Rotary drum granulators are heavy mechanical systems designed for continuous 24-hour duty. A single unplanned stop on a granulation line costs more than just the maintenance intervention – it costs the production lost during startup, the off-spec material generated as the process restabilises, and the recycle excursion that typically follows when the line comes back up cold. For a plant running three shifts a day, 330 days a year, the difference between a reactive maintenance culture and a predictive one tends to show up clearly in the annual production figures.
The good news is that drum granulators are well-suited to condition-based monitoring. The key failure modes – bearing wear, gear mesh deterioration, liner failure, shell misalignment – all produce detectable signals long before they cause a breakdown, provided you’re measuring the right parameters and comparing them against an established baseline. We’ll cover which parameters to monitor, where sensors belong, how to set meaningful alarm thresholds, and how to build a practical predictive maintenance programme around them.
Why Drum Granulators Are Well-Suited to Condition Monitoring
The drum granulator has a small number of critical mechanical systems, each of which produces a measurable signature during normal operation. The trunnion roller bearings carry the drum’s full rotating mass and produce characteristic vibration and temperature patterns that change as they wear. The ring gear and pinion mesh at a known frequency based on the number of teeth and the drum speed, and changes in that mesh frequency component in the vibration spectrum can indicate tooth wear or misalignment. The shell itself moves at a known rotational speed, and deviations from the expected riding ring tracking pattern indicate alignment drift that will eventually accelerate trunnion roller and riding ring wear.
Because these systems run under steady loads during normal granulation, the baseline condition is stable and relatively easy to characterise, and deterioration tends to develop gradually rather than catastrophically, which gives a meaningful warning window between the first detectable signal change and the point at which intervention is required. The slow speed cuts both ways, though, and it is worth being straight about it: a drum turning at around 10 RPM is a difficult case for conventional vibration monitoring, not an easy one. Defect energy at the trunnion bearings is low, and because acceleration amplitude falls with the square of frequency, a general-purpose accelerometer will often return nothing usable at these speeds. Low-frequency accelerometers, velocity or displacement measurements, and techniques such as acoustic emission or shock pulse are what earn their place on the drum itself. The drive motor and gearbox, which turn at normal speeds, are conventional measurement points.
Modern drum granulators designed for condition monitoring readiness – including those with pre-installed vibration and temperature sensor mounting pads on the trunnion pedestals and bearing housings – allow this programme to be implemented without structural modifications to the machine. Where the mounts are already in place, adding IIoT sensors is a matter of installation and commissioning the data acquisition system rather than a fabrication project.
Parameters to Monitor
Trunnion roller bearing vibration. Vibration measured on the trunnion roller bearing housings is the primary indicator for bearing condition, shell imbalance, and drive-induced excitation. Accelerometers mounted on the bearing housing (not on the drum shell itself, which rotates) measure both overall vibration level and the frequency spectrum. Changes in the overall vibration level provide a simple alert trigger; spectral analysis of the time series provides more detailed diagnostic information – bearing defect frequencies, gear mesh components, and sub-synchronous signals associated with shell wobble or riding ring eccentricity.
Trunnion roller bearing temperature. Rising bearing temperature at constant operating speed and load is one of the more reliable early indicators of inadequate lubrication or bearing degradation. An RTD or thermocouple installed in the bearing housing provides a continuous temperature reading that can be trended and alarmed. The alarm threshold should be set relative to the established baseline temperature rather than against an absolute value, since normal operating temperature varies with ambient conditions, drum speed, and bed load.
Drive motor power draw. Monitoring the power consumed by the drive motor – available from the VFD in most modern installations – provides a continuous indication of the torque demand on the drum. Filling factor is not directly measured on a running plant, so in practice this signal has to be normalised against total material flow, meaning fresh feed plus recycle, rather than against a filling factor figure. Rising power at constant drum speed and constant total flow can indicate increased bed friction (for example, from product build-up on the liner or from an over-filled bed), gear mesh deterioration, or trunnion roller misalignment. It’s a relatively coarse signal compared to vibration, but it’s easy to access without additional sensors on plants where the VFD is already in place.
Gear mesh temperature. For spur-tooth ring and pinion drives, an infrared spot sensor aimed at the tooth faces, or a periodic thermographic survey, is the practical way to read gear mesh condition and lubrication effectiveness. A fixed contact sensor near an open mesh largely reads the air around it, since the mesh is open and moving, so non-contact measurement is what gives a usable tooth-face temperature. Sustained elevated temperature at the gear mesh, when not explained by increased ambient temperature or load, tends to point toward inadequate lubrication or surface degradation on the tooth faces.
Riding ring tracking. Some axial float of the riding ring is normal, and often deliberate: trunnion rollers are commonly skewed slightly so the ring migrates back and forth across the roller face, which spreads wear instead of concentrating it in one band. What matters is the pattern rather than the movement itself. A ring that drifts consistently in one direction and sits hard against its thrust roller indicates a thrust imbalance that, if uncorrected, leads to accelerated wear on the riding ring face and the trunnion roller edge. This is typically monitored by periodic manual measurement during inspection stops, though some plants install proximity sensors to provide a continuous axial position signal.
Sensor Placement
Vibration sensors belong on the stationary components of the drum assembly, not on the rotating shell. The correct locations are the trunnion roller bearing housings, the gearbox input and output bearing housings, and the drive motor bearing housings. Mounting an accelerometer on the drum shell itself will produce a rotating signal that’s difficult to interpret and doesn’t provide useful bearing diagnostic information.
For bearing temperature, the sensor should be positioned in the bearing housing at a point that sees the bearing’s lubricant exit temperature, not the ambient temperature of the surrounding structure.
Cable routing from the sensors to the data acquisition system should use a path that avoids the hot zones around the dryer and cooler if those are in proximity, and should be protected against the mechanical and chemical environment of the granulation area – conduit or armoured cable is appropriate in most fertilizer plant environments.
Establishing Alarm Thresholds
One of the most common mistakes in setting up a condition monitoring programme is applying generic alarm thresholds from an international standard without establishing plant-specific baselines first. ISO 20816 (and the earlier ISO 10816 series) provides a framework for vibration severity classification by machine type and mounting, but note the scope: its measurement range is aimed at machines running from roughly 600 RPM upward. It applies to the drive motor and gearbox measurement points on a granulation line; it does not cover the trunnion and shell points, which turn two orders of magnitude slower. Used inside that limit it’s a useful starting point – but the absolute vibration level that should trigger an alert on a specific drum depends on that drum’s baseline condition, not on a universal table value.
The right process is to collect baseline measurements during a period of known good operation – ideally during commissioning or shortly after a planned maintenance stop when bearings have been inspected and confirmed in good condition – and use those measurements to set the alarm thresholds. A common approach is:
Alert threshold: baseline value + a defined increment, typically 50–100% above baseline for overall vibration, depending on the machine and the rate of change observed. This should be treated as an indicative range rather than a fixed rule; the appropriate increment depends on how stable the baseline is and how quickly the machine has historically shown deterioration once vibration starts to rise.
Action threshold: worth giving a number rather than defining it only as higher than the alert level. A common convention is around four times the baseline overall level, against roughly twice the baseline for the alert – the point at which investigation is required regardless of other operational priorities.
For bearing temperature, a threshold of 10–15°C above the established baseline temperature under comparable operating conditions is a reasonable alert level for most drum granulator applications, though the specific value should be set against your bearing type, lubricant specification, and observed temperature stability. These are indicative ranges – the right thresholds are ones that give you enough warning time to plan an intervention, without generating so many false alerts that the monitoring system gets ignored.
Moving from Reactive to Predictive
The shift from reactive to predictive maintenance on a drum granulator happens in stages, and it doesn’t require implementing everything at once.
The first stage is data collection. Even before acting on any monitoring data, simply establishing what the normal vibration, temperature, and power signatures look like during stable operation gives you the baseline against which you’ll interpret future readings. This is the stage that most plants skip because it doesn’t feel like it’s doing anything – but without a baseline, all subsequent measurements are contextless.
The second stage is trend monitoring. Rather than waiting for a threshold alarm, routine review of trending data catches gradual changes that stay below the alert threshold but are moving in a direction that indicates developing deterioration. A bearing whose vibration level has been rising consistently over six weeks, even if it hasn’t yet reached the alert threshold, deserves investigation before it does.
The third stage is integration with planned maintenance. The value of condition monitoring is that it allows maintenance to be scheduled on the basis of actual equipment condition rather than fixed time intervals. A bearing that’s still showing stable condition at the scheduled replacement interval may be safely deferred; a bearing showing deteriorating condition ahead of its scheduled replacement date can be addressed at the next planned stop rather than waiting for failure.
This progression – from threshold alarms to trending to condition-based scheduling – represents a genuine improvement in maintenance efficiency, but it requires consistent data collection, operator discipline, and a maintenance management culture that trusts the monitoring data enough to act on it.
Ceylan Machine & Process manufactures rotary drum granulators with pre-installed vibration and temperature sensor mounting provisions on trunnion pedestals and bearing housings, designed for IIoT-based condition monitoring integration. For technical enquiries on condition monitoring specifications or to discuss your maintenance programme requirements, contact our engineering team.
