Data and Visual Analytics for Cyber-physical Systems : Current Situation and Strategies for Action

Abstract: Today, cyber-physical systems (CPS) exist everywhere in different sizes, with different functionalities and capabilities. CPS often support critical missions that have significant economic and societal importance. They require software systems, communications technologies, sensors/actuators, embedded technologies, and physical systems to work together seamlessly, and they are seen as a driving force behind digital transformation. This dissertation describes the research work carried out to investigate applicability of data and visual analytics for CPS to overcome three main challenges: interoperability, complexity, and sustainability.To this end, several case studies are used to effectively implement and test different data and visual analytics solutions to aid stakeholders when they make decisions on interoperability, complexity, and sustainability for CPS. These studies raised questions about issues found to be of importance for the success of data and visual analytics approaches, including accessibility, availability, quality, volume, and variety of data—issues. Moreover, additional studies are used to show the benefits of blending different approaches, such as systems thinking and design thinking, and the current data analytics readiness of the Swedish industry is assessed through a questionnaire completed by more than a hundred respondents. The data and visual analytics are positioned between digitalization and machine intelligence as a research focus. Data and visual analytics is the next step after digitalizing the information by adding analytical capabilities to the data. It is also an important phase before developing machine intelligence applications. Earlier studies clearly show that only a fraction of companies have machine intelligence applications across the enterprise. One important reason behind this is the lack of strong digital capabilities that big data and advanced data analytics technologies could bring. The findings of the work carried out as part of this thesis show the importance of this middle phase—data and visual analytics—for the success of not only the CPS but also these two concepts—digitalization and machine intelligence.This thesis concludes by highlighting that currentdata and visual analytics approaches in CPS are closely dependent onthe availability, accessibility, quality, volume, and variety of the data. Notably, the huge amount of industrial data that exists in CPS manufacturers data repositories does not always mean that this data is useful, especially for analytical purposes. To this end, firstly, the CPS industry should concentrate its efforts to collect useful data that will benefit the industry by providing analytical insight intothe environments where CPS is produced and operated. Secondly, the industry should make necessary organizational changes such as considering to employ data scientists, analysts, and business intelligence developers and make data accessible tothese people for further usage. Thirdly, the data management procedures and data analytics roadmaps of companies should be created and shared with other employees, and necessary mechanisms needto be considered to improve and guarantee the quality of the data. Lastly, the variety of data needs to be addressed by the industry. Data and visual analytics provides an opportunity to extract patterns; to evaluate the interoperability, complexity, and sustainability; to create an overview of the current challenge by providing different viewpoints adapted to different stakeholders, focusing on key concerns for the respective stakeholder; to optimize performance, automation, and cooperation of distributed CPS, development environment, and teams; and overall, to improve any of the challenges that are mentioned above by, basically, providing a better understanding.To this end, I suggest that the industry discuss the next step after digitalization and address the challenges related to the availability, accessibility, quality, volume, and variety of data by considering user-centric approaches and organizational needs of the future development and manufacturing environments. The CPS industry should plan and act on these challenges as part of its data analytics strategies to expedite the machine intelligence applications of the future.