Entwicklung einer Methode zur objektorientierten Spezifikation von Steuerungen

The still ongoing trend of increasing the level of automation within manufacturing systems and the also growing ratio of decentralized control systems leads to a stronger impact of software-based functions in these control systems. Looking from the programming point of view, it can be observed, that for developing these control applications within manufacturing systems heterogeneity is the main issue which has to be overcome. Migration of further technologies coming from the area of information technology (IT) enforces this kind of heterogeneity. Concerning these obstacles an integrated method for developing cost efficiently these control systems is needed. For a very similar purpose object-oriented methods based on the Unified Modelling language (UML) were developed for different domains within the IT. These methods were chosen as a starting point for the new method (Object-oriented Method for Developing Technical Multi-Agent-Systems, ODEMA) should be developed within the present PhD thesis. As a first step, present techniques for specifying and programming control systems were analysed, also as the already mentioned object-oriented methods for developing embedded realtime-systems were taken under consideration. A defined set of features, the so called method-axioms, was applied throughout analysis work as requirements which served as a proofing guideline for different description languages and methods. As a result it was shown, that the description languages for control systems can be seen as graphical programming languages (e.g. IEC 61131). Therefore the architectural aspects of control systems (system-decomposition and structure) could not be described sufficiently. It was also derived that description languages which describe causal and timed behaviour of control systems can be found also as diagrams within UML, which therefore can be understood as a toolbox for the approach of this PhD thesis. As a second step basic models which describe architectural aspects of control systems, were analysed. These basic models cannot be used as description languages, but they can be seen as mental models used for illustration of basic principles. Therefore they must be integrated as a part of the UML-based description language, which can be accomplished by extension mechanisms the UML is offering. Following this rough approach UML CASE-Tools can also be applied by ODEMA. Based of the understanding, that a method consist of a procedure and description language, conception and developing of the ODEMA-method was distinguished in two different lanes. At first, a procedure model which describes steps, products and participating roles was derived and assembled to a iterative, incremental, use case driven and architecture centric procedure model. Integrating the basic models, mainly the models of production-agent, functionblock and control-component, into the description language of ODEMA was the second lane resulting in a platform-independent UML-Profil. Procedures for proofing consistency of syntax and semantics of a particular specification cover both lanes and are also part of ODEMA. Finally, the method was validated by applying a case study which specifies holonic controlled material flow. As first result of validation it can be pointed out, that UML lacks instantiation of UML-Functionblocks modelled by UML-subsystems and a formalism for describing operations on data and objects (action language). Also a set of basic and platform-independent data types and physical units is missed. These open issues concerning to the UML-Standard would probably be solved by the next releases of the UML. Moving the UML forward to a formal description language is a further step requested by application domains faced by safety requirements of software-based systems. Future UML releases would not cover this issue as far this can be foreseen concerning the actual UML roadmap. Therefore additional techniques were developed so far. The so called reference model based on transformation and transition systems is an example for this purpose. As conclusion, this PhD thesis points the way to an object-oriented method for specification of control systems within manufacturing systems as an important step towards a common method overcoming heterogeneity