How a Transformer Vibrates — And What a Healthy Signature Looks Like
Every power transformer vibrates. Two distinct physical mechanisms drive it. Both carry information. Reading them correctly — in real time, across multiple sensors, corrected for temperature and weather — is what continuous condition monitoring actually requires.
Two Mechanisms, Two Signatures
The first mechanism is core magnetostriction. The silicon-steel laminations of the core change dimension slightly as the alternating flux magnetizes and demagnetizes them with each power cycle. The strain depends on the magnitude of the flux, not its polarity. The core therefore vibrates at twice the line frequency: 120 Hz on a 60 Hz grid, 100 Hz on a 50 Hz grid. Higher even harmonics appear as the core's magnetic and mechanical behavior becomes more nonlinear under heavier excitation. This vibration is present whenever the transformer is energized, including at no load, because it tracks the applied voltage.
The second mechanism is load-driven winding forces. Load current in the windings interacts with the leakage flux that the same current produces. The resulting Lorentz force on the conductors scales with the square of the current. A doubling of load current produces four times the force. These forces act in both the radial direction (outward from the core) and the axial direction (along the core axis). Their magnitude also depends on the clamping condition and the geometry of the coils, which is why two transformers carrying identical loads can produce different winding vibration signatures.
The core signature tracks voltage. The winding signature tracks current. That separation is the physical basis for distinguishing core behavior from winding behavior, and for detecting when either departs from what it should be.
Two Paths to the Tank Wall
Vibration reaches VIE's sensors through two distinct routes, and they carry different information.
The structure-borne path travels through solid contact: the core frame, clamping structures, tie rods, yokes, support beams, the base, and the welds and stiffeners of the tank wall. This path carries information about mechanical tightness and the integrity of the core frame and clamping system. A change in clamping force, a loosening lamination stack, or a shift in the structural coupling changes what arrives through this path.
The fluid-borne path travels as pressure waves from the vibrating core and windings into the liquid dielectric, then into the tank wall where the oil couples to it. The density and viscosity of the oil depend on temperature and condition. The velocity and amplitude of signals along this path therefore change with the thermal state and the quality of the oil. A change in oil condition changes what arrives through this path, even if the core and windings are mechanically undisturbed.
Why VIE Uses Three Axes, Not One
A single-axis sensor blends the structure-borne and fluid-borne signals together into one ambiguous trace. A triaxial sensor resolves them by direction.
The axis normal to the tank wall — the out-of-plane axis — responds primarily to the fluid-borne pressure waves striking the wall from inside. The two axes parallel to the tank surface (the in-plane axes) respond primarily to the structure-borne motion transmitted through the tank and its supports.
VIE tracks all three axes on each sensor, analyzes each one individually, and then looks at how the relationship between the out-of-plane and in-plane signals changes over time. A change that appears in the out-of-plane axis points toward a winding or oil-related shift. A change in the in-plane axes points toward a clamping or frame-related shift. Both arriving simultaneously indicates something more complex. A single-axis measurement cannot make that distinction. Three axes can.
Weather, Temperature, and the Co-Located Sensor
Vibration without thermal context produces false indications. Temperature changes the properties of the oil and the structural response of the transformer, so it changes the measured vibration signature even when nothing is wrong internally.
Each VIE sensor includes a temperature element in the same housing. This is co-location, not approximation. VIE measures surface temperature at the exact point where it measures vibration, removing the spatial mismatch that arises when temperature and vibration come from separate instruments in different positions on the tank.
VIE then couples those surface temperatures with local weather data: ambient temperature, solar load, wind, and humidity. Weather drives a large portion of tank-surface and top-oil temperature independently of internal load. VIE models the expected thermal state from weather and load, subtracts the environment-driven and load-driven variation, and isolates the residual. That residual is what indicates a developing fault. The same model corrects the temperature dependence of the fluid-borne path, which keeps the vibration comparison valid across seasons, day-night cycles, and variable load conditions.
This correction matters more as operating conditions grow volatile. A transformer serving a large data center or a renewable interconnection experiences rapid load swings that stress windings and clamping. A continuous, weather-corrected method observes the mechanical response to those swings as they happen, not between sampling visits.
A Virtual Model, Not a Learned Baseline
Most condition monitoring systems learn a baseline from the asset's own history and flag departures from it. That approach has a fundamental problem: it assumes the asset was healthy when monitoring began. Operators routinely install sensors on transformers of unknown prior condition. A baseline learned from a unit that already carries a defect absorbs the defect as normal, hiding the very problem the system exists to find.
VIE uses a different reference. From the approximate geometry of the specific transformer, VIE builds a virtual model of how a unit of that construction should vibrate and conduct heat. Every measurement is compared to this model, not to the asset's own past. This means VIE can identify a defect that was present when the sensors were installed, not only one that develops afterward.
VIE refines the virtual model with every sample. Sensors typically record 3 to 4 times per hour, so the model converges on an accurate, unit-specific reference continuously, over weeks and months of operation. It never treats an existing fault as the normal state.
Rate of Degradation, Not Just a Flag
Working from the virtual model, VIE's AI accounts for how load, temperature, and weather should move the expected signature. It separates a genuine departure from a benign operating swing. It identifies a developing problem early, where the signature first separates from the model, rather than where it crosses a hard threshold.
The second output is equally important: VIE trends the rate at which a parameter is changing. A slow drift and a sharp acceleration are different situations, and they call for different responses. The rate of change sets the length of the intervention window — whether an operator has weeks to plan a repair or days to act. That turns the platform from an alarm into a planning tool.
What a Healthy Signature Looks Like
A healthy transformer produces a vibration pattern that is stable and repeating, with amplitude and harmonic content consistent with what the virtual model predicts for its current load, voltage, and weather conditions.
Stability is the signal. When the pattern tracks the model closely over time, the core, windings, oil, and structure are behaving as expected. When it begins to separate from the model, that separation (its direction, its magnitude, and especially its rate) is the diagnostic information.
A healthy signature is not the same on every transformer. A unit at 80% load will show different absolute amplitudes than one at 30% load. A transformer serving a variable industrial load will show more dynamic variation than one on a stable distribution circuit. The virtual model accounts for all of these differences. What it cannot explain is what VIE flags.