Starting a Condition-Monitoring Programme Without Buying Everything at Once

A condition-monitoring programme is a commitment, not a purchase. The failure mode is rarely the instrument: it is the programme that starts with eighty machines, falls behind by month two, and quietly stops — leaving data nobody trusts and alarms nobody answers. A half-run programme that nobody acts on is worse than none, because it teaches the plant that vibration monitoring does not work. So the first decision is not which DAQ to buy. It is which machines earn a place on the route.

Rank every candidate on two axes: what a failure costs (production stop, safety exposure, spare-part lead time) and how likely a failure is (duty, age, history, whether it has a standby). The unspared, high-consequence machines in the hot corner go on route one. A first route of ten to fifteen machines measured properly every time beats eighty measured badly — and the matrix gets reviewed each quarter, so nothing is excluded forever.

A route is a fixed list: which machines, which points on each machine, in which order, at which settings. Forty machines at six points each, twice a month, is a day's work with a portable chain — a PhonoVibe Q or O, an accelerometer, and TVIB fit in a carry bag, and a two-channel PhonoVibe D covers a smaller starting route. The discipline that makes the data comparable is boring on purpose: same sample rate, same frequency range, same averaging count, every visit.

Pick an interval you can sustain, not the one that sounds impressive. Twice a month is a common rhythm for critical rotating machinery, but a monthly route walked every month beats a weekly route abandoned after six weeks. Put the route on the maintenance calendar with a named owner — consistency is worth more than frequency, because the trend only exists if the visits keep happening.

The single most valuable measurement in the life of a programme is the one taken while the machine is healthy, because every diagnosis you make afterwards is a comparison against it. Machines have personalities: a perfectly healthy gearbox does not read like a perfectly healthy fan, and neither reads like the chart in a textbook. Without a baseline, you are left guessing whether today's reading is normal for this machine — and guessing is exactly what the programme exists to replace.

At each point on the route, capture a time waveform and a spectrum and record everything alongside it: sample rate, frequency span, averaging count, mounting method, and the exact sensor position. That bundle is the machine's fingerprint. It is also why the measurement chain matters early — a PhonoVibe DAQ ships with a factory calibration certificate valid for one year, so the sensitivity behind every number is a stated, dated figure rather than an assumption, and a change in the trend is less easily blamed on the instrument. (Traceability in the metrological sense is a separate exercise — an unbroken calibration chain back to a national standard — and it is what the T-Calibro bench calibrator exists for.)

Trending only works if the measurement is repeatable, and the quickest way to ruin repeatability is mounting. The same machine, on the same day, reads differently from a spot a few centimetres away — a reading taken on a cooling fin is not comparable to one taken on the bearing housing, and neither tells you anything if next month's technician picks a third spot. Mark every point permanently: a paint mark at minimum, an adhesive pad for a dedicated location, and a quick-release magnetic mount for walk-around collection.

The rule to enforce is simple: a change in the trend must mean the machine changed, not the technician. Photograph each point during the baseline visit and put the photos in the route sheet. Low-noise cabling earns its keep here too — coaxial cable that flexes under vibration generates its own spurious signal, which is one more way to see a change that is not the machine.

A severity chart cannot know your machine. A fan that has read 4 mm/s since the day it was commissioned, and still does, is far less likely to be developing a fault than one that drifted from 1.5 to 3 over three months — even though the second machine looks better on paper. (The steady 4 mm/s machine may still be worth investigating on its own merits; what it is not is changing.) Absolute limits are a useful backstop; the early warning lives in the trend, which is why alarm limits are the last thing to set, not the first. You cannot set a sensible limit for a machine whose normal you have never measured.

After half a dozen route visits you know the machine's normal scatter, and alert and alarm bands can be set relative to it: alert means plan the work, alarm means act. In TVIB, band cursors let you put alarm bands around the frequencies that matter — bearing defect frequencies, gear mesh — so rising energy in those bands flags trouble while the overall level still looks respectable. The office routine from there is short: compare today's spectrum against the baseline and the last three visits.

A representative rollout, not a specific customer: a mid-size plant scores its machine list and picks twelve for route one. Month one is baselines — every point measured healthy, pads fixed and photographed, settings written down, with a four-channel DAQ covering motor and driven-end bearings at each stop. Months two and three are the route, twice a month, same order every time. By day ninety there are five or six trend points per machine and a first draft of alert bands. Nothing dramatic has been found, and that is the point: the plant now knows what normal looks like, which is the one thing it could not have bought.

Because here is the honest part: programmes fail on skill and follow-through, not on hardware. Someone has to read the spectra, and someone has to turn a rising trend into a work order. If the alert fires and nothing is scheduled, the programme is theatre — and expensive theatre reads worse than none. Budget for the person before the fourth instrument.

Building that person is a ladder. The free TIERA 101 primers at 101.tieraonline.in get a new technician reading a spectrum — they are free primers, not accredited ISO certification, and their hours do not count towards formal ISO 18436 training requirements. For a credentialed analyst there is the formal TCAT programme, Category I through IV, detailed on our services page, with proctored examinations at exams.tieraonline.in. In between, ToLearnVibe drills fault recognition on 100+ simulated scenarios covering the ISO 18436-2 Cat I and Cat II syllabus, and a TMFSS gives the team controlled, repeatable faults to practise on before the real machine presents one.

Everything route one needs fits in a carry bag: a two-channel PhonoVibe D for a small starting route (a four-channel Q when you want motor and driven end in one stop), an IEPE accelerometer with a magnetic mount and adhesive pads for the marked points, low-noise cabling, and TVIB — the analysis software is bundled with every PhonoVibe, so trending spectra against your baseline costs nothing extra.

You do not have to make the first calls alone, either. Talk our engineers through your machine list and we will help you shortlist route one and, once a few visits of real baseline data exist, set alert and alarm bands around the frequencies that matter for those machines — from your machines' normal, not a chart's.