A flexible Architecture for Intelligent Management Systems Ralph Holland-Moritz, Ralf Vandenhouten TH Wildau [FH], Wildau, Germany ralph.holland-moritz@th-wildau.de, ralf.vandenhouten@th-wildau.de Abstract — For an intelligent management system which evaluates data by means of artificial intelligence there are different approaches which come into question. Each approach has its own advantages and disadvantages. To get the best out of all, different systems have to be combined in a proper way. This is where Complex Event Processing comes into play. I. INTELLIGENT MANAGEMENT SYSTEM An intelligent management system can be divided into four groups of components. These groups are input components, intelligent components, data converters and reporting components. A facility management system is used as an example of such a system. In this environment the input components are sensors which provide input data for the intelligent system. Sensors are defined as technical components measuring one or more physical or chemical properties or material type of goods in their environment. The measured property is called a feature or attribute of a sensor and has a specific value at a given time. Figure 1. Components in a simple architecture The reporting components, forming the output of the intelligent management system, are generated reports with status information or schematic representation like ground plans of the building with connected state indicators. The reporting component is the interface for the facility manager or security agent, giving him an overview of the events inside the building or location. Between input and output components there is an interlayer consisting of intelligent components. These intelligent components evaluate the input data coming from the events and generate output data as result of the evaluation process with the help of their knowledge. The intelligent components can be divided into two types depending on how the knowledge is stored or gained. These two types are the learning components and the knowledge-based components. A learning component is able to gain knowledge from changing findings and to draw conclusions from simulation or normal operation. The learning component is applicable for event sequences which are unknown or not describable. Under knowledge-based components we understand systems which evaluate new statements with the help of a knowledge base. The knowledge base is formed out of statements which are defined as rules, logic statements or semantic connections. With the help of this component changes in the operational application can be evaluated by means of a rule base, defined by a so called expert. For evaluating the statements of the knowledge-base and a subsequent conclusion a rule interpreter is used which interprets the facts by means of predefined rules. Knowledge-based components work well for the reasoning of events by means of a knowledge base consisting of elementary rules. It can be used where states of the system are known a priori. Different types of intelligent components need different types of input data and provide different types of output data. Some intelligent components have their own event memory, others don’t. The ones without memory need a system state copy every time they are executed. To connect the input data from the input components with the intelligent systems and the output data from the intelligent systems with the output components we need data converters for input and output conversion. II. REQUIREMENTS In an intelligent management system with different intelligent components there are requirements to ensure flexibility and other non-functional requirements. The first non-functional requirement is the requirement of parallel processing. Under parallel processing the parallel execution of the intelligent components is understood. That means that the components should work independently from each other. In an intelligent management system the focus is on time dependant events which have to be evaluated time-critically. Because of their different behavior the intelligent components can evaluate events in different speed. Parallel processing therefore means that one component is independently from another in the sense of not having to wait for the other’s termination. When dealing with different intelligent components it is previously unknown how these components work together. It cannot be determined whether component "A" and "B" evaluate before "C" and if "C" uses the output of "A" and "B" or whether they evaluate parallel or in another way. Thus the intelligent components cannot be connected in a fixed way. This means that a loose coupling is needed for connecting input and output of the intelligent components. In order to be not restricted to specific intelligent components and to be able to further develop the intelligent system or extend the system with new intelligent input or output components a requirement of the system is to be easily extensible. This means that the LINDI 2011 • 3rd IEEE International Symposium on Logistics and Industrial Informatics • August 25–27, 2011, Budapest, Hungary – 83 – 978-1-4577-1841-0/11/$26.00 ©2011 IEEE