Over the last two decades, a concept called Digital Twin has evolved rapidly. Yet, there is no unified definition of the term. Based on a literature study and an industrial case study, an overarching definition of Digital twins is presented. Three characteristics were identified – representation of a physical system, bidirectional data exchange, and the connection along the entire lifecycle. Further, three sub-concepts are presented, namely: Engineering Twin, Production Twin, and Operation Twin. The presented paper thus formulates a consistent and detailed definition of Digital Twins.

A Digital Twin as a virtual representation of a physical system is becoming a key technology. While potential benefits are evident, there is no approach in literature or practice comprehensively supporting its introduction. In an industrial case study, a generic procedure model for the conception and implementation of a Digital Twin was developed. The relations between use cases, usage data, and virtual models resulted in a target concept as well as requirements for the implementation. Thereby, companies can access the potentials of a Digital Twin taking into account their specific situation.

The growing digitization affects all areas of engineering. Together with fast-paced trends, it drives complexity and uncertainty in many domains. Yet, its potentials are manifold and, in most cases, outweigh the disadvantages. Beneath terms such as "big data", "digital twin", the term "data-driven engineering" has evolved over the last years. However, neither in literature nor in industry, there is a unified definition or understanding of the term. The presented research is based on a literature review as well as an industrial case study. Several databases were screened systematically for the literature review and forward and backward searches were used additionally. The case study was conducted in a collaboration with a company in the climate system sector. First, a literature-based distinction between the terms model-based, modeldriven, data-based, and data-driven as well as definitions of data-driven engineering were investigated. Representatives of the company then evaluated these findings in a workshop and together with the industry partner a consistent definition was developed. The authors define data-driven engineering as a framework for product development in which the goal-oriented collection and use of sufficiently connected product lifecycle data guides and drives decisions and applications in the product development process. Further, promising use cases for the industry partner regarding data-driven engineering were formulated. The use cases were initially evaluated and prioritized regarding their cost-benefit ratio. Symbioses with other strategies of the company such as Digital Twins, model-based engineering, and solution space engineering are outlined. For academia, the presented findings provide a consistent definition that can be used as a promising direction for future research. Especially a procedure model for the systematic conception and implementation of data-driven engineering would be beneficial. For industry, this paper provides insights on potentials of data-driven engineering, a differentiation from related concepts, and very concrete use-cases serving as a starting point for a company-specific implementation.

No abstract

In this contribution, an 'approach for model-based fault detection and isolation (FDI) of sensor and process faults for nonlinear processes is presented. The process is decomposed into several subprocesses and for each a nonlinear model is identified. This model bank consisting of fuzzy models (Takagi-Sugeno type) is used to generate several different estimates for process outputs and states. Comparing these estimates with actual measured ones leads to residuals which indicate the state of the system and provide information about the source of possible faults. The two ways to implement a model, as a parallel or as a series-parallel model lead to different FDI results. Hence, this different sensitivity is also investigated in this contribution. The practical applicability is illustrated on an industrial scale thermal plant. Here, seven different process faults and eight different sensor faults can be detected.