The main trigger for the future vision of Industry 4.0 is the global trend towards individualized mass production (batch size 1) in conjunction with the resulting increase in variant diversity and shortening product life cycles. However, the degree of automation is still low today, especially for lower quantities, as the corresponding costs for procuring and setting up the necessary equipment counteract the wage advantages of automation. The effects in the area of assembly are particularly serious, as this represents the largest cost factor within production.

In order to meet these challenges, it is necessary to enable low-cost configuration and commissioning of machines and systems. This applies equally to new planning and rescheduling.

While the OPC UA communication standard, which is becoming increasingly popular in the context of Industry 4.0, merely provides the basis for data modeling, the data models and concepts required to implement changeable systems must be described in the form of information models and provided in the form of OPC UA servers, for example.

Describe functionalities in an abstract way

The use of so-called 'capabilities' or 'skills' is a promising approach to standardizing the system integration and control of equipment in production plants. With their help, device functionalities can be described in an abstract manner and are therefore independent of specific domains or manufacturers. While OPC UA provides the necessary data semantics and networking, the functionalities of any equipment can be represented in a manufacturer-independent manner by describing standardized skills. In addition to simplified rescheduling and replanning, this concept enables standardized control of individual equipment.

A striking and frequently used example of a skill is 'Move object'. This initially only describes that an object is to be moved to a certain position. This can be carried out by an industrial robot from any manufacturer (manufacturer-independent), but also by other equipment such as a conveyor belt, an overhead crane or a person (domain-independent). The challenge, however, is to standardize these skills in such a way that the parameters they contain can be used for all equipment.

As part of a working group of the VDMA's Integrated Assembly Solutions (IAS) department, a demonstrator or concept was developed for Automatica 2018 in collaboration with various component manufacturers and scientific institutes, which not only allows assembly processes to be modeled using skills, but also allows the equipment used to be controlled directly via OPC UA based on skills. The main challenges here are the definition of a uniform skill catalog in which the same functionalities of different device classes and manufacturers are described in a uniform manner and the creation of a suitable metamodel to map the control in OPC UA.

The standardized skills can now be used in engineering to model the individual process steps. These can be determined beforehand with the help of an assembly precedence graph, for example. Various levels of detail or composition are conceivable. For example, the 'Pick&Place' skill can exist directly as a process step and be executed coherently by a Pick&Place unit. However, modeling is also possible via a sequential order of the atomic skills 'Move' and 'Grasp'. This depends heavily on whether the skills are provided directly by the various operating resources or by automated compositions - for example by means of higher-level controllers.

The control architecture: Skills offered in OPC UA can be called up by the higher-level station controller via methods.

© VDMA

The assignment between the skills modeled on the basis of the process steps and those offered by the equipment is ideally automated. The first step is to check whether the requirements set by the product and the process can also be fulfilled by the respective equipment. In the case of the 'Move object' skill, for example, the requirements can be the product weight and the required end position. It is also possible to carry out this assignment manually in an engineering environment such as the 'Codesys Application Composer' using the user's implicit knowledge. In both cases, higher-level software can then carry out the automated setup of the equipment and thus drastically reduce the effort involved in planning or rescheduling a system. This achieves a high degree of changeability.

While it is still possible to use a proprietary control architecture with this approach, all equipment can also be controlled directly on a capability basis. As part of the VDMA R+A OPC UA demonstrator, an initially technology-independent metamodel was developed for this purpose, which enables the uniform modeling of directly executable skills of the individual components. One of the most important elements is the description of a generally valid state machine, which enables control by a higher-level system and always reflects the current status of skill execution.

The overall architecture of the demonstrator

OPC UA in the form of a client/server architecture is used to implement the described skill-based control within the demonstrator. The individual operating resources provide their skills with the associated state machines as information models in the form of an OPC UA server, while the control is carried out by various clients. The architecture of the demonstrator cell was logically divided into individual hierarchy levels for this purpose. In order to achieve a high degree of coverage between the logical-functional architecture and the control architecture, fully integrated operating resources are used in the demonstrator. These are automation components or (sub)stations that are equipped with local intelligence in the form of a small controller, which is used to ultimately execute the skills. This project-neutral, internal implementation of the component functions is carried out by the component manufacturer, as they have the greatest expertise in operating the component and therefore also guarantee the optimum operating settings. Specifically, the demonstrator uses industrial controllers such as the Effectuator from Elrest as well as commercially available small controllers such as the Raspberry Pi 3. If it is not possible to integrate the controller into the component or station, an external industrial controller (CPX from Festo or PFC200 from Wago) is used to control the actuators. The hierarchically structured functional architecture remains unaffected by this.

To demonstrate the independence of the runtime environment, either Codesys or Forte (4DIAC) is used on the controllers in the demonstrator, which communicate with the operating equipment on a skills-based basis via OPC UA. The use of integrated components enables the encapsulation of the operating equipment and thus the change from centralized to decentralized control architectures or the interchangeability and changeability of individual system parts.

However, due to the current lack of deterministic real-time capability of OPC UA, this architecture has limitations, such as the purely sequential processing of individual skills. Therefore, the implementation of synchronized and time-critical process steps within the demonstrator is currently being dispensed with or - as already described in the example of Pick & Place - these must be offered as a composite skill by a higher-level controller, which then also takes over the necessary real-time-critical control of the individual underlying components. However, the use of the Ethernet standard TSN, which is currently being tested, in conjunction with the publish/subscribe mechanism of OPC UA would in all likelihood solve this shortcoming.

In short, even if proprietary fieldbus and industrial Ethernet interfaces continue to dominate production networking in the future, the demonstrator shows that the use of OPC UA is not only interesting for pure data acquisition, but can also be used to implement a control architecture. Especially with the introduction of deterministic real-time through OPC UA Pub/Sub and TSN, the course has finally been set for comprehensive OPC UA-based production control.

Authors:
Patrick Zimmermann is a research associate at Fraunhofer IGCV in Augsburg;
Benjamin Brandenbourger is a freelance consultant in the field of IIoT.