Facility managers can take advantage now of artificial intelligence (AI) to revolutionize their roof and façade inspection data gathering with results that make possible strategic maintenance planning and scheduling with defensible budgets.
Done manually, the traditional approach is often expensive with the use of scaffolding, swing stages and other access equipment. It is time-intensive, exposes technicians to safety risks, is limited in the data gathered and subjective in the data’s interpretation. With traditional methods, it is challenging at best for a facility manager to make building maintenance and capital expenditure recommendations with confidence.
The barriers to entry in using the fast-emerging AI technology are low and begins with understanding AI-enabled assets, how they work and what integrating them into your facility assessment process can mean for your business.
A Primer on Facility Assessment-Related AI Technologies
AI leverages technology to more safely and thoroughly collect data, analyze it, use it to predict future needs and automate manual processes. Among the AI-enabled technologies that are revolutionizing facility assessments:
Drones with CV Cameras
Drones and cameras integrated with computer vision (CV) algorithms scan building surfaces to pinpoint defects. CV excels at identifying visible defects on exposed surfaces such as cracks, spalling, corrosion, leaks, discoloration, efflorescence, joint failures, and other forms of deterioration. Drones provide mobility and access to hard-to-reach areas and are equipped with GPS, stabilization, and sometimes LiDAR for navigation.
LiDAR/3D
LiDAR/3D scanning produces accurate surface geometry that can show changes in shape (deflections, bulging) which may indirectly indicate internal degradation. LiDAR (Light Detection and Ranging) is a remote sensing method that uses laser light to measure distances and build detailed 3D maps. How it works:
- A LiDAR sensor sends out very short bursts (pulses) of laser light.
- The laser hits an object (like a wall, tree, or road surface) and bounces back.
- The system measures how long it took for the pulse to return.
- Every laser pulse gives a 3D coordinate (X, Y, Z) point.
- Millions of these points form a point cloud (detailed 3D representation of the environment) that software converts into a surface or model for visualization, measurement, or simulation.
Thermal/Infrared Imaging
While not literally seeing inside, thermal/infrared imaging cameras can highlight temperature anomalies that suggest hidden problems (moisture intrusion behind walls, insulation gaps, electrical hotspots). Camera sensor types can include:
- RGB cameras (like regular cameras) offer high-resolution images/video.
- Thermal/infrared cameras can detect heat signatures (moisture intrusion, electrical hotspots).
- Multispectral sensors can detect material degradation. Multispectral sensors see “beyond visible light” to help reveal hidden defects.
Machine Learning
AI algorithms analyze historical maintenance data, sensor readings, and environmental factors. The analysis can be used to predict when an asset is likely to fail. This makes possible the scheduling of proactive maintenance before a breakdown occurs. How machine learning works:
- Data input – the machine (computer system learning the algorithm) is fed a lot of data (text, sensor readings, numbers, images).
- Training – the algorithms (set of step-by-step instructions the computer follows to solve a problem or complete a task) look for patterns or relationships in the data.
- Model creation – the system builds a mathematical model that captures those patterns.
- Prediction – the model is used on new data to make predictions which can inform recommendations.
- Continuous improvement – the ongoing input of data enables the model to continuously improve accuracy.
Natural Language Processing (NLP)
Facility Condition Assessments (FCAs) generate huge amounts of data. NLP algorithms can analyze unstructured text data from sources like inspection reports, technician notes, maintenance logs, work orders, even emails. NLP can make sense of this text and provide valuable insights that would be nearly impossible to extract manually from dozens, hundreds or even thousands of documents. How NLP works:
- NPL reads the data, such as a technician’s notes (“… cracking along south wall, high severity”)
- NLP converts the data into structured categories (Defect type: crack; Location: south wall. Severity: high).
- NLP automates organizing data and report analysis.
How the AI Technology Works Together
Flight and data capture: Drone flies along a pre-programmed path (or manually controlled) while its cameras take continuous images or video of the structure.
Image Processing: Data is uploaded to a CV-enabled platform and algorithms process each image frame-by-frame.
Defect detection: CV models look for patterns – linear features (cracks), color anomalies (corrosion or water damage), surface roughness (spalling or concrete loss), heat anomalies (insulation failures or leaks).
Classification and tagging: After each defect is classified (crack, rust, deformation), metadata (location via GPS, size, severity) is added.
Reporting / digital twin integration: Results are turned into a 3D map or digital twin (virtual replica of a physical object, system, or facility that continuously updates with real-world data) of the facility. It’s more than just a model – it’s a living, data-driven mirror of the real asset.
Getting Started
AI sounds intimidating, but it’s about layering new tools onto whatever the current process is. Starting out, facility managers can make the most of information already in hand – work with a consultant to gather all the inspection documents associated with a building, use NLP to read and organize the data, identify what is most degraded and prioritize those issues for maintenance or repair. Next steps could include:
- Continue FCAs as are done at present (manual inspections, contractor reports, maintenance logs).
- Identify pain points (slow inspections, inconsistent data collection, reactive repairs).
- Work with your consultant for a small use case such as a drone and CV for roof inspections (more safe, fast), or sensors for HVAC (track energy use, temperature, vibration).
- Learn how to read the AI results and trust the outputs starting with side-by-side comparisons (AI v human inspection).
- Prove return on investment and scale gradually.
Modern facilities generate a ton of data. AI can sift through this data to find hidden patterns, predict future trends, and identify areas for improvement enabling a shift to proactive, data-driven maintenance planning and creating strategic, justifiable budgets.
While in its early days for application to FCAs, AI is a powerful data gathering tool and models continue to improve as computational capacity increases. It can be a cost differentiator as getting consistent, quantifiable data – such as accurate numbers behind how much of a defect condition there is – informs scheduling, unit pricing and the best approach for repair.
While AI is a game changer in all areas relating to data gathering for façade and roof assessments, it is not perfect. There still is a need for an engineer / technician to validate some of the data coming out of these systems and to confirm what the data means for remedial property correction. Even so, using AI in FCAs empowers facility managers to proactively manage real estate assets and make fiscally responsible, forward-thinking decisions.
Jeff H. Poe, Jr., PE, RBEC, is a senior engineer with Terracon. He regularly prepares specifications and detailed drawings for structural and building enclosure repairs, replacements and restorations and is practiced in multi-family housing, K-12, higher education, healthcare, commercial and government projects.
Greg Walterscheid, eMBA, FMP, is a national director for Terracon’s Facilities Consulting practice focused on building performance consulting – providing building operators with expert and specialized consulting on all systems from the parking lot to rooftop, supporting owners’ efforts to extend system service life and operation while optimizing return on investment.
