Keywords: fault detection and diagnostics, building information model, parameter estimation.

The building information model (BIM) is a key component for storing different building data, including the geometry, geographic information, quantities and device properties. In this work, we show how this data can be used to intelligently design fault detection and diagnosis code to cover the entire building life-cycle, from design to operation. The main contribution of this work is to show how to use the BIM to structure the condition monitoring and diagnostics models, and also how BIM data can be used for updating model parameters following building commissioning.

We show how to use a BIM to structure fault detection and diagnostics (FDD) models for building applications, and also how BIM data can be used for learning model parameters for updating the FDD parameters following building commissioning. We propose an approach for generating FDD rules using a generic meta-model together with the data defined in a BIM or building management system design database. Our meta-model is a detailed model that identifies a key set of properties of a system, e.g., connectivity and functionality of the devices that comprise the system. We then show how we can tune the parameters of the FDD rules using data from a building simulation model, or from actual building data collected in a data warehouse. We illustrate our approach using a lighting systems model within an intelligent building application. More details of our meta-model approach can be found in [1].

We use a two-phase approach for parameter estimation: (1) initialization, and (2) tuning. During the initialization phase, we compute the initial values of the parameters using data simulated by a model. We developed a hybrid systems model for the lighting system, which we used to simulate data for normal and faulty conditions. Using this simulation model speeds up the process of initialization, and it allows us to simulate faulty data (just by setting fault conditions in the model to true); by contrast, to obtain faulty data in a real building would require considerable work, and potentially would entail destructive testing. During the tuning phase, we take the initial rules and fine-tune the thresholds using data collected from the real building, as stored in the data warehouse.

1

M. Behrens, G. Provan, M. Boubekeur, A. Mady, "Model-Driven Diagnostics Generation for Industrial Automation", Proc. 7th IEEE International Conference on Industrial Informatics, 24-26 June, 2009.

purchase the full-text of this paper (price £20)

go to the previous paper
go to the next paper
return to the table of contents
return to the book description
purchase this book (price £140 +P&P)