• multimodal
  • Condition Monitoring
  • Predictive Maintenance

Multimodal Condition Monitoring: What Vibration, Ultrasound and Temperature Each Catch

Alex Vedan

Updated Nov 16, 2026

10 min.

Key Points

  • Vibration, ultrasound and temperature each respond to a different stage of a developing fault, so each one moves first on different failure modes.
  • Ultrasound usually flags lubrication and friction problems before vibration or temperature change on rolling-element bearings.
  • The right signal mix depends on the asset, its speed and its failure modes, which is why the decision table below starts from the fault.

Multimodal condition monitoring means watching one machine through mechanical signals (vibration, ultrasound, temperature) and electrical signals (current, voltage) together. A dry bearing on a centrifugal pump, a worn gear set and an overloaded motor winding each show up first in a different signal, and often days or weeks apart.

This guide is a working reference for reliability engineers and maintenance managers. It covers what each signal catches, where each one needs help, and a fault-by-signal decision table for pumps, motors, gearboxes, fans, conveyors and screw compressors. A single signal sees some faults late, such as lubrication starvation in vibration data, which is why combining signals reduces monitoring blind spots.

What Is Multimodal Condition Monitoring?

Multimodal condition monitoring means tracking mechanical and electrical signals on the same asset. Tractian combines Smart Trac Ultra, which captures vibration, ultrasound, temperature and rotational speed from one sensor, with Energy Trac for electrical data on the motor circuit.

The signals are independent because they respond to different physics. Vibration responds to mechanical forces, ultrasound to friction, impacts and turbulence above 20 kHz, and temperature to heat from friction, load and electrical losses. When two of them move together, the probability of a false alarm drops; when a single one moves, the pattern itself narrows down the cause.

Tractian multimodal monitoring combines these mechanical signals with electrical and operational data and the asset's maintenance history, so the alert says which component is degrading, how fast, and what to inspect.

What Does Vibration Monitoring Catch?

Vibration catches mechanical faults that change how a rotating shaft moves: unbalance, misalignment, looseness, bearing defects and gear wear. It is the richest single signal on rotating equipment because each of those faults leaves energy at a predictable frequency.

Faults vibration reads clearly

  • Unbalance: a rising peak at 1x running speed in the radial direction, common on fans with buildup on the wheel.
  • Misalignment: 1x and 2x peaks with high axial readings across a coupling.
  • Looseness: a series of running-speed harmonics, sometimes with half-order peaks.
  • Bearing defects: energy at the ball pass frequencies of the outer race (BPFO) and inner race (BPFI), clearest in the envelope spectrum.
  • Gear wear: gear mesh frequency with sidebands spaced at the speed of the damaged shaft.

Learn more about vibration analysis.

Where vibration needs a second signal

Vibration responds late to lubrication starvation. A bearing running on a thin film generates friction and high-frequency noise well before the race or rolling elements are damaged enough to produce clear defect frequencies.

Slow-speed machinery is the other gap. Below roughly a few hundred RPM, impacts carry little energy and velocity readings stay flat while a defect develops. Load and speed changes also move vibration amplitude on healthy machines, so a rise on a variable-duty compressor needs operating context before anyone opens a work order.

What Does Ultrasound Catch?

Ultrasound catches friction, impacts and turbulence in the range above human hearing, which makes it the earliest of the three signals for lubrication problems on rolling-element bearings. It also picks up cavitation in pumps and early bearing damage on slow-speed shafts.

Lubrication and friction

When grease thins out or contaminates, metal surfaces start touching through the film. That contact raises the ultrasound noise floor first. Temperature follows as friction builds, and vibration shows bearing defect frequencies only after surface damage begins.

Continuous ultrasound also changes how lubrication gets done. A technician can watch the reading while greasing and stop when it settles back to baseline, which avoids both under-greasing and the seal damage that comes with over-greasing.

Slow-speed bearings and early defects

On conveyor pulleys, kiln rollers, mixers and paper machine rolls, ultrasound picks up impact energy that velocity readings miss at low RPM. On pumps, the collapse of vapor bubbles during cavitation produces broadband ultrasonic noise that shows up before impeller erosion does.

Ultrasound does not resolve which component is at fault as precisely as a vibration spectrum. That is why it pairs naturally with vibration: ultrasound gives the early warning, and the vibration spectrum confirms where the problem is and how far it has progressed.

What Does Temperature Catch?

Temperature catches heat from friction, overload and electrical losses, so it moves first on overloaded motors, blocked cooling and failing cooling fans. On bearings it usually confirms a problem that ultrasound or vibration already flagged.

A rising bearing housing temperature next to a rising ultrasound reading separates a real lubrication issue from sensor noise. On motor frames, temperature rising with current points to overload or voltage imbalance. Ambient changes and duty cycle move temperature too, so trends need to be compared against the asset's own baseline at the same load.

Contact temperature on a sensor is one point. Infrared thermography surveys complement it on electrical panels, steam traps and refractory, where a thermal image covers more surface than a mounted sensor.

Which Signal Sees Each Fault First?

The first signal to move depends on the failure mode: ultrasound leads on lubrication and slow-speed bearings, vibration leads on unbalance, misalignment, bearing race damage and gear wear, and temperature leads on overload and cooling problems. The table below maps common faults on named assets to the first signal, the confirming signal, the typical warning window and the first action.

Asset and faultFirst signal to moveConfirming signalTypical warning windowFirst action
Centrifugal pump: drive-end bearing lubrication starvationUltrasound: friction raises the high-frequency noise floorBearing housing temperature, then high-frequency vibration once metal contact startsWeeks to monthsRegrease the bearing and confirm ultrasound drops back to baseline
Centrifugal pump: cavitationUltrasound and broadband high-frequency vibrationSuction pressure and flow from the process sideDays to weeks before impeller erosion showsCheck suction conditions, NPSH margin and valve positions
Electric motor: outer race bearing defectVibration: energy at the bearing defect frequency (BPFO) in the envelope spectrumUltrasound; temperature rises lateWeeks to monthsPlan the bearing replacement for the next scheduled stop
Electric motor: winding overheating from overloadTemperature at the motor frameMotor current and voltage balanceHours to daysCheck load, cooling fan and supply voltage before the insulation degrades
Helical gearbox: gear tooth wearVibration: gear mesh frequency with growing sidebandsOil analysis wear metals; gearbox temperatureWeeks to monthsInspect tooth contact, check lubricant and order the gear set
Centrifugal fan: unbalance from buildup on the wheelVibration: rising amplitude at 1x running speedTemperature rarely movesWeeks to monthsClean the wheel and balance it in place
Belt conveyor drive: motor to gearbox misalignmentVibration: 1x and 2x peaks, strong axial componentCoupling and bearing temperatureWeeksLaser-align the coupling and check for soft foot
Slow-speed conveyor pulley: early bearing defectUltrasound: impacts too weak for velocity readings at low RPMVibration in the acceleration envelope as the defect growsWeeks to monthsLubricate, then inspect the bearing at the next planned stop
Rotary screw compressor: airend bearing defectVibration: rising bearing-defect peaks in the envelope spectrumUltrasound on the bearing housing; discharge pressure and duty cycle unchanged, which rules out a load changeWeeks to monthsPlan the airend bearing inspection for the next scheduled stop

How to read the warning windows

The warning window is the typical time between the first measurable change and functional failure, the P-F interval for that fault. These ranges come from general condition monitoring practice and move with speed, load, lubricant and how often the asset is sampled. A pump bearing that starts and stops many times a day will run through its window faster than the same bearing on steady duty.

Two habits make the table useful in practice. Start from the failure modes that matter most on the asset, then check that the signal mix in place sees each of them early enough to plan the repair. A fault first seen in temperature, with a window of hours, needs a different response plan than one caught by ultrasound months ahead.

How the Signals Work Together on One Asset

Signals work together by confirming or ruling out each other's readings: one signal raises the question, and a second one answers it.

Walkthrough 1: a centrifugal pump bearing running dry

Picture an ANSI process pump on a cooling water loop, driven by a 75 hp motor. Its drive-end bearing is on a fixed regreasing schedule, and the last shot of grease went in two months ago.

  1. Ultrasound rises first. The high-frequency noise floor on the drive-end bearing climbs above its baseline while vibration and temperature stay flat.
  2. Temperature follows. A few days later the bearing housing runs several degrees warmer at the same load, which confirms friction and rules out sensor noise.
  3. The team acts before vibration moves. A technician regreases the bearing while watching the ultrasound reading and stops when it settles back to baseline.
  4. The fix is verified. Ultrasound holds at baseline over the next week, and no bearing defect frequencies ever appear in the vibration spectrum.

The outcome is a ten-minute grease job on a planned route instead of a bearing replacement and seal rebuild after an unplanned stop.

Walkthrough 2: a screw compressor with a developing airend bearing defect

Now take an oil-injected rotary screw compressor feeding plant air on a demand that changes through the day. Vibration on the airend climbs above baseline, and because load changes also move vibration on healthy compressors, the rise alone does not confirm a fault.

  1. Mechanical layer: the vibration rise shows energy at a bearing defect frequency in the envelope spectrum, and ultrasound on the same bearing housing climbs with it. Two independent signals moving together point to a real defect.
  2. Electrical layer: current signature and power factor stay stable and balanced, which points away from the drive motor and toward the airend.
  3. Operational layer: discharge pressure and duty cycle sit in their usual range, so a change in demand does not explain the rise.

The planner schedules an airend bearing inspection for the next planned stop and orders the bearing set ahead. Checking operating context first is what separates a real defect from a load-driven rise on a healthy machine.

Energy and current data add one more independent check on the same machine. Learn more about how multiple data streams improve condition monitoring.

Which Signals Should Each Asset Get?

Each asset should get the signals that see its most likely failure modes early enough to plan a repair, so the mix follows the asset's criticality and failure modes. A criticality analysis sets which assets get continuous coverage in the first place.

  • Electric motors: vibration for bearings, unbalance and misalignment; temperature for overload and cooling; current where winding and rotor bar faults are a concern.
  • Centrifugal pumps: vibration plus ultrasound, since lubrication and cavitation lead the failure list; process pressure and flow help separate cavitation from mechanical faults.
  • Gearboxes: vibration for gear mesh and bearing faults, temperature for lubricant condition; periodic oil analysis adds wear-metal trends on large units.
  • Fans and blowers: vibration first, since unbalance and looseness dominate; ultrasound on bearings that are hard to reach for manual greasing.
  • Slow-speed conveyors, mixers and rolls: ultrasound first, with vibration in the acceleration envelope to track defect growth.
  • Screw and centrifugal compressors: vibration, temperature and operational context together, because load changes move readings on healthy machines.

Where Electrical and Operational Signals Fit

Electrical and operational signals cover faults and context that mechanical sensors on the machine can't measure directly. Winding insulation breakdown, rotor bar damage and voltage imbalance show up earliest in current and voltage data, read with motor current signature analysis.

Operational data (pressure, flow, duty cycle, setpoints) explains why a mechanical reading moved. Tractian adds the electrical layer with Energy Trac, installed non-invasively without stopping the process, and correlates it with the mechanical signals on the same asset.

How Tractian Runs Multimodal Condition Monitoring

Tractian runs multimodal condition monitoring by pairing Smart Trac Ultra, which measures vibration, ultrasound, temperature and RPM in one device, with Energy Trac electrical monitoring on the same asset, plus operational data where failure modes call for it. Smart Trac Ultra combines always-on ultrasound and vibration in a single sensor, with sampling from 0 to 64,000 Hz.

On the analysis side, Tractian's industrial AI checks a deviation against the asset's learned baseline under its current load, correlates the other signals and maintenance history, and returns a diagnosis with the probable cause and a recommended action. Auto Diagnosis covers the failure modes in its detection library, and human review by Tractian analysts is available when a case needs it.

  • Fast installation: sensors mount with adhesive or a drill-and-tap stud, connect over LTE with no plant Wi-Fi or cabling, and install in a day with support from Tractian specialists.
  • Built for harsh areas: IP69K sealed, ATEX/IECEx compliant and Class I Div 1 certified, with NFPA 70 Class 1, 2, and 3 (Division I) compliance.
  • Connected to existing systems: alerts and work orders flow into the CMMS, ERP and EAM tools a plant already runs through Tractian's integrations.
  • Scoped to the asset: an on-site Asset Risk Assessment sets which assets get which signals, based on criticality and failure modes.

Buyers comparing hardware can line up vibration and ultrasound sensors side by side on signals, connectivity and certifications.

Frequently Asked Questions

Is multimodal condition monitoring the same as sensor fusion?

Sensor fusion is the analysis step inside multimodal condition monitoring. Multimodal monitoring covers measuring mechanical and electrical signals on the same asset; sensor fusion is how software correlates them into one diagnosis. A program can collect several signals without fusing them, which leaves the cross-checking to people.

Can ultrasound replace vibration monitoring?

No, ultrasound gives earlier warning on lubrication, friction and slow-speed bearing faults, while vibration identifies the specific fault and component through its frequency content. Unbalance, misalignment and gear mesh problems are read from the vibration spectrum. The two work best measured together on the same bearing.

How early can ultrasound detect a lubrication problem?

Ultrasound typically detects lubrication starvation before temperature rises and before vibration shows bearing defect frequencies, often weeks ahead of surface damage. The exact lead depends on bearing size, speed, load and grease type. Continuous measurement matters here, because a monthly route can miss the start of the rise.

Does every asset need vibration, ultrasound and temperature?

No, the signal mix should follow each asset's failure modes and criticality. A fan dominated by unbalance risk may need vibration and temperature, while a slow-speed conveyor pulley benefits most from ultrasound. A criticality and failure mode review sets the scope before sensors go in.

What sampling rate does a vibration and ultrasound sensor need?

The sampling rate has to be at least twice the highest frequency you want to analyze. Bearing defect and gear mesh analysis on fast machines needs several kHz of bandwidth, and ultrasound sits above 20 kHz. Tractian's sensor samples from 0 to 64,000 Hz.

See Which Signals Your Critical Assets Need

Start with the three assets whose failure would stop production, list their top failure modes, and check them against the decision table above. Tractian can map that list to a sensor plan and show live vibration, ultrasound and temperature data on your own machines. Explore the Tractian condition monitoring platform.

Alex Vedan
Alex Vedan

Director

Alex Vedan, Marketing Director at Tractian, develops impactful strategies that empower industrial clients across North America and LATAM to achieve operational excellence. By aligning innovation with customer needs, he ensures Tractian solutions drive meaningful improvements in efficiency and reliability.

Share

Start Exploring Tractian Condition Monitoring