5 Ways to Use Graph Neural Networks in Facility Maintenance

Artificial intelligence is taking the world by storm, whether by scientists, researchers and commentators speculating on its practicalities and pitfalls or by businesses and individuals tangibly applying it to make their lives easier. While we’ve covered some of the broad applications of AI in the business world on our blog (see our articles on workplace technology innovation), we wanted to dive into AI applications in the facility maintenance space, particularly the uses of graph neural networks in facility maintenance.

What are Graph Neural Networks?

First, let’s break down graphs and neural networks. Simply put, graphs are visual descriptions of how a set of objects are connected. For example, a marketing team might use graphs to measure the performance of our social media posts. Mostly, the graphs we use describe how types of content (the set of objects) perform over time (the connection).

Neural networks are the systems at the heart of deep learning, which is a subset of machine learning, that processes large amounts of data to make predictions. Neural networks are meant to mimic the way the human brain’s neurons communicate with each other using a system called backpropogation. In basic terms, once an AI machine makes a guess about a problem or offers a prediction based on existing knowledge, the system will then send information back to the machine to tell it whether it was right or wrong. This is the most common way to train AI machines.

Graph neural networks (GNNs) are neural networks that use data in graph form to make predictions. According to Nvidia, “Graph neural networks apply the predictive power of deep learning to rich data structures that depict objects and their relationships as points connected by lines in a graph.” GNNs are particularly useful for finding patterns among data points.

Uses of GNNs in Facility Maintenance

The nature of facility maintenance data lends itself well to improving facility efficiencies by leveraging the underlying graph structure (an object and its relationship to other objects) of facility-related data. Facility data includes energy use, occupancy levels, layout and design, and other systems that interplay to support a facility. GNNs can analyze these systems to improve their use and the overall efficiency and experience of the facility.

Here’s how GNNs can enhance facility management:

1. Enable Preventative Maintenance: The various systems and technologies within facilities are often interconnected; the success of one system is dependent on the efficacy of another. GNNs can analyze the complex data of these interconnected systems (including historical data) to identify hidden patterns to anticipate system and equipment failures, mitigating downtime.
2. Maximize Energy Efficiency: Optimizing energy consumption, especially in industries with high energy consumption, is critical to financial success and to appropriate stewardship of natural resources. In the same way that analyzing current and historical data of the facility’s systems can help with preventative maintenance, analyzing the energy consumption of different systems and machines within the facility can identify flaws in the architecture and system connections that waste energy. Plus, saving energy saves money!
3. Better Utilize Space: By collecting data from occupancy sensors, access control systems, and environmental sensors, GNNs can analyze how spaces are being used and by how many people. They can identify underutilized spaces, suggest layout changes, and help facility managers make better design decisions.
4. Improve Emergency Response: In analyzing the existing emergency response plans, layout, evacuation routes, and safety equipment, a GNN can identify potential bottlenecks or unaddressed safety concerns, run emergency simulations, and suggest improvements. This will mitigate risk by improving emergency response times.
5. Optimize Workflows: Especially in facilities with significant on-site presence, like a manufacturing facility, GNNs can analyze the dependencies and interactions between various workflow components to optimize workflows. This may include suggestions to reallocate resources and improve scheduling. GNNs can also place work orders for maintenance and track inventory. This all improves operational productivity.

Implementation & Security Concerns

There are many useful applications for GNNs in facility maintenance and for AI in general. Still, it’s important to acknowledge a few of the concerns about GNNs.

First, for GNNs to be truly effective, facility teams need to implement them correctly. The GNNs must be able to collect accurate data and receive quality training and validation.

Second, facilities need to carefully consider the security and privacy of the GNN system. Especially in facilities that handle sensitive data, like hospitals, maintaining a secure system for analyzing the data is critical to protecting the integrity of the facility and the people inside it.

With thoughtful, safety-first implementation, using graph neural networks in facility maintenance could revolutionize the efficiency of facilities of all types! Let us know what you think about the use of AI in the FM space in the comments below.