to create context-sensitive decision support services in an eco-process engineering systems .... integration of control and maintenance of production system.
Context-sensitive decision support for improved sustainability of product lifecycle Sebastian SCHOLZEa1, Oliver KOTTEa, Dragan STOKICa, Cristina GRAMAa a ATB-Bremen, Institute for Applied Systems Technology, Bremen, Germany
Abstract. The paper presents a new approach to applying context awareness in order to create context-sensitive decision support services in an eco-process engineering systems setting. Optimizing the life-cycle of industrial products is subject to the options of continuously updating them by incorporating cutting edge technologies, replacing worn out pieces by new improved ones, and conceptually changing components of the product itself. While new products benefit from cutting edge technologies, the highest impact is achieved by upgrading existing products in operation, leading to the “long life eco-products” concept. The proposed approach uses context awareness to enable evaluating the performance of engineered products based on the whole life-cycle, so that product engineering teams can exploit this information to adapt the design, operation, and disposal strategies of products. One of the key assumptions of this approach is that the results of analysing contextual information and previously taken decisions can be used to reconfigure applicationspecific services positioned anywhere along the life-cycle of a product; by monitoring the development life-cycle and the product enriched with context from ambient intelligence, the services supporting the life-cycle can be configured in order to have faster update iterations. The presented approach will be validated against three different business cases.
Keywords. Context awareness, context modelling, collaborative decision making, product life-cycle
Introduction The optimization of industrial products’ life-cycle is a process of continuous update by incorporating cutting edge technologies, replacing worn out pieces by new improved ones, and conceptually changing components of the product itself. While new products benefit automatically from cutting edge technologies, the highest impact is achieved by upgrading existing products in operation, leading to the “long life eco-products” concept. The proposed approach uses context awareness to enable evaluating the performance of engineered products based on the whole life-cycle, so that product engineering teams can exploit this information to adapt the design, operation, and disposal strategies of products. One of the key assumptions of this approach is that the results of analysing contextual 1
Corresponding Author.
information and previously taken decisions can be used to reconfigure applicationspecific services positioned anywhere along the life-cycle of a product; by monitoring the development life-cycle and the product enriched with context from ambient intelligence, the services supporting the life-cycle can be configured in order to have faster update iterations. The approach has been applied within an industrial setting, where context awareness is used to support more flexible and timely reconfiguration of product life-cycle services [1]. Figure 1 below presents the general concept of the architecture in which the context awareness approach has been implemented:
Figure 1: Proposed Concept
The Services Generator Module is used for reconfiguring life-cycle services; it uses a context extraction service to detect changes in the context of the monitored system. A Simulation Module and a Decision Making Module are supporting services which can be used (potentially in conjunction with the Service Generator) for adjusting the parameters of life-cycle services. The paper is structured as follows: Section 1 provides a brief overview of the key addressed area: context sensitivity. Section 2 presents the proposed concept of context sensitive decision support in eco process engineering systems based on context awareness, while Section 3 explains in more detail the context sensitivity approach and context model used. Section 4 addresses the application of the proposed approach in a specific industrial setting, showing the current situation, the targeted objectives, and the way the solution is applied for achieving these objectives. Section 5 presents conclusions and future work.
1. State-of-the-art Context Awareness is a concept propagated in the domains of AmI and ubiquitous computing. It is the idea that computers can be both sensitive and reactive, based on their environment. As context integrates different knowledge sources and binds knowledge to the user to guarantee that the understanding is consistent, context modelling is extensively investigated within knowledge management research. Existing research on context can be classified in two categories: context-based, proactive delivery of knowledge, and the capture & utilization of contextual knowledge. Using context information (for context-sensitive or ubiquitous computing) is an active area of research, with various context capture methods and context languages defined. Starting with the pioneering work at XEROX PARC, other notable frameworks
are Context Toolkit (Berkeley), CAMELEON project, C-OWL and the Kimura System. The current research on knowledge context is primarily oriented towards capturing and utilization of contextual data for actionable knowledge [2]. In addition, it has been shown that knowledge context could be used to classify and organize knowledge so as to realize unified management of distributed, heterogeneous knowledge in a networked enterprise [3]. However, such initiatives have specific goals, so an intense study of collaborative work and its patterns is necessary to devise a suitable context model. The overview of collaboration support systems and context-sensitive collaboration systems provided in [4] illustrates a lack in provision of knowledge context. A couple of systems to handle context awareness were proposed by the research community, many of which are directed at the needs of wireless networks and mobile computing [5], [6], [7], [8], [9] and [10]. An approach for context awareness and context identification in collaborative work was developed in the EU-funded K-NET project [11], together with context models for collaborative work. The key hypothesis of K-NET was that the context under which knowledge is collectively generated and managed can be used to enhance this knowledge for its further use. Another EU-funded project in which context identification approaches have been applied is Self-Learning [12]; the key assumption of the project is that a context awareness approach allows adaptation and integration of control and maintenance of production system. The goal of the project is to develop self-learning production systems that are able to self-adapt in response to changes in the context in which they operate, including changes in process and equipment parameters. Although there are various types of context-aware systems, in general a contextaware system follows four steps to process context-awareness [13]: 1) Acquisition of context information, 2) Storing acquired context information into a repository, 3) Controlling the abstraction level of context information by interpreting or aggregating context data and 4) Utilizing the context information for services or applications. Context modelling can solve the problem of how to represent the context information. Based on the formal description of context information, context can be processed with contextual reasoning mechanisms. Since contextual information has some inherent features (it can be considered incomplete, temporal, and interrelated), context reasoning can exploit existing reasoning mechanisms to deduce high level, inferred context from low-level, raw contextual information. Furthermore, contextual reasoning can be used to verify and possibly solve inconsistent context knowledge due to imperfect input. For example, Luther et al. [14] show the needs for ontology support and reasoning in mobile applications; their case study is conducted with the Protégé knowledge workbench [15] for ontology modelling and OWL editing, and the RACER inference engine [16] for proof checking, ontology validation and classification. A more flexible use of ontological reasoning is presented by Forstadius et al. [17]. Their framework utilizes context-awareness for service classification. Their model enables a flexible way of describing context-based rules, which can be used for constructing prioritized service lists and for recommending available services. In addition, the model provides a way for describing context-triggered actions, e.g. notifications. Three main categories of reasoning can be distinguished: deductive reasoning, event-condition-action reasoning and statistical reasoning. The notion of context has been the subject of debates among researchers in different domain areas. Based on the requirements and characteristics of each of these domains, the term ‘context’ has been interpreted in different ways and different approaches have
been applied to capture con-textual information. In order to find out the commonalities of different views and establish a shared understanding of what context is, Bazire & Brézillon [18] analysed the scientific literature of several fields and came up with a vague notion of a set of constraints that can influence the behaviour of a system in a given task. Nevertheless, a widely accepted definition of context is the one given by Dey [19]: “Context is any information that can be used to characterize the situation of an entity. An entity is a person, place, or object that is considered relevant to the interaction between a user and an application, including the user and applications themselves.” Schmidt goes one step further to define ‘the situation’ as a relevant subset of the state of the world at a given point in time (including the respective knowledge of history and expectations for the future at that point in time) [20]. With the recent advance of context-aware computing, an increasing need arises for developing formal context modelling and reasoning techniques. The basis for contextaware applications is a well-designed context model. A context model enables applications to understand the user’s activities in relation to situational conditions. Typical context modelling techniques include key-value models, object-oriented models, and ontological methods [21]. Formal techniques would facilitate context representation, context sharing and semantic interoperability between heterogeneous systems. Context modelling for collaboration teams presents additional challenges, as these teams are highly dynamic in their constitution and objectives, reside in distributed environments and are usually knowledge-intensive [22]. Context models, both informal and formal, can be found in the literature. Informal context models are often based on proprietary representation schemes without facilities for sharing the understanding about context between different systems. Cooltown [23] uses a Web-based context model in which every object has a corresponding Web description. Existing formal context models support formality and address a certain level of context reasoning. Henricksen et al. [24] model context using both ER and UML models, allowing contexts to be managed with relational databases. Ranganathan and Campbell [25] represent context in the Gaia system as first-order predicates. However, none of these addresses formal knowledge sharing, and none show a quantitative evaluation for the feasibility of context reasoning. In addition, the problem to be addressed is the gap existing between industrial actors and customers, because of different vocabularies and ontologies utilized. Work on ontology has become popular in the ICT for the development of the Semantic Web and areas such as knowledge engineering, database design, information retrieval and integration. Ontologies support the specification of reusable libraries of Problem Solving components, to drive knowledge acquisition and to allow semantic information retrieval. They are spreading through the Web, such as Dublin Core, FOAF, DOAP, increasingly gaining consensus and agreement world-wide and fostering better integration and access to data.
2. Proposed concept The objective of the presented approach is to develop a novel eco process engineering system based on context awareness, which will constitute a comprehensive platform enabling a dynamic composition of services adaptable to the different products and operating conditions, supporting the Product Service System.
This novel service oriented framework will allow industries to evaluate the performance of engineered products considering their whole life cycle rather than only early stages such as design and manufacturing (see Figure 1). The capabilities resulting from the research will enable the capitalization on trustable global and local sustainability intelligence. Product engineering teams can exploit this intelligence to adapt design, operation and disposal strategies through managed “eco-constraints” relevant to their market contexts. This allows for faster and more flexible decision making along the product life-cycle: by analysing the context in which previous decisions have been taken, future decision making can be improved [1].
Figure 2: Reference Architecture
Following a bottom-up approach, the Platform is divided into 4 main conceptual blocks (see Figure 2): Virtual Collaborative Network (VCN): The VCN main functionality is to provide a main point of access for non-expert end-users through user interfaces, including capabilities for supporting the aggregated GUIs provided by the Service Generation Module. It provides the technical infrastructure for the distribution of users into groups; content management and sharing mechanisms; workflow engine for contents production, consumption and transformation. In addition, the VCN will provide a Knowledge Base, which will contain actual data, historical data, identified constraints and objectives from the collaborative networks, KPIs and Life-Cycle Inventory data. Service Generator (SG): The Service Generator’s main functionality is creating, updating, and deploying configurations for application-specific services. It interacts with the Simulation Module and the Decision Making Module to compute parameters for updated configurations. The SG can also store / retrieve configurations in/from the Virtual Collaborative Network. Simulation Module (SM): The simulation module of the presented platform provides a capability for running numerical analyses related to the life-cycle assessment process. It provides simulation services through an abstract service interface that allows higher-level components of the platform to use simulation as an interchangeable service, according to the principles of the SOA design paradigm. As basis GeSim, which is a discrete event simulator developed by VTT, will be used for the Simulation Module.
Decision Making Module (DMM): DMM is an interactive system intended to help decision-makers to use data and models to identify and solve problems and to make decisions. On the one hand, there will be some Traditional indicators for measuring the overall factory performance, but on the other hand, there will be new KPIs established and defined in the DMM. These KPIs will be traced to the measured constraints, and the DMM will show, in each BC, where the constraints are, and how to measure the performance of the organization, in order to take the best decision in a range of values proposed by the SM taking into account the localized constraint. The KPIs taken into consideration will mostly reflect eco constraints. Inputs from the SM will support the decision process, by simulating the potential evolution of relevant KPIs and allowing thus taking informed decisions. The Data Analysis will be based on the well-known Open Source Pentaho Suite, which includes Business Intelligence products providing data integration, OLAP services, reporting, dash-board and data mining.
3. Context Extraction and Context Modelling One of the key assumptions is that analysing contextual information and previously taken decisions can be used to update an existing configuration and to reconfigure a particular deployed application-specific service, by monitoring the development life cycle and the product enriched with context from ambient intelligence. The Context Extractor is responsible for identifying changes in the context of the environment. The current identified context will be used to extract available context knowledge. The results of the Context Extractor are then used for updating the system behaviour. The purpose of the context model is to define a fundamental data model for context extraction. The context model’s main role being to model the context, it will not try to fully describe the context, but to index context in order to help with its identification. Therefore a key research point is the definition of „holistic” and dynamic context model and ontologies to enable context awareness, taking into account the context of the engineering approach (e.g. processes, equipment and product information, users, teams, etc.). The results from EU projects K-NET [11] and Self-Learning [11] are the baseline for context extraction, interpretation and modelling in a flexible and distributed environment. Based on this context awareness approach, knowledge should be generated that is necessary to provide the envisaged decision support related to short-term operational decisions, which should be optimized regarding their effect on specific subtasks in engineering processes, as well as long-term decisions related to the overall engineering process. The proposed building blocks should be able to establish an awareness about situations and based on this provide an appropriate support, presenting to the user only the key values and characteristics that really matter. The context extraction process is presented in Figure 3 below. The monitoring data is called “standardized” because it has to be formatted according to the context extractor’s interfaces. Whenever the monitoring data is sent to the Context Identification component, the identification process is triggered. This process involves analysing the monitoring data and reasoning on it to generate as much contextual information as possible. At this step, the determination of the new (or old) current context (the context based activity the monitored system is currently in) is postponed, because further processing is required, which itself needs the newly identified knowledge context.
If the system basically has no record of any on-going context (a context that the system is involved in at the moment) of the current system, it simply creates a new one as the system’s current context. Otherwise, all on-going contexts of the system or device are ordered according to relevance metrics, which are computed based on a comparison of the newly identified context with currently on-going contexts. If the relevance value of the first on-going context is above a certain configurable threshold (e.g. > 0.9), that context is selected as the system’s or device’s current context; if no on-going contexts have relevance values above the threshold, a new context is created and set as the current one. After the current context has been determined, the similarity measure between the current context and history contexts is computed. Finally, the results (the current and similar contexts) are sent to subsequent software components.
Figure 3: Context Extraction Process
The definition of the context model was based on several considerations. First, the question of how to represent different knowledge items in a contextual manner does not have a simple answer. Second, different kinds of contexts should be represented in a common “language” when possible, but the representation must be extensible enough to support domain and application specific concepts. A semantic model provides a representation flexible enough to support common modelling of context in a structured way, as well as domain specific extension to the model, thus it was chosen for representation purposes. Especially with the advance of the semantic web, RDF/OWL based ontology modelling is a useful tool to meet the requirements. Thus, the context model underlying the identification of the context is an ontology that defines a fundamental data model for context extraction. This ontology does not attempt to exhaustively describe the context, but to index it in order to help its identification. In order to apply such an ontology-based context model to a particular situation, two ontologies are needed: a generic and a sector-specific context model. Both context ontologies will model the knowledge context (including information of activity, resources, etc.). The generic part of the ontology addresses the link between the higher level services, with their goal oriented behaviour, and the constrained, sector-specific, infrastructure services. The ontology can be easily extended to better suit the application domain, by defining domain-specific subclasses for the core concepts in the generic context model. This extension does not require a redesign of the system, since the core ontology remains unchanged and the extension can be used to further elaborate the core concepts.
4. Application The presented approach is being applied in three industrial business cases, to validate the proposed decision making solution within different application domains: BC1: Engineering maintenance services for optimizing maintenance and increasing availability of wind turbines. The decision processes involved here deal with the best maintenance route to be taken for optimal maintenance of multiple wind turbines. BC2: Power grid control systems for improved identification of improved monitoring of grid load and safety limits. Decision support methods are supporting the user in detecting trends in the grid load, ensuring thus that good load balancing decisions will be taken. BC3: Support for optimized design and manufacturing of aircraft wings. Here, the decision support system addresses the optimal layout of manufacturing facilities in the design phase. Table 1 lists the identified main specific technical aspects of the three business cases, which are addressed by the generic proposed solution. Table 1: BC-specific technical aspects addressed by the proposed solution BC
Main Area of Interest
Technical issues to be addressed
BC1
Services for optimizing maintenance and increasing availability of wind turbines
Monitor parameters of wind turbines to detect preventive maintenance needs
BC2
Power grid control systems for improved identification of maintenance needs and monitoring of grid load and safety limits
Monitoring of cable temperatures and grid load to support decision making of load balancing
BC3
Support for optimized design manufacturing of aircraft wings
Use simulations to estimate production rate, energy consumption, emissions, hazardous material waste
and
The different BCs illustrate the wide applicability of the generic proposed solution. The methodology for applying the proposed concept is presented next in the setting of the BC2. This application scenario addresses new product and customer support services for high voltage cables, to provide cable-monitoring services that help to redefine grid load and safety limits, enabling a significantly improved utilization of grid capacity. Especially the consistently increasing share of renewable energy in the European energy grid (e.g. by offshore wind parks), together with the decentralized and discontinuous production of renewable energy, will have a high impact on the grids, asking for a significant higher grid capacity (e.g. with respect to energy distribution and transport of electricity to energy storages). Due to this, electric utilities urgently need technologies and support tools facilitating an optimal usage of existing grid capacities by an improved load and security management. Usage can be improved by better maintenance and monitoring. By “better maintenance” we understand avoiding unnecessary maintenance, or better predicting future maintenance needs. Monitoring can also be improved if the raw monitored data is processed before presenting it to the human operator, making decisions easier to take. Targeted are supporting services easily adaptable to customer specific needs, facilitating analysis / optimization and maintenance of cable systems, with respect to an optimal and secure use of full cable capacity (up to 40% additional capacity). Thus, the new services will increase available grid capacities; resources required for grid expansion will also be reduced (i.e. investments, raw material, demand for land, etc.), to
achieve a European electricity grid ready for an increasing share of renewable energy. The most challenging problems in this context are the required flexibility/adaptability of services to the diversity of customer needs, as well as the provision of online context information from the cable (e.g. conductor temperature / hotspot for detection of cable faults, etc.) and ambient condition (e.g. thermal characteristics of environment), required to analyse the maximum cable load. Moreover, knowledge on cable (e.g. dimension and construction, temperature profiles) and grid specific constraints (intelligent deduction of load predictions, etc.) are basic parameters. The implementation of the proposed solution – the automatic adjustment of parameters based on changing context, e.g. changing ambient conditions – leads to minimization of errors and keeps the utilization high, as well as the overall quality. Furthermore, improved decision making will be supported taking into account previously taken decisions as well as current context, leading to better decision making proposals. For this scenario, the context monitoring serves as a basis for identifying adjustment parameters. The monitored context refers in this BC to the environment of the cables, such as the material, which surrounds them (soil, under water, in open air), the ambient temperatures, the voltages going through the cable. This context of the cables is used, together with previously taken load balancing decisions and their contexts, for ensuring the appropriate load balancing decisions will be taken.
5. Conclusions and Future Works A novel approach for the realization of eco process engineering systems based on context awareness is presented. The proposed solution addresses decision support in eco engineering systems, based on a context aware approach. The wide applicability of the proposed solution is presented, together with an example of customizing the generic solution for a specific industrial setting. The proposed platform and services are under development. Early testing of first prototypes already indicated promising results of the proposed approach. Further research will focus on advanced algorithms on extracted context to (semi-) automatically update the context model. In addition, the context model itself will be addressed by further research to allow better utilisation of the presented model for other companies as well as for other application domains.
Acknowledgement This work is partly supported by the EPES project of European Union’s 7th Framework Program, under the grant agreement no. FoF-ICT-2011.7.3-285093. This document does not represent the opinion of the European Community, and the European Community is not responsible for any use that might be made of its content.
References [1] [2]
EPES2012 Mladenić 2009
EPES Concept, EPES Project, www.epes-project.eu, 2012. Mladenić, D., Active Project Website., Retrieved January 10th, 2012, from http://www.active-
[3]
Tao 2004
[4]
Ahn 2005
[5]
Bellavista 2006
[6]
Toninelli 2008
[7]
Gu 2005
[8]
Kim 2007
[9]
Chang 2006
[10] Feng 2004
[11] K-NET2009 [12] Self-Learning2012 [13] Lee 2011 [14] Luther2005 [15] Noy 2001 [16] Haarslev 2001 [17] Forstadius 2005
[18] Bazire 2005
[19] Dey 2001 [20] Schmidt 2006
[21] Strang 2004
[22] Hyung 2005 [23] Kindberg 2000 [24] Henrickson2002 [25] Ranganathan 2003
project.eu/fileadmin/public_documents/D231_context_definition_and_represent ation.pdf, (2009). Tao, Y., Tianyuan, X., Linxuan, Z., Context-centered design knowledge management, Computer Integrated Manufacturing Systems, 10 (2004). Ahn, H. J., Lee, H. J., Cho, K., & Park, S. J., Utilizing knowledge context in virtual collaborative work, Decision Support Systems, 39(2005). Bellavista, P., Corradi, A., Montanari, R., Toninelli, A., Context-aware semantic discovery for next generation mobile systems, IEEE Communications Magazine, 44 (2006). Toninelli, A., Corradi, A., Montanari, A., Semantic-based discovery to support mobile context-aware service access, Computer Communications , 31 (2008). Gu, T., Pung, H. K., Zhang, D. Q., A service-oriented middleware for building context-aware services, Journal of Network and Computer Applications, 28(2005). Kim, S., Suh, E., Yoo, K., A study of context inference for Web based information systems, Electronic Commerce Research and Applications 6(2007). Chang, J. W., Kim, Y. W., Design and implementation of middleware and context server for context awareness, High Performance Computing and Communications, Proceedings, LNCS 4208(2006). L. Feng, P. M. G. Apers and W. Jonker, Towards context-aware data management for ambient intelligence. In: Database and Expert Systems Applications, Proceedings, 3180, pp. 422-431, 2004. K-NET Project, FP7-215584, http://www.k-net-fp7.eu/, 2009. Self-Learning Project, FP7-NMP-2008-228857, http://www.selflearning.eu/, 2012 Lee, S., Chang, J., Lee, S., Survey and Trend Analysis of Context-Aware Systems, Information - An International Interdisciplinary Journal, 14, 2011. Luther, M., et al., Situational reasoning – a practical OWL use case. Chengdu, Jiuzhaigou, China, 2005. Noy, N.F., et al., Creating Semantic Web contents with Protege-2000. IEEE Intelligent Systems, 16(2001), 60.Haarslev2001 Haarslev, V. and R. Moller, RACER system description. Siena, Italy: SpringerVerlag, 2001. Forstadius, J., Lassila, O., Seppänen, T., RDF-Based Model for Context-Aware Reasoning in Rich Service Environment, Proceedings of the Third IEEE International Conference on Pervasive Computing and Communications Workshops. IEEE Computer Society, (2005) Bazire, M.; Brézillon, P., Understanding Context Before Using It. Modeling and Using Context: 5th International and Interdisciplinary Conference CONTEXT (2005). Dey, A.K ., Understanding and Using Context. Personal and Ubiquitous Computing Journal, 1(2001), 4-7. Schmidt, A., Ontology-based User Context Management: The Challenges of Dynamics and Imperfection, Proceedings of the International Conference on Ontologies, Databases and Applications of SEmantics (ODBASE 2006), On the Move Federated Conferences (OTM), Springer, Lecture Notes in Computer Science (LNCS), 4275 (2006), 995-1011. Strang, T, Linnhoff-Popien C., A Context Modeling Survey, Workshop on Advanced Context Modelling, Reasoning and Management as part of the Conference on Ubiquitous Computing - The Sixth International Conference on Ubiquitous Computing, (2004). Hyung, Jun A., et al., Utilizing knowledge context in virtual collaborative work. Decision Support Systems, 39(2005), 563. Kindberg, T., et al., People, places, things: Web presence for the real world, Third IEEE Workshop on Mobile Computing Systems and Applications, (2000). Henricksen, K., J. Indulska, and A. Rakotonirainy, Modeling Context Information in Pervasive Computing Systems, Pervasive Computing, (2002), 79. Ranganathan, A., Campbell, R.H.. A middleware for context-aware agents in ubiquitous computing environments. Springer-Verlag, Rio de Janeiro, Brazil, 2003.
