Digitalization of HVAC schema drawings
Background
In the building domain 2D-schema of e.g., the heating ventilation and Air conditioning (HVAC) or the building automation (BA), are still one of the most important documents to transport information between different stakeholders. The level of information is increasing depending on the building lifecycle constantly, and the information content focuses on different stakeholders (e.g., planning engineers, facility managers, etc.). Building Information Modelling digitises the building domain and its processes in general. Nevertheless, the 2D-schema of a HVAC-plant is often decoupled from a digital BIM model, because it is an abstract and simplified representation of the building plant and corresponding exchange circuits. One of the biggest problems is that changes are not automatically synchronized nor checked between the model and the 2D-schema. As the 2D-Schema is still one of the most important documents for BA-engineers, it is important to assure correspondence between schemas. Furthermore, for most of the existing aging building stock, schemas are digitally stored in formats that offer little machine-readable semantic information about different elements ofschema, or, in worst case, they are stored in non-digital (paper) form. Digitalization of old schemas and automatically extracting semantic information about the elements and systems is relevant and non-trivial problem.
The following master thesis should help solve these issues and improve the planning processes.
Thesis should leverage existing know-how, models and accomplishments from the field of deep
learning scene segmentation and object recognition.
Thesis objectives
The aim of the thesis is to extract information out of the schema and help digitise them, in order to link them into digital model. Following objectives should be achieved:
- Extract all relevant information (e.g. components, connections and dependencies and properties) out of an existing building scheme (HVAC and BA), regardless of whether it comes from .dwg, .pdf or printed paper
- Identify how these components are related in the schema.
- Store the information in a suitable framework such as ontologies. Check which existing approaches exist (recognize and save information) and develop an application / algorithm for a stable proof-of-concept.
Thesis learnings
The student will gain knowledge in following areas
- HVAC planning processes, understanding of complex 2D-schemas
- Computer vision, segmentation, object recognition, text recognition
- Data management: master and extend on related ongoing linked data efforts, suitable data
- Storage approaches
Proposed time schedule, milestones and intermediate reports
The student will be employed and compensated financially by AIT Austrian institute of Technology. The work will be done in Vienna. The thesis is expected to start under January 2022 (wk. 3) and be completed in June 2022 (wk. 23).
Contact persons:
Supervisor at KTH, Department of Energy Technology
Supervisors at AIT, Center for Energy
Stefan Hauer , Research Engineer
Milos Sipetic , Research Engineer