A case study in predictive maintenance: Soundsensing & CloudAMQP

The most costly equipment failures are the ones no one sees coming. In hotels, office buildings, and other commercial spaces, HVAC (Heating, Ventilation, Air Conditioning, and Cooling) systems often malfunction in silent, slow motion. This creates a 'blind spot' where a minor, fixable issue—like a rattling motor or an unusual vibration—can quickly escalate into an emergency repair or costly hours of downtime.

Soundsensing Soundsensing aimed to eliminate this risk by building a robust, predictive maintenance platform that ensures systems are monitored, protected, and consistently running.

Device overview image from Soundsensing.com

The challenge: from raw data to real-time insights

Soundsensing’s solution relies on a network of intelligent sensors placed directly inside ventilation systems. These sensors are the "eyes and ears" of the operation, continuously monitoring for subtle changes in sound, vibration, and temperature.

The challenge was clear: How do you get a constant stream of high-volume data from these sensors and turn it into actionable insights in real time, all while ensuring the system remains stable?

The solution: a decoupled system with RabbitMQ (hosted by CloudAMQP)

To manage this data flow, Soundsensing chose the message queueing software RabbitMQ as the central nervous system for its solution. At its core, a message queue (or "queue") works much like a line at a coffee shop: requests (or messages) come in, wait, and are processed in order (First-In, First-Out, FIFO).

In system architectures, queues create a robust, asynchronous link between different components. This mechanism achieves decoupling, which is precisely what most systems handling a constant, high-volume flow of sensor data require for stability.

The sensors as data producers

The sensors act as tireless data producers. They send their data via an IoT gateway, which sends the data as messages to the queues in RabbitMQ. This asynchronous handling enables sensors to continue their tasks uninterrupted, eliminating the need to wait for data analysis.

This approach ensures stability. If a downstream server were to go offline, no data would be lost. All messages are held safely in RabbitMQ's queue, waiting to be processed when the system is ready.

Jon Nordby, a co-founder at Soundsensing, highlights this key benefit:

"We've trusted RabbitMQ for years. It's been the perfect fit for us from day one, acting as our reliable data traffic controller. Our sensors just fire off their data as messages to the queue, and they don't have to wait for the analysis to be done. It completely decouples our data collection from our processing, which is critical for system stability and reliability."

Error detection mapping, image from Soundsensing

The machine learning models: Processing and interpreting the sensor data

On the other side of the broker is Soundsensing's analysis system, powered by machine learning models, which acts as the consumer. It picks up messages from the queue one at a time. By constantly comparing the incoming data to what a healthy system "sounds" and "feels" like, these models can quickly and accurately detect anomalies that signal an impending problem. When issues are detected, the system automatically sends alerts to property managers, allowing them to take preventive action before equipment fails.

The result: predictive maintenance at scale

Thanks to this architecture, Soundsensing's solution is both scalable and reliable. RabbitMQ — operated by CloudAMQP as a managed service — ensures that every message is delivered correctly. The system's scalability allows property managers to easily expand their monitoring capabilities by simply adding more sensors without worrying about infrastructure limitations.

The data system helps spot problems before they happen, saving time and money for building managers. Benefits include:

Fewer surprises: Keep systems running smoothly and avoid unexpected breakdowns.
Lower costs: Replace parts when needed, not in a crisis.
Less downtime: Fix issues early and keep operations running.
Smaller environmental footprint: Fewer site visits mean lower emissions and less wasted travel.

Soundsensing helps property managers stay ahead of problems rather than react to them.

This article was written by Lovisa, part of the team at CloudAMQP. We manage the RabbitMQ platform that powers Soundsensing’s predictive maintenance pipeline. CloudAMQP runs and scales RabbitMQ for thousands of companies, such as Soundsensing. Have questions about message queues or reliability at scale? Reach out to me anytime!