Tuesday, October 22nd 3:00 PM – 3:50 PM
Joel Bender, Cornell University
The artificial intelligence and machine learning hype curve is in full swing, and it’s just as prevalent in building automation as it is in every other sector where there is an expectation that throwing massive amounts of data and computational horsepower can magically make life better. The reality is that it takes a tremendous amount of effort to wrangle the data into shape, filter out the noise, and align it with appropriate decision making processes.
The Semantic Web hype curve is now on the “Slope of Enlightenment” where some actionable information can be extracted from a tangled web of relationships. During its “Peak of Inflated Expectations,” there were grand ideas of unified ontologies that were carefully constructed by linguists and supervised by logicians. Now the reality is that there is no grand theory of everything, there are thousands of models of things and relationships between them, and to quote George E. P. Box, “all models are wrong, but some are useful.”
The BACnet Applications Working Group has been working for many years on defining “profiles” for building automation equipment and has struggled with creating models, a distinctly top-down effort. Project Haystack is defining a point classification vocabulary, a bottom-up effort delicately side stepping the naming convention problems that the BAS industry has struggled with for decades.
If Project Haystack is defining “words” with combinations of “tags,” then Brick is defining “phrases” that are appropriate combinations of words to use for a concept, and bridges the divide between the Haystack world and the Semantic Web.
There is a proposal now before the BACnet AP Working Group that will enable BACnet devices to present semantic web data in the form of an RDF graph using Project Haystack “tags,” Brick classes, from vocabularies developed by vendors, from vocabularies for specialized applications, from Semantic Sensor Network ontology, or any combination of them. This proposed addenda is the first step of many, enabling ASHRAE Standard 223P to construct a building systems ontology.
This presentation will be about the development of a graph model that contains the physical topology, space utilization and classification, network topology including the building automation systems, building systems topology, trend point data, and facilities maintenance data. It will showcase Klarman Hall, a Cornell University LEED Platinum building.