Nowadays, to be on edge, modern manufacturing automation systems have to be involved. Usually they consist of numerous different IT systems located at business management/operation and process control levels. It is broad class of applications domain where business IT and control systems are converged to make a large whole with the aim to improve performance as the result of the macro optimization and synergy effect. This domain is called Industrial IT. Frequently the systems are distributed geographically in multi-division organizations.
To deploy the mentioned above convergence the systems have to be integrated – must interoperate with each other. From integration we should expect improved performance as a result of synergy and macro optimization effects.
After integration the systems should make up a consistent system, i.e. each subsystem (as a component) must communicate with each others. The final information architecture is strongly dependent on organization, culture, type of technology and target process. Communication is necessary for exchanging data for production state analysis, operation actions scheduling, supervisory control and task synchronization in the process as a large whole.
Vast majority of enterprises declare that difficulties with the integration of the existing systems are the most important obstacle to expand the process control and business management support. Other major integration problems are diversification of the existing systems, their quantity and non-unified data architecture.
Integration process results in Large Scale Distributed Network Control Systems (LSDNCS). Systems belonging to this class are usually created in a multi-step integration process. To succeed, the process has to be governed by a well-defined information and communication architecture.
System integration means necessity of the information exchange. To exchange information we need an association between components. Going further, to instantiate association, i.e. to make the component interoperable, we need at the same time a common:
- information representation – a language (data type),
- underlying communication infrastructure – a transport (protocol + medium),
We must be aware that establishing an association we are actually building information architecture – system structure. It is worth stressing that selection of the architecture development has a great impact on the final robustness, maintainability, expandability, dependability, functionality, and last but not least implementation costs.
Generally we have three possibilities:
Peer to peer approach: Common integration practice is to achieve short-term ad-hoc objectives by manually creating/proprietary dedicated point-to-point links between the subsystems everywhere it is useful (see fig. 1). Using randomly this approach we can establish numerous independent links ((k+n)(n+k-1)/2 where k, n – number of business and process control components appropriately). The number rapidly grows, e.g. it is equal 1770 links for n=10 ; k=50, and finally we have to deal with rapidly growing complexity leading to the communication chaos, which is difficult to be maintained.
In this model the information and communication architectures are closely coupled. This approach is very popular, but adversely affects all of the solution features.
Totalitarian approach: One option to overcome the communication chaos problem is to use an “all in one” product dedicated to both functions: process control and business management (see presentation). Usually, it is provided as one complex, total system – let’s call it a supper-system. Most of the MES (Manufacturing Execution System) vendors offer theirs products as a panacea for all problems of the chaotic system integration.
Actually supper-system does not solve the problem, it only hides it under a not transparent cover and makes the solution very difficult to expand and vendor related forever.
In this model the system distribution is reduced, and as a consequence many associations can be instantiated on the same platform without necessity to communicate. This model reduces complexity by reducing communication needs.
If strictly observed it could be a dead end.
Process Observer approach: The Process Observer is a consistent, homogenous real-time representation of the process control layer. It is a kind of the virtual layer, which is a “big picture” of the underlying process layer composed of unit data randomly accessible by means of a unified and standardized interface (see presentation). It allows sharing data from plant floor devices by the process and business management systems, using international standards of data exchange. Process Observer is like a bridge connection between the plant-floor control and the process and business management levels.
Thereby, the structure of the links becomes systematic and the existing functionality of the upper layers is preserved. Now, they can gather the process data in a unified, standardized way (see fig.2).
Using the Process Observer archetype the number of links between components can be substantially reduced and, what is very important is a linear function of the number of nodes.
Process Observer model greatly reduces the whole complexity and decrease dependency by decoupling application associations and underlying communication routes. Additionally, it allows applying systematic design methodology and building information architecture independently of underlying communication infrastructure.
Real-Time Communication for Large Scale Distributed Control Systems (Proceedings of the International Multiconference on Computer Science and Information Technology pp. 849–859 ISSN 1896-7094)
Process and business layers robust integration (white paper)
Communication management in the Process Observer Archetype (Proceedings of the 16th conference “Polish Teletraffic Symposium 2009”)
Large Scale Distributed Process and Business Management Integration (Proceedings of the 14th International Congress of Cybernetics and Systems of WOSC)