Object-oriented models for advanced automation engineering

Emanuele Carpanzano

1999, Control Engineering Practice

visibility

…

description

13 pages

link

1 file

Abstract

The paper focuses on how the object-oriented paradigm can be exploited as a valid modelling technique in many engineering fields and, in particular, in advanced automation engineering. The most typical features of object-oriented modelling techniques include the encapsulation of data, functions to access such data into objects, and the hierarchical composition of objects. On the other hand, the large majority of complex systems, like industrial plants and distributed control systems, can be decomposed into a hierarchical structure that facilitates system comprehension, design and control. For the above reasons, the object-oriented modelling technique is suited to real world modelling, in particular to representing the hybrid dynamics (continuous-time, discrete-time and discrete-event) of industrial plants.

Figures (13)

Fig. 1. A typical module for physical system modelling.

In Fig. 4 the OMT/UML diagram representing the basic classes of the proposed object-oriented modelling framework is sketched. In particular, the hierarchy among the represented classes is outlined by illustrating the aggregation relationships through the proper connector defined by the OMT/UML formalism.

Fig. 3. Object-oriented representation for the function-block model.

Fig. 2. Function block model.

Fig. 5. Model of the overall control system with sensors and actuators.

In order to evaluate the performance of a given control system realistically, through analysis or simulation, it seems reasonable to define a physical model for the process under control and a physical model for the control system (i.e., the overall hardware and software architecture, along with possible communication nets), and to properly integrate the two sub-systems. To this aim, in general, it is necessary also to model the physical behaviour of sensors and actuators. Then, exploiting previous definitions, the model of the whole control system can be sketched as in Fig. 5. Clearly, if the behaviour of sensors and actuators can be simplified, then they can be modelled without individual modules, simply by appropriately connecting the causal input/output control and event terminals (Fig. 6). Fig. 6. Model of the overall control system without sensors and actuators.

Fig. 9. Link model in MOSES. The FBCad model for the control system is constituted by an FB network representing the overall distributed control system. As an example in Fig. 11 part of the FB network for the supervisor PC software of a specific robot mission (ACQUIRE_SAMPLE) is reported. Composite FBs can be exploded into FB networks, as shown for instance in Fig. 12, where part of the FB network of the composite FB Supervisor is illustrated. For each FB, causal events and data terminals can be distinguished. Moreover, for each FB an ECC with algorithms associated to its states is defined, as previously discussed. In Fig. 13 the ECC of the FB Transport of Fig. 12 is reported; here transition conditions and priority attributes are associated with the edges. Such information is essential to properly define the algorithm responsible for the scheduling operations required to correctly simulate the control system. More about the modelling and implementation details of the considered application can be found in (Carpanzano, et al., 1998). The OOM concepts discussed in the previous sections have been strongly exploited in this

Fig. 7. Overall control system model in MOSES. environments mentioned; therefore the interested reader is referred to the existing literature (Carpanzano, et al., 1998; Maffezzoni and Girelli, 1998); this paper only discusses how the OOM concepts are exploited by such tools. In Fig. 7 the model of the control system built in MOSES is shown: in particular, the model of the six- link industrial manipulator and the FBCad model of the controller can be distinguished. The behaviour of sensors and actuators has been simplified. As a consequence, they are modelled without individual modules, simply by connecting the causal control terminals, as discussed in Section 5 with reference to Fig. 6. Moreover, no events have been considered for the physical system dealt with. Actually, no event terminals are defined for the SMART-3S model. The last one is composed by six aggregate models LINKi, 1Sis6, and by the simple model base (Fig. 8). Each link is an aggregate of a rigid body and a direct drive joint (DDJ), i.e. a revolute joint equipped with a dc motor and a voltage-controlled power supply (Fig. 9). In the models there are causal control terminals (i.e. input (Con_in) and output (Con_out) control terminals), and two types of acausal physical terminals (i.e. mechanical terminals (Me), that export linear and angular positions, velocities and accelerations, forces and torques, and electrical

In the present paper the concept and the features of object-oriented modelling (OOM) are discussed, with reference to the problem of control system modelling and design in advanced automation engineering. Although with different characteristics, it is shown how OOM is useful in the definition of general classes of both process and _ control components. Specifically, the object-oriented approach is exploited to define models capable of capturing the hybrid dynamics of complex physical systems and control-system devices. Such a powerful paradigm makes the definition of new design methodologies and the adaptation of existing ones easier, allowing control engineers to deal with Fig. 11. FBCad: part of the PC supervisor FB network for the mission ACQUIRE_SAMPLE

Fig. 10. Equations of the rigid body model in MOSES. 7. Concluding remarks

complexity, and so keeping control of the quality of design phases and _ predicting overall plant performance. Future work includes the definition of suitable interpreters of the hybrid modular models presented here and the integration with future related standard, e.g. IEC 1804 on the definition of function blocks for process control. Fig. 13. FBCad: execution control chart for the FB Transport. Acknowledgements. This work has been supported by the MURST Project "Ingegneria del controllo".

Fig. 12. FBCad: part of the FB network for the composite FB Supervisor. Department of Automatic Control. Lund Institute of Technology. ° Arzén, K. and C. Johnsson (1996). Object-oriented