Use Case
Predictive Anomaly Detection from Building IoT Data
Built an automated anomaly-detection system using IoT sensor data from HVAC, power, water, and other building equipment.
Overview
Consolidated a massive volume of IoT sensor data from building equipment and quantified anomaly signs as scores for visualization. The data-driven monitoring platform enabled both workload reduction and faster response.
Benefits
- Automated equipment monitoring through data-driven anomaly detection
- Standardized algorithms accommodate property-count growth
- Reduced operational costs through automation and faster response
Target Industries
- Building Facility Management
- Energy
- Facility Operations
Challenges Before / Changes After
Before
- Huge volumes of sensor data monitored visually
- Threshold judgments depend on individuals with inconsistent criteria
- Monitoring cannot keep up as properties grow
After
- Data-driven anomaly detection automates monitoring
- Standardized algorithms accommodate property-count growth
- Automation and faster response realized
Implementation
Sensor Data Consolidation & Preprocessing
Consolidated varied data sources (HVAC, power, water, etc.) with missing-value imputation and normalization.
Multi-Method Anomaly Detection Model
Combined statistical methods and machine learning for anomaly detection tailored to equipment characteristics.
Anomaly Score Visualization & Alerts
Visualized anomaly scores on a dashboard and designed alert-notification workflows.
Input Formats
IoT sensor data (temperature, humidity, power, water volume, etc.)
Output Formats
Anomaly scores, alert notifications, analysis reports
Integrations
IoT platforms, BMS, notification systems
Project Summary
Team & Timeline
- PoC → operational design → production rollout
- Facility management / IoT platform team / a.s.ist engineers
- Built data-processing infrastructure on cloud
Outcomes
- Reduced lead time for anomaly detection
- Scalable expansion of monitored properties
- Reduced staffing for monitoring operations
Contact Us
We propose end-to-end support — from PoC through operational design — for anomaly detection using IoT data.
- Consolidation and preprocessing of IoT sensor data
- Anomaly detection model construction and validation
- Alert notifications and visualization dashboards