1 Building Automation System Integration into a Cloud-based BIM-FM ModelJenn McArthur and Marin Litoiu

2 Team Cloud IoT Architecture– York UniversityMarin Litoiu Cornel Barna, Marios Fokaefs, Brian Rampasad, Yar Rouf FM Data Analytics and Visualization Jenn McArthur Elizabeth Leriche, Clarice Medina, Brandon Bortoluzzi FuseForward Connie Yee, Mark Damm

3 Concept: System Integration for Facility AnalyticsBIM can be used for relationship & trend identification to support root-cause analysis in facilities management The collection and classification of work order requests creates a bottleneck in some FM contexts Case study conducted at Ryerson University to investigate the use of BIM and data analytics to support the resolution of maintenance requests Machine learning algorithms developed to classify maintenance requests; collect information to support root cause analysis; and prioritize response to requests

4 Inherent Visualization Capability in BIMRoomID Current Faults Current Alarms Setpoint Temperature Actual Temperature Supply Airflow Supply Air Temperature SLC-401 22oC 22.4oC 150 l/s 18.6oC SLC-402 1 26oC 200 l/s SLC-403 23.1oC 400 l/s SLC-404 18oC 0 l/s - SLC-405 SLC-406 SLC-407 33.1oC 800 l/s 11.6oC BIM can be used for relationship & trend identification to support root-cause analysis in facilities management The collection and classification of work order requests creates a bottleneck in some FM contexts Case study conducted at Ryerson University to investigate the use of BIM and data analytics to support the resolution of maintenance requests Machine learning algorithms developed to classify maintenance requests; collect information to support root cause analysis; and prioritize response to requests

5 Dashboard Proof of ConceptBIM can be used for relationship & trend identification to support root-cause analysis in facilities management The collection and classification of work order requests creates a bottleneck in some FM contexts Case study conducted at Ryerson University to investigate the use of BIM and data analytics to support the resolution of maintenance requests Machine learning algorithms developed to classify maintenance requests; collect information to support root cause analysis; and prioritize response to requests

6 Challenge: 4 V’s of Big DataValue: Improved Building Performance (5000 sensors reporting per building; multi-year historical trends) Volume (Sampling every 15 seconds) Velocity (Text, integer, unstructured sources) Variety (Need to identify aberrant readings/ sensor faults) Veracity BIM can be used for relationship & trend identification to support root-cause analysis in facilities management The collection and classification of work order requests creates a bottleneck in some FM contexts Case study conducted at Ryerson University to investigate the use of BIM and data analytics to support the resolution of maintenance requests Machine learning algorithms developed to classify maintenance requests; collect information to support root cause analysis; and prioritize response to requests

7 Sensors->Cloud StreamingData Collection and Verification labeled maintenance requests available QA performed on 45,000 prior to use Data Mining used to provide insight on overall trends Machine Learning Term Frequency (TF) Term Frequency – Inverse Category Frequency (TF-ICF) Random Forest Visualization (Building Information Model) concurrent with above

8 Sensors->Cloud StreamingCloud analytics Queries/ Jobs Sensor Sensor Network Online Spark streaming Ingestion ID Lookup Table Clean Schema Cassandra Data Warehouse MQ (Kafka) Batch Spark Data Collection and Verification labeled maintenance requests available QA performed on 45,000 prior to use Data Mining used to provide insight on overall trends Machine Learning Term Frequency (TF) Term Frequency – Inverse Category Frequency (TF-ICF) Random Forest Visualization (Building Information Model) concurrent with above Visualization Analytics

9 Next Step: Mobile Sensing Stream-to-CloudTexas Instruments SimpleLinkTM Bluetooth Sensor Talk through example – L1 (Orange); L2 (green); consolidation of existing “cause codes” (blue) into L2 categories

10 Current work Brian Rampasad (Master’s): “Mobile Sensors Streaming to Cloud” Schema design and its influence on analytics latency Yar Rouf (Master’s): “Distributed processing for IoT” In a highly distributed IoT, how do we distribute the analytics to the edges?