Marek Obitko

Big Data Challenges in Industrial Automation

Within the industrial domain including manufacturing a lot of various data is produced. For exploiting the data for lower level control as well as for the upper levels such as MES systems or virtual enterprises, the traditional business intelligence methods are becoming insufficient. At the same time, especially within internet companies, the Big Data paradigm is getting higher popularity due to the possibility of handling variety of large volume of quickly generated data, including their analysis and immediate actions. We discuss Big Data challenges in industrial automation domain, including describing and reviewing relevant applications and features. We pay special attention to the use of semantics and multi-agent systems. We also describe possible next steps for Big Data adoption within industrial automation and manufacturing.

Semantics in Industrial Distributed Systems

Abstract Industrial distributed systems aim at robust and flexible control of industrial processes, for which the traditional centralized approaches are not sufficient. The general problem of distributed systems is that they are still tightly coupled from the point of view of system integration and are still far from openness that would enable cooperation at a larger scale without any human intervention. To achieve better operation and integration in an open reconfigurable environment, explicit semantics is needed to capture the meaning during communication. Relevant research area is the field of ontologies and semantic web. We show how semantics can be employed in industrial systems, in particular in distributed agent-based systems, and especially using semantic web research. We review current state of the art and, based on our own experience, we discuss potentials and challenges as well as differences and similarities of applications of semantics and ontologies in industrial systems when compared to WWW oriented research.

Understanding Data Heterogeneity in the Context of Cyber-Physical Systems Integration

The current gradual adoption of the Industry 4.0 is the research trend that includes more intensive utilization of cyber-physical systems (CPSs). The computerization of manufacturing will bring many advantages but it is needed to face the heterogeneity problem during an integration of various CPSs for enabling this progress. In this paper, we describe various types of heterogeneity with emphasis to a semantic heterogeneity. The CPSs integration problem is classified into two different challenges. Next, we introduce the approach and the implementation of the semantic heterogeneity reduction with the focus on using Semantic Web technologies for a data integration. Then, the Big Data approach is described for facilitating the implementation. Finally, the possible solution is demonstrated on our proposed semantic Big Data historian.

Big Data Semantics in Industry 4.0

The Industry 4.0 is a vision that includes connecting more intensively physical systems with their virtual counterparts in computers. This computerization of manufacturing will bring many advantages, including allowing data gathering, integration and analysis in the scale not seen earlier. In this paper we describe our Semantic Big Data Historian that is intended to handle large volumes of heterogeneous data gathered from distributed data sources. We describe the approach and implementation with a special focus on using Semantic Web technologies for integrating the data.

Big data analysis for sensor time-series in automation

The trend of large scale data production is observed not only within web companies, but is entering also other domains including automation domain. Smart sensors and smart devices contribute to growing amounts of data that need to be processed. An example of processing is prediction for better control, clustering for more effective maintenance, or improving the overall production in general. The so called Big Data paradigm shows new ways of handling bigger amounts of various data, including providing technologies that are able to handle them in an effective way. This paper examines the utilization of Big Data technologies for industry automation domain. The approach is illustrated on time series data measured from a passive house with the goal of detecting specific events. We show how the Big Data technologies allow data analysis that would be hard with traditional approaches.

Diagnostics of Distributed Intelligent Control Systems: Reasoning Using Ontologies and Hidden Markov Models

Abstract The distributed intelligent control systems based on multi-agent systems paradigm bring many important features, including their flexibility and extensibility. These features are even more apparent when the agents use ontologies as a base for their knowledge management – such ontologies can be regarded as models that to some degree drive the operation of the system. However, much of the information from knowledge bases of agents can be used also for other important ability of a control system – the diagnostics. In this paper, we demonstrate and discuss two approaches to diagnostics – one based on description logic reasoning and the other one based on Hidden Markov Models. Both of these approaches are illustrated on sample scenario from a transportation system.

Architecture for Explicit Specification of Agent Behavior

Abstract In the manufacturing domain high attention is paid to the flexibility, adaptability, and robustness of the production system. These features can be more easily achieved in a distributed system, such as holonic or multi-agent system. However, these features can be hardly maintained forever by hard coded system behavior – usually system update is needed from time to time to satisfy these requirements. We discuss the necessity of explicit definition of both declarative and procedural knowledge and propose explicit procedural knowledge handling. Sharing and distribution of such knowledge is discussed and is illustrated on an implemented transportation system example. Such a solution greatly increases the possibility of system integration, openness and flexibility, all without having to restart the running distributed system.

Visualization of ontologies in multi-agent industrial systems

The development in the field of multi-agent systems used in the industrial distributed control systems has demonstrated the possibility of flexibility, fault tolerance and integration. It was shown that using ontologies for explicit description of the world has many advantages, including easier integration and extensibility. However, any more complex system, including multi-agent system, needs visualization for its easier development, debugging and monitoring. The fact that the knowledge in the system is expressed in a standardized semantic way contributes to the possibility of providing a uniform view on the system as a whole as well as on the individual agents. In this paper we describe and discuss possibilities of visualization of the state of the industrial control system based on ontologies. We demonstrate the visualization possibilities on the MAST system.

Ontology Mapping Approach for Fault Classification in Multi-Agent Systems

Abstract One of the most important abilities of control systems is diagnostics. The ability to detect faults, to explain them to an operator, and possibly also to propose and execute a recovery is an important feature of an advanced control system. We present ontology mapping approach for error classification in this paper, with the focus on multi-agent systems. Fault descriptions are kept in global error ontology which facilitates reusability of this approach as well as easier maintenance. The fault classification method utilizes HMM-based diagnostic system for automatic fault detection and offers effective and easy option of describing faults and inferring non-trivial dependencies using reasoning. The fault classification system is demonstrated on a testing example of automobile camshafts process.

Enabling Semantics within Industry 4.0

Manufacturing faces increasing requirements from customers which causes the need of exploiting emerging technologies and trends for preserving competitive advantages. The apriori announced fourth industrial revolution (also known as Industry 4.0) is represented mainly by an employment of Internet technologies into industry. The essential requirement is the proper understanding of given CPS (one of the key component of Industry 4.0) data models together with a utilization of knowledge coming from various systems across a factory as well as an external data sources. The suitable solution for data integration problem is an employment of Semantic Web Technologies and the model description in ontologies. However, one of the obstacles to the wider use of the Semantic Web technologies including the use in the industrial automation domain is mainly insufficient performance of available triplestores. Thus, on so called Semantic Big Data Historian use case we are proposing the usage of state of the art distributed data storage. We discuss the approach to data storing and describe our proposed hybrid data model which is suitable for representing time series (sensor measurements) with added semantics. Our results demonstrate a possible way to allow higher performance distributed analysis of data from industrial domain.