measurable and random. “Measure point” specific terms: “signal bond” and “power bond”, that we defines sensors or softsensors that are software algorithms use in compliance with bond graph theory to define the calculating a value from other variables. Blocks filled in real device interconnections. We consider a measuring and yellow color depend on a type of a particular plant, in control system as one domain, which interacts with the real our case an example passive house classes are depicted plant domain via properties “measures” and “controls”. We (further classes were omitted for better readability). The explicitly define disturbances, i.e. factors that influence the simulation ontology comprises the description of available real plant and that are usually not desired from a control simulation libraries and final simulation models including point of view. For example, when controlling a temperature their interfaces. Since industrial devices and tools usually in a house, disturbances are the sun, humans, or opening and requires information about not only the connections, but closing doors as they impact on the controlled variable. Sim- port numbers as well, the individuals of “Port” are depicted. ulation models are interconnected with a real plant via the They represent all available ports and specify their signal relation “simulates”, expressing that some real plant device types. Last but not least, “Power Bond Decomposition” is approximated by the particular generic simulation block. labels doubles of ports that decompose power bonds, i.e. According to our industrial experiences, this relationship is they define signal routing in signal-oriented tools. not usually 1:1 but 1:n, i.e. one plant device can be simulated Although a structure of the automation ontology could by more than one simulation blocks. seem complicated at first, it provides powerful support for An important issue for simulation model design is a de- diverse engineering tasks. Furthermore, it is not expected to composition of power bonds for signal-oriented simulators, be modified by hand in a general-purpose ontology editor, such as MATLAB-Simulink. Figure 3 depicts a description but either in specialized editors implemented for automation in the plant ontology and the corresponding schema in purposes or via specialized tools such that a user does not MATLAB-Simulink. We can see that each power bond is interact with this ontology at all. The latter case is used in decomposed into two signal bonds, i.e. two interconnections the passive house use-case presented in this paper, as the in the signal-oriented simulator. In our approach, translation plant ontology is created semi-automatically by the parser 15 First Workshop on Industrial Automation Tool Integration for Engineering Project Automation owl:Thing Plant Simulation Port Signal Passive Measure Simulation Power Port User- Actuator Disturbance Controller Element Point Library Decomposition defined PB Wall 1 PB Wall 2 State Room Room HVAC Unit Envelope State Flow Elements Bldsimlib Room input Port Wall Input 1 signal input Port Wall Input 2 signal Wall Room Exterior Bldsimlib Wall output Port Wall Output 1 signal room1 simulates output Port Wall Output 2 signal simulates Figure 4. Automation ontology: Machine-understandable knowledge-base formalizing industrial plant, simulation and other engineering knowledge. Figure 5. An example of room description in a House Builder WPF configuration file config.xml. of a passive house floor plan. Although the description of in a Java loop taking the free ports and decomposing the simulation model blocks, i.e. a simulation ontology, a port power signals by doubles of signal interconnections in case ontology and a power bond ontology must be entered by of signal-oriented tools. humans, we are implementing a tool that supports this work and makes it easy and user-friendly. V. U SE - CASE P ROJECT: I NTEGRATED D ESIGN OF Since simulation parameters, i.e. parameters required for PASSIVE H OUSE S IMULATION M ODEL the parametrization of simulation blocks, can differ from The goal of this section is to provide a solution for parameters of real plant devices in both count and scale, the research issue RI-2, i.e. to create a simulation model every generic simulation block has the parameters described for a passive house, whose floor plan is drawn in House in the simulation ontology. A special kind of parameters Builder WPF software. The simulation model is created by are initial conditions that are managed in a similar way selecting and interconnecting generic simulation blocks from as parameters in our approach. Known parameters of real a universal library Bldsimlib, implemented in MATLAB- plant devices are stored in a plant ontology. Nowadays, the Simulink. This task is challenging as House Builder WPF translation of plant parameters to simulation parameters have and the universal simulation library have heterogeneous data not been satisfactorily solved in the presented prototype. models. Therefore, the semantic engine and the parser are referred Figure 6 depicts a use-case passive house floor plan in as semi-automatic, meaning that structural issues are solved House Builder WPF tool. The use-case passive house is automatically but the parameters are managed manually. a simplified house, consisting of two rooms, walls and While an automation ontology is created, the seman- interior equipment. As it is very simplified, it enables to tic engine can semi-automatically assemble the simulation show our approach in a clear way and it does not pose model for a particular plant. It finds all real plant devices any restriction in generality. An exemplary piece of the (i.e. individuals of the plant ontology) and according to the House Builder config.xml file is shown in Fig. 5. The ontology property “simulates”, a set of simulation candidates figure depicts the House Builder representation for the left and their interfaces are retrieved for each plant device via room of the passive house floor plan. Rooms can be found SPARQL queries. Consequently, the interconnections are set via keyword RoomPolygon, whereas walls via keyword 16 First Workshop on Industrial Automation Tool Integration for Engineering Project Automation Universal library for environmental quantities modeling of residential buildings Objects in rooms Zones (rooms, exterior, ground) Interactions between zones and ventilation N_man out 1 1 1 1 N_woman Interaction Exhaust fan Human out t (deg C) 2 2 2 2 N_child interaction exhaust K_activity ROOM CO2 (ppm) human 1 1 RH (%) Extender 1 1 2 O/C Window 2 Ventilator p (kPa) {0;1} 2 <0;1> 2 out Heat source ocwindow ventilator heat (W) bldsimlib_room heat 1 out 1 1 1 Window 2 Damper t (deg C) extender 2 2 2 Universal source <0;1> CO2 (ppm) bldsimlib_window damper source EXTERIOR RH (%) 1 1 p (kPa) 2 Interior DOOR Exhaust Gas cooker sun {0;1} 2 Exhaust bldsimlib_door Outlet cooker wind HVAC unit bldsimlib_exterior 1 Source 1 Plant WALL 2 Source 2 t (deg C) Inlet Figure 6. Floor plan of the use-case passive house created in House plant out bldsimlib_wall hvac t (deg C) Builder WPF software. Bath/shower Ground CO2 (ppm) 1 Leakage 1 1 RH (%) 2 (ext) (ext) 2 Ext bathroom leakage p (kPa) p_1 Blower door tester 1 1 1 p_2 bldsimlib_ground Leakage (simple) p_3 2 2 blower_door_tester LinePoint. First step of our solution is a parsing of leakage_s House Builder config.xml file that gives information consequently used to create individuals of passive house Figure 7. High-level overview of a universal simulation library Bldsimlib in MATLAB-Simulink. ontology, i.e. the instances of rooms and walls are created and their interconnections are inserted as object ontology properties. The presented version of the algorithm supports rooms and walls only, but the functionality is planned to be extended for further elements support. Some elements must exceeds the size of this paper, we will address it in future be recognized in a non-intuitive way, as House Builder does work. The high-level overview of the used Bldsimlib library not express them. For example, there is no graphical element is depicted in Fig. 7. The emphasized generic simulation representing doors or windows, but there are so-called blocks are supported in the current version, i.e., simulation sensors measuring and visualizing position of sun-blinds blocks of exterior, room, and wall are used in generated or position of door. This is done by intended functionality model. of House Builder WPF oriented on passive houses, where windows are expected usually equipped with sun-blinds and The semi-automatically generated model of the passive furthermore, the floor plan need not to be absolutely exact house is depicted in Fig. 8. The upper left block represents in all details as it monitors the operation of the whole an exterior, i.e., it defines borderline conditions for the automation system of the house. simulated house. Simulation blocks in the central column The parser reads the config.xml file and for known represent rooms and the blocks in the right part of the figure keywords creates plant ontology individuals. The parsing is approximate walls. The depicted simulation model does not done via Java code and creating individuals is realized by export its outputs into a file or MATLAB Workspace and ARQ/Jena methods. Another alternative would be to express although it would be possible to set all variables as outputs House Builder WPF elements in a specialized ontology and automatically, we expect either the user of a simulation map its concepts onto plant ontology, but the first approach model would define the outputs or in the planned extended was used for this prototype as it is easier and reaches version, the outputs will be defined by ontology property satisfactory results. “measures” as well as the inputs will be entered via ontology While having the passive house representation in the property “controls”. In this use-case simulation model, a plant ontology, that is a tool-independent representation of limitation of the current implementation is apparent: The the object, the semantic engine generates the simulation walls that separate two same areas are not merged into one model. The simulator-independent part of the engine is wall. In other words, each room in the generated model has implemented in Java. It is called from supporting MATLAB four walls, but they could be merged into two walls without script, where simulation blocks and signal interconnections changing the functionality - one wall to exterior and one to are entered via MATLAB API using functions add_block the other room. This issue could be solved on the level of the and add_line. As we assume that there is available a configuration file parser, but we consider this problem being simulation library comprising generic simulation blocks, general. As it occurs in many real-life systems, we plan to the creating of models means selecting appropriate library handle this situation on a plant ontology level. The problem blocks, entering them into a simulation model file, setting exceeds this paper; it is planned for future work. Last but not their name uniquely and interconnecting them according least, the positions of blocks are done by a rectangular matrix to the real plant structure. To run the simulation model, and signal wires use auto-routing available in MATLAB; the also simulation parameters must be added, but as this issue positioning is planned to improve. 17 First Workshop on Industrial Automation Tool Integration for Engineering Project Automation that is the Excel script widely used in civil engineering out 1 t (deg C) out 1 WALL 2 to calculate and evaluate thermal and other properties of 2 CO2 (ppm) t (deg C) t (deg C) wall101 passive houses. Other future work issues are the support for EXTERIOR RH (%) ROOM CO2 (ppm) 1 1 simulation model input and output management, that will WALL 2 p (kPa) RH (%) 2 t (deg C) work not only on the simulation model and visualization sun wall102 wind p (kPa) 1 level but as well on the run-time level. Further future work m101 1 ext 2 WALL 2 topics are a positioning of simulation blocks and signal out t (deg C) wall104 wires to obtain better readability for humans, and automatic t (deg C) 1 1 WALL 2 merging simulation blocks on the ontology level. ROOM CO2 (ppm) 2 t (deg C) RH (%) wall105 ACKNOWLEDGMENTS 1 p (kPa) 1 m102 2 WALL 2 t (deg C) The authors would like to thank to the partners from wall106 the Christian Doppler Laboratory for Software Engineering 1 1 WALL 2 Integration for Flexible Automation Systems for the discus- 2 t (deg C) wall107 sions and feedbacks. This work has been supported by the 1 1 Christian Doppler Forschungsgesellschaft and the BMWFJ, WALL 2 2 t (deg C) Austria. wall103 R EFERENCES Figure 8. Simulation model of the use-case passive house that was [1] P. Novák, “Modelling and Control of Environmental Param- generated semi-automatically by the parser and the semantic engine from eters of Passive Houses,” in POSTER 2010 - Proceedings of passive house floor plan and generic blocks from Bldsimlib library. the 14th International Conference on Electrical Engineering, 2010. VI. C ONCLUSION AND F UTURE W ORK [2] T. Moser, R. Mordinyi, W. Sunindyo, and S. Biffl, Canadian Semantic Web: Technologies and Applications. Springer, This paper describes the usage of ontology-based repre- 2010, ch. Semantic Service Matchmaking in the ATM Domain sentation of plant, simulation, and other engineering knowl- Considering Infrastructure Capability Constraints. edge. The information is structured in order to generate sim- ulation models semi-automatically and to support the design- [3] A. Gómez-Pérez, M. Fernández-López, and O. Corcho, Onto- logical Engineering with examples from the areas of Knowl- phase of automation system. The presented methodology is edge Management, e-Commerce and the Semantic Web, second demonstrated on the real industrial use-case project dealing printing ed. London: Springer-Verlag, 2004. with the design of simulation model for a passive house. As the approach is oriented (but not limited) for signal-oriented [4] T. Gruber, “A translation approach to portable ontology spec- simulators, such as MATLAB-Simulink, the attention is paid ifications,” Knowledge Acquisition, vol. 5(2), 1993. into decomposition of power bonds to the doubles of signal [5] J. Hu and H. Zhang, “Ontology based collaborative simulation bonds, supported by such tools. The use-case illustrates framework using HLA and Web Services,” in World Congress generating simulation model structure from the floor plan on Computer Science and Information Engineering, vol. 5, created in the particular graphical tool. As the data models 2009, pp. 702–706. of available simulation library and floor plan configuration [6] L. Lacy and W. Gerber, “Potential modeling and simulation file have different structure, it is shown that the information applications of the web ontology language - OWL,” in Pro- is recognized in the parser and afterwards, plant ontology ceedings of the 2004 Winter Simulation Conference, vol. 1, individuals are created. Consequently, the simulation model Dec 2004. is created semi-automatically by the semantic engine that is general-purpose in terms of applicability on various types of [7] G. Silver, O.-H. Hassan, and J. Miller, “From domain ontolo- gies to modeling ontologies to executable simulation models,” industrial plants. in Proc. of the 2007 Winter Simulation Conference, 2007, pp. The proposed methodology guarantees avoiding structural 1108–1117. errors, reduces manual and error-prone work, and saves development time and costs. The restriction of the presented [8] U. Durak, S. Güler, H. Oǧuztüzün, and S. K. İder, “An methodology is a need for entering simulation parameters exercise in ontology driven trajectory simulation with MAT- LAB SIMULINK (R),” in Proceedings of the 21th European manually as well as some pieces of information are not cov- Conference on Modelling and Simulation, Prague, 2007. ered in House Builder WPF, such as a topology of the HVAC system. In future work, we plan to extract the parameters [9] P. Gawthrop and G. Bevan, “Bond-graph modeling,” Control from a so-called Passive House Planning Package (PHPP)7 Systems Magazine, IEEE, vol. 27, no. 2, pp. 24 –45, 2007. 7 http://www.passiv.de/07 eng/phpp/PHPP2007 F.htm 18