Project Objective and Description

VAPi-KI is intended to close gaps in the digital process chain in the area of requirements elicitation and verification by comparing them with measured values. The resulting hybrid models for better prediction of requirements allow a reduction in product development costs through the data-driven synthesis of physical models. The methods should support an earlier correction of requirements and thus also lead to a reduction in the time and costs required for product development.

The objective of the project is a technology based on artificial intelligence (AI) with which data from requirements management for a new development can be linked and processed with measurement data from an existing solution and thus verified. This digitally enriches the processes of determining and comparing requirements, allowing optimization potential to be discovered.

In product development, requirements describe the desired properties of a product as measurable variables. However, the actual properties and functions of the product can only be measured in the usage phase. This phase is dominated by phenomenological experiences. However, methods for the digitally supported traceability of findings from implementation and use are often still lacking. The goal of the project is to close this gap and increase the probability of success of innovative development projects. Artificial intelligence methods will be used to shift the requirements determination and validation of innovative technical systems from an approach based on empirical knowledge to a data-driven approach using the example of forming machines. The scientific focus of the project is the recognition and use of the relationship between shape and property. This is to be represented in a graph database and enriched with information using AI methods. The determined shape-property correlation can be used to compare the measured characteristic values with existing requirements in order to revise them and derive new/missing requirements. The basis for this is formed by sensor data, which makes the product behavior measurable, and information from the requirements, which enables a final evaluation. In particular, the linkage of word vectors from Requirements Engineering (RE) with numerical datasets from measurement technology presents a scientific challenge, as these represent fundamentally different data types.

CONWEAVER supports the project by designing and implementing a suitable graph based on the scalable and high-performance Linksphere platform. The graph technology closes the gap between the world of words and the world of measurement data by semantically linking the relevant aspects.