Mobile video as a first-pass triage tool for live power equipment
On site, the first sign of trouble is rarely a reading — it is a sound, noticed by someone who cannot prove what they heard. A phone recording turns that impression into data that can be analysed, compared and acted on, in minutes, with no specialised equipment.
Field maintenance runs on a skill that is rarely written down: an experienced technician often hears a problem before any instrument reports it. Thermal overload, abnormal discharge, excessive vibration — all of them change the sound of a machine well before a sensor threshold is crossed.
So the most common starting point for a real diagnosis is not “I measured this” but “I heard that”. The difficulty is what happens next. An impression of a sound cannot be attached to a work order, compared against last month, or handed to a colleague. It has to become data before it can become a decision.
A phone recording is the cheapest available instrument that converts an auditory impression into something structured: a signal that can be measured, plotted, compared and argued about.
What makes mobile video useful is not the audio alone. It is that a phone records two different kinds of evidence on a single shared time base, with no configuration and no synchronisation work.
Separately, each is weak. Together, they let you test whether an acoustic feature actually belongs to the component you suspect — because both tracks carry the same clock.
Phone audio is not laboratory-grade. It is band-limited, automatically gain-controlled, and captured through a small microphone in an uncontrolled acoustic environment. What survives all of that is structure: equipment anomalies tend to produce consistent spectral patterns, and those patterns remain detectable even when absolute levels are not trustworthy.
The practical consequence is that you should read relationships and patterns, not absolute sound pressure levels. Four feature families do most of the work:
| Feature family | What to look at | What it points to |
|---|---|---|
| Frequency structure | Whether 50 Hz and its harmonics dominate or are persistently elevated; harmonic ratios; amplitude modulation | Electromagnetic imbalance, mechanical resonance, or load-driven amplification of an existing response |
| Band energy distribution | Energy concentration within bands rather than isolated peaks; sustained increases in a range | Structural resonance, enclosure radiation, or rising noise from auxiliary machinery such as fans and pumps — highly directional for field troubleshooting |
| Time-varying behaviour | Periodic fluctuation and rhythmic patterns in a time-frequency plot | Rotating components, attachment vibration, or load cycling |
| Impacts and transients | Short-duration impacts, sudden energy release, non-stationary events | Event-driven faults — arcing, collision, discharge — as distinct from continuous operating noise |
The distinction in the last row matters most in practice. Continuous noise tells you about a state; a transient tells you about an event. Confusing the two is the most common way a field acoustic assessment goes wrong.
Audio analysis can establish that something is abnormal. It cannot, on its own, establish what is abnormal. That is the video track's job, and it does it in three ways:
Mobile video does not replace a vibration analyser, an acoustic camera or an electrical test. Framing it as a cheap substitute for instruments is the fastest way to discredit it. Its value is upstream of all of them: it answers, in minutes, the three questions that determine what happens next.
| Question | What the recording provides | Where it leads |
|---|---|---|
| Is there an anomaly? | Spectral features compared against normal operation, instead of a subjective impression | Either close the item out, or continue |
| What is the likely cause? | Audio-visual correlation separating electromagnetic imbalance, structural resonance and electrical discharge | A hypothesis specific enough to test |
| How should we proceed? | Enough evidence to rank the options: temporary reinforcement, sensor validation, or escalation | Targeted vibration, acoustic or electrical diagnostics — on the right asset |
None of these is a definitive conclusion, and it would be wrong to present them as one. They are actionable insights that let a frontline decision be made now, with the reasoning recorded.
Effective maintenance technology does not have to rest on complex algorithms or expensive hardware. It has to extract stable, verifiable features from data that can actually be obtained, and turn them into interpretable judgements and defined next steps. Mobile video does that unusually well:
It bridges the gap between an experienced engineer's intuition in the field and the deeper diagnostic work that follows — providing evidence for the investigation while narrowing down where that investigation should point.
Stated plainly, so the method is not oversold:
Never compromise safe working distance or approach limits to get a better recording. If the equipment cannot be recorded safely from an approved position, it does not get recorded.
HVPACE is the trading name of Power Asset Condition Engineering Limited, a New Zealand-based supplier of test and diagnostic instruments and selected specialty equipment — such as cable sheath voltage limiters and other low-volume, mission-critical components — for high-voltage power systems. Backed by hands-on engineering expertise, including field-tested condition assessment techniques and AI-based signal analysis, we support customers with diagnostic guidance, methodology development and fault investigation, so they get the most value from the equipment they buy.