Application and Adaptation of a Process Model for Data-Driven Validation of the System of Objectives

Abstract

Development practice shows that it is currently almost impossible for developers themselves to actively use data of customer's use of machines to derive decisions in the context of the development of mechatronic systems. In the light of the steadily growing data volumes and the associated costs, it makes sense for companies to make this data accessible to as many developers as possible. Therefore, an initial process model for the data-driven validation of the system of objectives is adapted to further improve the methodical support of developers to actively use machine usage data in the development process. To get a better understanding on how the initial process model for the data-driven validation needs to be adapted to meet the challenges developers are facing in using data, a validation study is executed. The study is carried out by 14 Data Science students under the supervision of company-internal domain experts. The main challenges as for example large data volumes, data availability, understanding of the technical system and unclear validation objectives can be assigned to the three causes: knowledge about data, organizations internal processes and combination of domain knowledge. To cope with the identified challenges, it is necessary to also incorporate the distinct selection of data and tools complemented by an estimation of the opportunities and risks of choosing a certain combination of data and tools. The study shows that adaptations to the process model increased the applicability, understandability and the perception of feeling supported in conducting analyses for validation purposes. It could be observed that a lack of knowledge about the technical system under investigation results in incomplete or incorrect information being shared. Therefore, a centralization of analysis activities in the development of mechatronic systems is not expedient.