(via www.4clicks.com – leading provider of software and services for efficient construction delivery including cost estimating, project-document-contract management, visual estimating/QTO, exclusive 400,000 enhancement of RSMeans cost data … for JOC, IPD, SABER, SATOC, MATOC, MACC, IDIQ, POCA, BOA … )
There is definately a role for the application of neural networks / AI within the AEC sector.
The technology would help to support the integrage of disparate data.
That said, it is the lack of awarencess of, and associated implementation of existent robust business processes that is primary issue within our industry.
For example, many FMers don’t appreciate the differences of CMMS, CPMS, CAFM… or the criticality of standardized reference cost data, UNIFORMAT, MASTERFORMAT, core benchmarks….FCIs, etc., or COBIE, OMNICLASS, etc. No do many appreciate the role of efficient construction delivery methods such as IPD (integrated project devliery) and JOC (job order contracting), the later a form of IPD specifically for facility renovation, repair, sustainability, and minor new construction.
Neural networks can parse disparate data (to a degree, they are about 80% to 99%+ accuarate depending upon source data and specific application, and thus still require human intervention such as manual mapping of “outliers”), FMers would do well to focus upon process, efficient construction delivery methods, etc.
(Source for following paragraph and above image –http://blog.aecsa.org )
Neural networks have been used in computer science for speech recognition, image analysis, adaptive controls, software agents, and statistical analysis for some time now. Your input layer [see image above] would consist of multiple models and code standards databases from different sources and phases of a project. This inputs are facets of an aggregate building model, interpreted by a middle layer (algorithms aka secret sauce) which results in the desired output (BIM) for specialized purposes (i.e. design, engineering, FM). They are in no way practical as the software architecture for a BIM authoring application as they are slower and require more computing resources to retrieve data. However, they would be useful for interpreting the relationships between nodes in an aggregate data structure where compute resources and storage space are not an issue… enter cloud computing.