Last Updated: 10-2020
Purpose – With the introduction of the Quality Assurance Act (QAA) and the increased interest around enabling corporate knowledge capturing to enhance the efficiency and effectiveness of construction projects, the application of data driven-decision making (DDDM) on construction project data has been initiated. To include DDDM in the early design phases of newly to develop construction projects, historical project evaluation data (HPED) has been considered. The steps from raw construction data towards HPED are complex and considered labor-intensive. Therefore, this research provides a framework that represents the start-to-finish process around the: planning-, capturing-, storing-, analyzing-, Act & reusing of construction knowledge data. This framework which is called the: BIM-Based Predictive Knowledge Management System (B-BPKMS), has the goal to provide insight around the influence of variables on the execution phase of processes. By mapping the influence of these variables, it becomes possible to optimize the execution phase based on the earlier obtained experience of the construction organization.
Methodology – To enhance the application of the B-BPKMS, a diverse amount of options have been elaborated to provide a general approach towards creating HPED. These options can be customized for each individual organization. Software tools such as Revit, Solibri, Synchro 4D, Asta Powerproject, SimpleBIM have been used to manipulate the IFC-data models and prepare them for analyzation. Within this analyzation, special attention has been given towards the multiple linear regression techniques and the supervised machine learning algorithm: Decision tree algorithm. The DTA is applied to classify HPED for process improvement purposes.
Results – The application of the B-BPKMS has been applied within towards a case-study to create the first customized HPED which can fill the object-oriented relational database for that organization. The final dataset is analyzed and shows the possibilities of applying supervised machine learning algorithms on HPED. Due to analyzing the outcomes of solely the dataset of the case-study, it is possible to determine the influence of different variables on the execution process of placing pre-fab concrete floor slabs (e.g. influence of wind, rain, height). Additionally, newly to develop projects can consult these figures in their development stage and create an as optimal as possible scheduling based on previously executed projects.
Scientific relevance – the novelty of the research is to illustrate the possibility to capture tacit knowledge codify this to explicit knowledge and implement this back in the organization to learn from previously executed projects while ruling out human biased perceptions. This all to minimize the gap between methodologies of the data science domain and the construction industry.