Christopher Haubeck

Operationalized definitions of non-functional requirements on automated production facilities to measure evolution effects with an automation system

Production facilities are systems under constant evolution due to frequent changes of requirements and unforeseen incidents. Complex interrelations within the facilitys physics and the control software may cause unintentional changes in its behavior after performing modifications. In order to detect such effects, we present, based on a detailed literature review, definitions of nonfunctional requirements on production facilities. The definitions are operationalized, when necessary, to allow an automatic examination of the fulfillment of requirements. This can be done by using dynamic information available in typical automation systems of production facilities. We show which information can be used for an automatic examination and which evolution scenarios affect the defined requirements.

Programming BDI Agents with Pure Java

BDI represents a well-known agent architecture that has been successfully adopted for expressing agent behavior in terms of beliefs, desires and intentions. A core advantage of the architecture consists in its underlying philosophical model that relies on intuitive folk-psychological notions to describe rational human behavior. A key challenge consists in making the ideas of the BDI model easily accessible for software engineers. For this purpose many different BDI programming languages have been devised that differ considerably in their interpretation of the attitudes and the used programming paradigm. In many cases, novel agent languages such as AgentSpeak(L) have been developed which expose a new syntax and semantics to the user. On the one hand this is positive because it allows for introducing a compact and concise notation, but on the other hand the language is very different from well-known and adopted mainstream languages. To remedy this problem it will be shown that the BDI model can also be realized in a completely object oriented programming language by exploiting its metadata capabilities. We will show how the BDI attitudes can be mapped to slightly enhanced object oriented counterparts and how common BDI use cases can be realized using the novel approach. A key advantage of the approach is that BDI programming more closely resembles object orientation and the learning effort is reduced, because existing concepts and tool chains can be further employed. The usefulness of the approach will be illustrated with an example application from the area of production automation.

Evolution of production facilities and its impact on non-functional requirements

Due to high acquisition costs, production facilities are to operate for many years or even decades to be profitable. During operation, application and customer requirements change rather frequently. Therefore, a process operator must constantly evolve the control software and the underlying system. This task is restricted by specific constraints in the domain of production systems (e.g. short reaction times, high dependency on physics, etc.) hindering the proper use of formal engineering processes, which results in a lack of explicit documentation. Under such circumstances, it is evident that long-living automation software systems require special strategies to deal with incomplete information. Moreover, due to the complexity of production plants, the interconnection between evolution scenarios and system requirements might be complex. Then, a link between evolution and fulfillment of requirements is to be defined. In an effort to give a structured overview of the resulting difficulties due to improperly performed evolution steps in production facilities, this contribution presents a categorization of evolution scenarios from a practical point of view. In addition, interrelations between physical process measurements and high-level requirements are shown. This paper aims at describing the occurring difficulties within evolving production systems from a practical point of view and establishing a first step towards exploiting process measurements for requirement-aware production systems.

Evolution Management of Production Facilities by Semi-Automated Requirement Verification

Abstract The concept described in this contribution utilizes available process data in production systems to enable to verify the fulfillment of (non-functional) requirements during operation. Therefore, the contribution presents a systematic approach to automatically derive property values out of signal traces by using adaptable runtime models. Those can be checked for the violation of limit values in order to verify the fulfillment or violation of requirements on these properties during an evolution. In an example application to a Pick and Place unit, the concept is used together with an anomaly detection method to support the operator during the evolution process by constantly providing information regarding requirement fulfillment.

Selected challenges of software evolution for automated production systems

Automated machines and plants are operated for some decades and undergo an everlasting evolution during this time. In this paper, we present three related open evolution challenges focusing on software evolution in the domain of automated production systems, i.e. evolution and co-evolution of (interdisciplinary) engineering models and code, quality assurance as well as variant and version management during evolution.

An active service-component architecture to enable self-awareness of evolving production systems

Production systems are typically long-living, interdisciplinary systems which undergo continuous evolution. However, especially in the industry of the production automation, any formalized documentation of evolutionary changes is often neither created nor adapted to the application. Accordingly, no knowledge artefacts exist that can be automatically processed in order to support the evolution process. Therefore, this paper proposes a software system which is capable to capture knowledge about the underlying production process. Based on so called “active service components” the corresponding software architecture enables the production system to acquire and keep knowledge about itself and to implement further functionalities based on this “self-awareness” in a uniform way. This is done by external behavior observation (without influencing any control code), which makes the architecture suitable for already existing plants in a non-invasive manner.

Supporting commissioning of production plants by model-based testing and model learning

During the commissioning phase of production systems the identification and correction of malfunctions is a tedious task mainly done manually by commissioning engineers. This task is of high importance because missed malfunctions may result in hazardous behavior during operation phase. At this point, regardless of the engineers expertise a systematic support can drastically decrease the risk of missed malfunctions. A promising systematic approach is to use engineering artifacts of the system design phase as an information source to identify unexpected behavior regarding the specification. This paper proposes such a systematic approach based on model-based testing resulting in automatic test case generation and execution which allows to support engineers with learned models representing the expected transient system behavior. Subsequently, the obtained models are used for detection of unexpected behavior during commissioning. The unexpected behavior is presented to a commissioning engineer who decides if the behavior (1) is correct and will be added to the models or (2) represents an identified system malfunction. The approach is evaluated on a demonstration plant.

Learning material flow models for manufacturing plants from data traces

Models describing the material flow of discrete manufacturing systems are important documentation artefacts and the basis for a comprehensive understanding of the underlying processes. The analysis of such models allows deriving important key performance indicators enabling the assessment of the current system implementation. However, manual modeling as well as up-to-date model maintenance is an error-prone and costly task. In an effort to allow for the automatic derivation of material flow models, this paper introduces the concept of Material Flow Petri Nets (MFPNs) and presents a learning algorithm for their automatic generation based on recorded PLC I/O data. The proposed algorithm has been evaluated on a case study of a laboratory plant with successful results.

Interaction of model-driven engineering and signal-based online monitoring of production systems: Towards Requirement-aware evolution

Due to market requests many production systems undergo an everlasting evolution process that increasingly shifts traditional development activities for production systems to later phases of their lifecycle. As one of these activities, this contribution aims on the need for a semi-automated requirement verification mechanism during evolution. In this context the contribution proposes an answer to the question how a posteriori as well as a priori verification can be combined and which synergies arise accordingly. To do so, two appropriated approaches, one using an interdisciplinary modeling framework for model-driven engineering, and one using a PLC-signal based monitoring technique are presented that enhance each other by comparing estimated and actual system characteristics as verifiable requirement description. The resulting combined approach and its synergies are illustrated in two scenarios of an evolution case study.

A Taxonomy of Anomalies in Distributed Cloud Systems: The CRI-Model

Anomaly Detection (AD) in distributed cloud systems is the process of identifying unexpected (i.e. anomalous) behaviour. Many approaches from machine learning to statistical methods exist to detect anomalous data instances. However, no generic solutions exist for identifying appropriate metrics for monitoring and choosing adequate detection approaches. In this paper, we present the CRI-Model (Change, Rupture, Impact), which is a taxonomy based on a study of anomaly types in the literatureand an analysis of system outages in major cloud and web-portal companies. The taxonomy can be used as an anlaysis-tool on identified anomalies to discover gaps in the AD state of a system or determine components most often affected by a particular anomaly type. While the dimensions of the taxonomy are fixed, the categories can be adapted to different domains. We show the applicability of the taxonomy to distributed cloud systems using a large dataset of anomaly reports from a software company. The adaptability is further shown for the production automation domain, as a first attempt to generalize the taxonomy to other distributed systems.