Wireless Condition Monitoring in a High-Density Oven Environment

The Challenge

Every condition monitoring deployment starts with the same question: where do you begin? Even for experienced teams, the answer depends heavily on what the customer already knows about their own assets. A criticality map and a failure history are invaluable, but rarely available in the detail you would want. The practical approach is to start small, focused on the assets causing the most pain.

For this customer, a medical equipment manufacturer, the pain was clear. A mixed population of critical assets needed eyes on them: motor-gearbox assemblies driving rotating oven bases (slow-speed), oil temperature on a pressure test stand, and cooling fans feeding airflow to the ovens from ceiling height.

They had tried condition monitoring before. A wired system had been installed previously, but cables were damaged during production and, more often, disconnected during maintenance cycles and never reconnected. The data went dark. Confidence in the approach went with it.

So the brief was specific: no wires.

The Environment

The site was not an easy one. The ovens are large, around 20 feet tall and 15 feet wide and deep, and arranged with just five feet between them. The motor-gearbox assemblies sit at the base of each oven, in tight, awkward spaces. The cooling fans are mounted at ceiling height and enclosed, making access and line-of-sight difficult for any sensing technology.

Any wireless system deployed here had to cope with steel, heat, height, and density, all at once.

The Solution

We deployed a wireless architecture built around multiple cellular-connected gateways. Cellular was a deliberate choice: it sidesteps the IT approvals, firewall rules, and network segmentation that so often slow these projects down on a customer's own infrastructure.

The gateways run an iTnnovate data server and operate as a shared pool rather than fixed pairings. The first gateway to see a sensor reads its measurement and forwards the raw acceleration data to the cloud. This matters in a dense, metal-heavy environment where signal paths are unpredictable: there is no single point of failure between a sensor and the platform.

In the cloud, the backend recognises each sensor by its UUID (Universally Unique Identifier), places the measurement into the plant's machine hierarchy, and processes the raw data into frequency and overall readings for acceleration, velocity and demodulation, applying the appropriate alarm thresholds. Sensor settings, such as sample rate, sample trace length, and collection interval, are managed centrally, so the on-site team is not touching device configuration.

For the slow-speed applications, we specified MEMS sensors. These operate reliably at the low frequencies this application demands, reading all the way down to DC, which traditional ICP accelerometers cannot do. (A useful field check: a disconnected triaxial MEMS sensor sitting on a bench will show 1g on its vertical axis. That is simply the effect of gravity, not a fault.)

Results After Two Months

The system has already earned its place. Within the first two months of operation, it identified belt slip and imbalance on the Air Conditioning Fan, issues that would previously have run until failure or a symptomatic inspection caught them. The findings are now with the engineering team, who are evaluating a design change to eliminate the root cause rather than simply repair the symptom.

That shift, from reactive repair to design-level improvement, is the real return on a condition monitoring programme. It starts with a single finding.

What's Next

Phase two extends the programme to additional assets and further work areas across the site. We are also introducing motion-detect measurement for standby equipment and compressors, assets without a predictable operating cycle, where time-based sampling wastes battery and misses events. Motion-triggered capture solves both problems.

The broader lesson from this deployment is a familiar one: start focused, prove the value on the assets that hurt most, and expand from there.