The benefits of the results for building management and maintenance management of bridges will be evaluated on the basis of several use cases, which will be implemented in at least two demonstrators. Further demonstrators will be decided in the definition phase during the requirements analysis.
Building management demonstrator: suitable inventory models will be created for the maintenance of technical building equipment (TBE) and for the optimization of space utilization and energy consumption. Concrete AI methods for the creation of building components, rooms and TBE will be developed and combined. For automatic creation, conventional 2D plans (floor plans, elevations or schematic plans) are evaluated using AI methods. The evaluation of actual images provides important information on the current state of the building. Concrete plant products with their properties are identified, which are then transferred to the inventory model on the basis of product data. In addition, a wide range of information on the structural and technical objects can be found in documents. These texts are automatically evaluated using retrieval methods to update individual objects. The results of the AI methods are integrated in a rule-based manner to consistently merge building and TBE elements into one inventory model. The demonstrator will also show how an existing digital as-built model can be updated based on images and documents. An actual as-built model can efficiently ensure operator accountability. The AI methods are integrated into existing BIM systems based on open interfaces. Due to the open approach of GAIA-X, existing systems of the companies ETA and HSG can be used, trained AI procedures for further building types (e.g. hospitals) and other technical facilities can be integrated and data provision can be made secure.
Demonstrator Maintenance Planning of Bridges: Appropriate as-built models must be created for efficient maintenance planning of bridges. The accuracy and level of detail of the as-built models must be defined for each specific project, e.g. the requirements for a new replacement construction are lower than for the repair of individual structural elements. For the demonstrator, concrete AI procedures are developed for the creation of building elements, structural connections and the building ground. Point clouds and 2D plans provide information on dimensions and non-visible construction elements. The analysis of images and texts (e.g. construction books, expert reports) contain information on materials, damage to the structure and the subsoil. The AI methods each provide individual results for bridge elements or bridge subsystems, which are combined with the help of expert knowledge. A secure cloud infrastructure is also necessary, as bridge structures in Germany are part of the safety-relevant infrastructure and distributed data sources are to be used. In addition to the use of images and text, information from databases of the federal and state road administrations will be integrated to address the concerns of public agencies. The AI procedures are used directly in the software systems of the partners ALP and AUS for bridge modeling (e.g. Allplan Bridge, A+S KorFin) on the basis of open interfaces. The evaluation of a generated as-built model is carried out for the demolition planning of a bridge and the repair of a noise barrier on a bridge. The demonstrator thus shows how domain-specific knowledge clusters and training models of the AI methods used are applied. Transferability to other infrastructure structures (e.g. tunnels, hydraulic structures) is given.
The results of BIMKIT will significantly improve the quality of as-built models and accelerate their creation considerably. The application focus is on the creation of as-built models for buildings for real estate management as well as for bridges for the planning of new replacement structures. The results will deliver many advantages not only for planners and operators, but also for users and facility management service providers. Specialist planners are provided with new options so that the current repetitive, time-consuming and manual creation of as-built models can be automated and carried out with higher productivity.