Digital technologies have increasingly shaped the global construction industry with the fourth industrial revolution (Industry 4.0) emerging as a widely discussed concept. In this context, Construction 4.0 represents the sector’s response to Industry 4.0 principles, particularly through innovative technologies such as Building Information Modeling (BIM). Accurately defining and verifying requirements is essential for project quality and safety, but complex projects often involve thousands of requirements that must be manually checked. As projects grow more complex, there is a clear need for automated verification. This research explores the process of semi-automated requirement verification. The main research question is: “How can a (semi-)automated process be developed to validate and verify requirements within a 3D model?”. This is supported by four sub-questions exploring current limitations, existing tools and methods, and key development stages of automated requirement verification.

A literature review highlights that project requirements are typically derived from the client’s vision and form the basis of contractual obligations. Successfully extracting these requirements is important for defining the project scope and achieving client satisfaction. The literature study also explores various international and national BIM standards, which are adopted to different extents across countries worldwide. Several verification tools were evaluated such as Solibri, Navisworks, BIM Assure, SMARTreview and Verifi3D. From the comparison, Solibri and Navisworks emerged as the tools that provide the most comprehensive functionalities. In terms of requirement structuring methods, the RASE language was identified as the most promising approach due to its clarity and expressiveness.

The research used the Oosterweel project as a case study that is being developed in Antwerp (Belgium) with the aim of completing the city’s highway ring road. The verification process of requirements is broken down in five key steps. The first step focused on the classification of
requirements into geometric and non-geometric requirements using a machine learning model in Python. The second step involved manual rule interpretation using the RASE mark-up language. The third step focused on preparing the model data by filtering the model and performing semantic enrichment. Step four looked at the actual execution of the verifications where different types of requirements were distinguished in classes based on their computational complexity. The last step focused on reporting of the verification results in a Common Data Environment.

Implementation challenges during this process included ambiguous requirement texts, inconsistencies in model data and a fragmented workflow. The insights from the literature review, findings and interviews were combined into guidelines which are divided into three levels: strategic, operational and BIM level. The strategic level focuses on guidelines from the organizational point of view. The operational level looks at how strategic intentions can be translated into practical actions. The BIM level addresses guidelines that relate to the practical
use of BIM. The guidelines are validated through interviews with professionals.

The results of this research show that a semi-automated approach to requirement verification within the BIM environment is feasible and significantly more efficient than current manual methods. Research limitations include the limited scope of one case study and the internal
validation of the guidelines. Based on the research, multiple recommendations were given for future research and practice. Practical recommendations include using a structured approach and the use of open standards. Recommendations for future research include the further
investigation of AI beyond the current machine learning model for requirement verification and the investigation of the impact of the automated verification process.