Predictive Gearbox Maintenance in Metallurgy: Spectral Analysis | Bently Nevada — KEG TRK

A gearbox on a rolling mill, roller-table drive, or main metallurgical-shop conveyor operates under shock and variable loads. Oil temperature, dust ingress, and misalignment after an emergency repair all accelerate tooth wear. When a gearbox "lets go," the entire line stops; the cost of replacing a gear set plus downtime far exceeds the cost of a predictive maintenance program.

Vibration spectral analysis is the primary method for early detection of gear defects. Bently Nevada systems — from route-based SCOUT to continuous 3500/42M with history stored in System 1 — give the engineer tools for trended spectral monitoring, not one-off readings that simply land "within limits."

The physics of gear-mesh vibration

For a gear with Z teeth rotating at frequency f, the shaft rotational frequency is 1x = f. The Gear Mesh Frequency (GMF) is:

GMF = Z × RPM / 60 (in Hz)

On a healthy gearbox, GMF appears in the spectrum as a relatively narrow peak with sidebands at ±1x, ±2x of the driving and driven shaft rotational frequencies. As a defect develops:

• tooth pitting and wear — rising GMF amplitude and sidebands spaced at 1x of the damaged gear;
• local tooth chip — GMF modulation, time-domain impulses;
• misalignment — rising 2x GMF and axial components;
• backlash and cracks — subharmonics (0.5x GMF), amplitude instability.

A detailed breakdown of failure mechanisms is in Gearbox Diagnostics; this article focuses on Bently Nevada's application in metallurgy.

Measurement points on a gearbox

1. Input-shaft bearing housing — H and V, Velomitor or accelerometer.
2. Output-shaft bearing housing — same; comparing input vs. output helps localize which stage has the defect.
3. Optional — motor-to-gearbox flange — for diagnosing misalignment of the coupled pair.

On gearboxes with output shaft speeds below 300 rpm, the low-frequency GMF can fall into the "blind spot" of a standard accelerometer — here a velocity sensor with good low-frequency sensitivity, or envelope analysis in System 1, is useful.

The predictive maintenance workflow

Stage 1 — Baseline

After startup or overhaul, record the "reference" spectrum: GMF, 1x, 2x, noise floor. Store it in System 1 tied to the gearbox serial number and the oil-change date.

Stage 2 — Periodic or continuous monitoring

• Route — SCOUT every 2–4 weeks across a fleet of 50+ gearboxes;
• Critical units — 3500/42M with an alarm on overall-level rise, plus a monthly spectrum pulled from System 1 on demand.

Stage 3 — GMF trend analysis

What matters is not the absolute level but the slope over 8–12 weeks. A GMF rise of 6–10 dB at constant load is a trigger for an internal inspection through the sight glass or an oil iron-content analysis.

Stage 4 — Confirmation and repair

A widening sideband spacing around GMF, matching the rotational speed of the damaged gear, is the classic confirmation of a local tooth defect. A shutdown is then scheduled for the next maintenance window rather than as an unplanned stop.

Examples from metallurgical sites

Roller-table gearbox on a rolling mill

Variable torque as the workpiece passes through modulates the load. The baseline is recorded both at idle and under load — different profiles. The alarm is tuned to a rise in the broadband level within the GMF band, not to an absolute ISO threshold.

Shear drive

Shock loading masks GMF at the moment of cutting. Analysis is done in the intervals between cuts, or the spectrum is averaged using Keyphasor synchronization to the shear cycle — a task for an experienced analyst in System 1.

Charge belt conveyor gearbox

Long-term partial-load operation, abrasive dust in the oil. GMF trending plus particle analysis of the oil is a combined approach. Vibration catches mechanical wear earlier than a visual inspection of the teeth through the access hatch.

Bently Nevada tools in the chain

Task	Tool
Mass fleet rounds	SCOUT Portable
Critical gearbox with no backup	3500/42M + Velomitor
Spectrum archive, comparison	System 1
Mechanic training	FFT signatures

Limitations of spectral analysis

• At the micro-pitting stage GMF may not yet rise — ultrasound and oil analysis help (see Gearbox Diagnostics).
• Misidentifying GMF (wrong tooth count) produces false conclusions — always cross-check against the gearbox nameplate.
• Variable speed (VFD) complicates the spectrum — RPM synchronization is required.

Economics

Replacing a gear set in a rolling-mill drive gearbox runs from hundreds of thousands of tenge to millions once removal is included. A spectral monitoring program across 20 critical shop gearboxes costs a fraction of that. One prevented failure pays for 3–5 years of the program.

Calculating gear-mesh frequency: an example

Roller-table gearbox: drive pinion Z1 = 22 teeth at 1480 rpm (motor input shaft); driven gear Z2 = 85 teeth, output 382 rpm.

GMF at the first stage: 22 × 1480 / 60 = 542.7 Hz GMF at the output (via the gear ratio): 85 × 382 / 60 ≈ 541.3 Hz (same stage, consistent figure).

The analyst compares the GMF amplitude against the baseline from 90 days earlier. An 8 dB rise at unchanged load is a trigger to open the sight glass or run an Fe oil analysis.

Envelope analysis for early-stage pitting

At the micro-pitting stage GMF may not yet rise. Methods include:

• Envelope (acceleration) — demodulation in the 2–10 kHz band around the housing resonance;
• comparing the envelope before/after at the same point and RPM;
• trending the "envelope RMS" parameter in System 1.

SCOUT and System 1 both support envelope calculation; for critical gearboxes, envelope is added to the list of trended parameters alongside GMF.

Oil and vibration: a combined approach

On metallurgical gearboxes it is worth combining:

• online/route vibration (Bently);
• wear-particle oil analysis (lab or online particle counter);
• sight-glass thermography (quarterly).

A correlation of "rising GMF + rising Fe in oil" is high-confidence evidence of a tooth defect. Oil analysis alone lags in time; vibration alone can trigger a false alarm when the lubricant grade changes.

A typical inspection schedule for metallurgical gearboxes

Criticality	Tool	Interval	Action on alarm
A — no backup	3500/42M	24/7	Shutdown within a 72-hour window
B	SCOUT + trend	2 weeks	Inspection, oil analysis
C	SCOUT	1 month	Scheduled repair

Common mistakes in spectral gearbox diagnostics

1. Wrong tooth count in the GMF calculation — always verify against the nameplate and datasheet.
2. Measuring at variable VFD speed without synchronization — a "smeared" spectrum.
3. Comparing spectra taken under different loads — GMF amplitude depends on torque.
4. Ignoring sidebands — they are precisely what localizes the damaged gear.

Integrating with the metallurgical shop's production schedule

Spectral gearbox monitoring only pays off when paired with shutdown windows. The reliability engineer hands the planning department a forecast: "roller-table gearbox No. 4 — bearing replacement or tooth inspection within 6 weeks." Planning then folds the shutdown into the next scheduled mill stop for liner change or furnace maintenance — no separate unplanned window needed. Combining auxiliary-drive vibration monitoring and gearbox spectral analysis on a single System 1 platform simplifies prioritization: the unit with the steepest GMF or BPFO trend gets repaired first — not "whichever sounds loudest during the rounds."

Training and certification of analysts at a metallurgical plant

A minimum training program for a gearbox spectral analyst is 24 academic hours: gear-train theory, GMF calculation, hands-on practice in System 1 and SCOUT using real spectra from the plant. KEG TRK delivers this training on-site, using the plant's own trend archive — which builds shop-floor trust in the program. We recommend appointing a single gearbox program owner (one reliability engineer) reporting quarterly to the chief mechanic: list of at-risk units, repairs completed, downtime avoided.

Spectrum archives and the legal weight of the data

In a dispute with a contractor over gearbox repair quality (gear-set replacement, alignment), the "before" and "after" GMF spectra in System 1 are objective evidence. Keep the archive for at least 5 years for critical rolling-mill and converter-shop gearboxes. Exporting the spectrum to PDF with date, RPM, and the engineer's signature is standard functionality in both System 1 and SCOUT; KEG TRK configures automatic backup of the station's database server.

The same GMF-analysis methods apply to gearboxes on conveyors and crushers at a mining and processing plant; the difference is lower shock loading and more stable RPM compared to a rolling-mill shear. A single methodology shared across the metallurgical plant and the processing plant simplifies rotating reliability engineers between sites within a holding company.

Gearboxes on rolling-mill auxiliary drives are a priority target for spectral monitoring: frequent starts and variable torque accelerate tooth wear. We recommend a 2-week route interval for Class A units and continuous 3500/42M monitoring on coilers and shears whose gearbox has no backup.

Calculation tools and templates in System 1

In System 1, each gearbox gets a record card: manufacturer, gear ratios, tooth counts per stage, bearing type, date of last oil change. Automatic GMF and BPFO markers in the spectrum reduce "wrong peak" errors. The "gearbox — monthly review" report template includes GMF trends, overall level, and a spectrum screenshot — ready-made material for a reliability meeting without manual assembly in Excel. KEG TRK supplies record-card templates configured for the typical Flender, SEW, and domestic gearboxes found at CIS metallurgical plants.

The connection to the gearbox diagnostics article is useful for combined programs: ultrasound and oil analysis for early pitting, Bently spectral GMF for advanced tooth wear. On a single gearbox the two approaches don't compete — they cover different stages of the P-F curve.

Reviewing GMF thresholds every six months is good practice: a load change after a mill upgrade shifts the "normal" mesh amplitude, and old alarm thresholds stop making sense.

Summary

Predictive gearbox maintenance in metallurgy relies on trended spectral analysis of mesh frequency and sidebands. Bently Nevada's SCOUT, 3500/42M, and System 1 form a proven chain from measurement to work order. KEG TRK helps build a program around your gearbox fleet.

For a consultation, see KEG TRK contacts.