Model-Driven Techniques for Integrating Data Engineering into Systems Engineering

Abstract

Data Engineering (DE) is gaining importance recently due to its ability to guide and drive decisions, e.g., by increasing efficiency and effectiveness in Systems Engineering (SE). At the same time, Model-Driven Engineering (MDE) methods are applied to SE, referred to as Model-Based Systems Engineering (MBSE), to manage the increasing complexity of modern systems in product development using models as primary artifacts. However, MBSE methods lack sufficient support for collecting and processing data to gain insights and knowledge, respectively. Although DE is promising to solve data collection and analysis challenges of MBSE, current state of practice and state of the art do not reflect the opportunities of DE in SE. Reasons for the gap are: first, a lack of knowledge in practice about the opportunities of DE and insufficient precondition elaboration to integrate with existing processes and technological environment. Second, unclear benefits of integrating DE into SE and communication issues among various disciplines lead to divergent expectations and missing acceptance in practice. Third, high efforts to implement DE applications lead to long duration and a bottleneck of available data scientists. In response to these issues, this thesis proposes a method with four steps for integrating DE into SE by leveraging MDE techniques. The first step focuses on participative workshops involving relevant stakeholders to gather knowledge, promote interdisciplinary communication, and validate findings using graphical modeling methods. The second step supports the integration of a desired DE implementation into existing processes. The third step supports the formalization of DE tasks using MDE techniques, aiming to drive the implementation of DE applications while fostering communication. The fourth step decomposes formalized DE tasks to enable code generation, reuse existing knowledge, and reduce implementation time.The proposed method extends existing state of the art methods, consolidated through a Systematic Literature Review. The underlying methods are validated in two use cases, highlighting the applicability in practice and the necessity to involve various disciplines as knowledge sources. Further, a user study indicate applicability and usability for non-programming engineers, e.g., mechanical engineers, and data scientists in practical samples. This method contributes to various research disciplines by introducing and evaluating a model-driven approach to facilitate the implementation and integration of Data Engineering in Systems Engineering.