Industrial ultrasound creates maintenance value when an acoustic indication becomes a traceable decision: locate the source, confirm it, classify the condition, release the right work, and measure the outcome after intervention. The instrument heterodynes high-frequency energy into an audible range and presents a comparable reading. One device can support leak surveys, contact checks on mechanical components, and electrical screening, yet those applications require different accessories, operating states, evidence, and confirmation rules.
A compressed-air campaign provides a rigorous application because location, estimation, repair, and verification can be separated. Displayed intensity is not automatically a flow rate, an energy figure, or a dollar value. Pressure, attributed flow, loaded hours, utilization, compressor performance, and electricity cost must remain visible. Engineering owns the conversion from acoustic evidence to a loss estimate and states what the estimate cannot prove.
Separate the phenomena before building routes
Turbulence from pressurized gas
A leak scan looks for contrast between a suspected point and background sound. Sensor direction, distance, system pressure, geometry, and neighboring sources change the indication. The inspector approaches progressively, uses a focusing accessory when appropriate, and identifies a repairable source at a fitting, hose, drain, flange, or valve. Tagging an entire noisy zone creates vague work orders and prevents meaningful post-repair testing.
Contact readings on bearings
A bearing route needs a marked contact point, repeatable probe force, known speed, load, and lubricant state. An absolute reading without the asset's own baseline carries limited authority. A change may accompany lubrication deficiency, contamination, race damage, fit, or process variation. If ultrasound supports controlled lubrication, the procedure limits quantity and observes response. Adding grease until sound falls can overfill the housing, increase temperature, and damage seals.
Electrical screening
An airborne signature can support a hypothesis of corona, tracking, or arcing, but cannot independently classify mechanism, severity, or repair. Voltage class, enclosure, distance, load, and access belong in the record. The task follows the site's risk assessment, approach boundaries, authorization, and electrical safety procedure. Opening energized equipment merely to obtain a cleaner signal is outside the authority of an ultrasound route.
Design a compressed-air route for repeatability
Route design starts with the system map and a decision question. The planner separates generation, treatment, main ring, and point-of-use branches, records normal pressure ranges, and identifies production sources that mask acquisition. A logical sequence reduces duplicate coverage and lets the team aggregate confirmed loss by area, component, and owner. Inaccessible points are recorded as coverage gaps rather than silently counted as healthy.
Every finding receives a physical tag and a link to the functional location or equipment in SAP PM. Evidence includes time, header pressure, line state, approximate distance, accessory, relevant device setting, background reading, point reading, contextual image, and exact location. When an instrument or engineering table estimates flow, the method, distance, pressure band, and uncertainty are retained with the result.
| AC-4 route stage | Required record | Gate to continue |
|---|---|---|
| Screen | Area, pressure, background, covered segments | Persistent contrast as direction and distance change |
| Pinpoint | Component, tag, image, sensor position | Source separated from adjacent equipment |
| Estimate | Didactic flow, assumptions, pressure, utilization | Reproducible calculation without direct dB-to-cost conversion |
| Treat | Notification, priority, access, material, tightness test | Repairable scope and safe work condition |
| Verify | Comparable pressure, repeat scan, found condition | Source removed or residual loss retained |
Completed case: the AC-4 network campaign
Didactic data: a packaging plant operates network AC-4 at 6.8 bar at the main header for 7,200 annual hours. The survey covered 412 connections across three production zones while pressure remained between 6.6 and 6.9 bar. Inspectors tagged 47 initial indications. Functional review excluded two engineered bleeds before leak classification, leaving 45 analyzable indications. Seven remained masked by continuous process blow-off and 38 were confirmed leaks. The reconciliation is 47 - 2 = 45; 45 - 7 = 38. People, assets, rates, and outcomes are educational data, not a customer result, benchmark, universal tariff, or PM Run promise.
| Class | Points | Estimated flow | Condition and disposition |
|---|---|---|---|
| A, high loss | 8 | 1.12 m³/min | Accessible flanges and hoses, first repair window |
| B, medium loss | 15 | 0.83 m³/min | Fittings and valves grouped by area and material |
| C, low loss | 15 | 0.45 m³/min | Drains and couplings in local work packages |
| Unconfirmed | 7 | No estimate | Repeat without blow-off or use another technique |
Confirmed loss totals 2.40 m³/min. The prioritized group contains all eight Class A points, totaling 1.12 m³/min, plus the four highest-flow Class B points, totaling 0.34 m³/min. These 12 points therefore total approximately 1.46 m³/min; 1.46 ÷ 2.40 × 100 = 60.8 percent after rounding. The other 11 Class B points total 0.49 m³/min, preserving the class total of 0.83 m³/min. This concentration determines sequence only, while all 38 confirmed leaks remain visible in the plan. The campaign owner refused to assign flow to the seven uncertain indications. Counting an unlocated source would overstate benefit, consume labor without a repair target, and make the verification denominator unreliable.
Show every assumption in the economic estimate
The example uses a didactic specific energy of 0.115 kWh per produced m³, a 75 percent average leak utilization factor, and an electricity cost of USD 0.14/kWh. Attributed annual energy is 2.40 m³/min × 60 min/h × 7,200 h/year × 0.75 × 0.115 kWh/m³ = 89,424 kWh/year. Indicative exposure is 89,424 × USD 0.14 = USD 12,519.36 per year.
The estimate does not replace system performance measurement. Pressure varies, compressors modulate, and reducing demand may not immediately unload or stop a machine. Utilities engineering checks whether the campaign changes power, loaded time, pressure stability, or avoided capacity. Until that relationship is demonstrated, the amount is labeled potential avoidable consumption instead of booked savings.
The approved scope included 31 confirmed points with available materials and safe access. Seven confirmed leaks stayed in backlog: four required a packaging-line outage and three needed special hose assemblies. The seven masked indications became a separate off-production survey. Management now had three controllable populations: released repair, blocked repair, and unresolved evidence.
Challenge the estimate before funding the package
The senior review varied specific energy from 0.100 to 0.130 kWh/m³ and utilization from 60 to 85 percent. That sensitivity produced an annual energy range rather than a false single-point forecast. Ranking did not change because the twelve largest sources remained dominant under every scenario. The financial figure supported packaging the work; pressure stability and production capacity remained the operational reasons for execution.
The functional review behind the population reconciliation checked whether a source served an intentional purge, cooling function, or safety requirement. The two engineered bleeds are the same two exclusions already deducted from 47, not additional findings outside the equation. Removing an intended flow without design authority could compromise equipment. This gate belongs before notification approval, especially where compressed air drives instrumentation or protects enclosures.
Convert findings into SAP PM work without overstating cause
One area-level notification grouped points that shared symptom, owner, and window; a leak on a critical valve received its own record. Description retained physical tag, component, pressure, estimated flow, method, uncertainty, and image. Damage coding followed plant taxonomy. The record did not claim gasket failure before disassembly because a loose joint, damaged seat, and cracked connector require different materials and learning.
After triage, the order contained operations to isolate, depressurize, repair, restore, leak-test, and remove tags. Work center, duration, kits, permits, and process release were checked against the schedule. The maintenance work order acceptance criterion required no localized indication and no adverse effect after repressurization. Financial potential was never a technical completion criterion.
Field feedback recorded 19 connections tightened under procedure, six seals replaced, four hoses changed, and two drains repaired. Three tags led to components different from planned scope, so the crew returned them for planning before consuming substitutes. Labor, components, found condition, action, and test entered confirmation, preserving SAP PM in industrial maintenance as the system of record.
Verify removal, stability, and recurrence
The first check occurred after stable repressurization at 6.7 bar. All 31 treated points showed no localized indication above background from the recorded positions. Estimated residual loss was 0.42 m³/min across the seven confirmed leaks still blocked. The team did not compare 0.42 directly with 2.40 without noting the 0.1 bar pressure difference and changed population.
The control plan used three exposures. Within 24 hours, operations confirmed pressure and functional stability. At seven days, inspection repeated treated points under comparable load. At 30 days, the complete route looked for recurrence, missing labels, and new sources. Final effectiveness required 31 stable repairs, dates assigned to seven blocked leaks, and a disposition for every masked indication.
Three indicators remain separate: confirmation rate, confirmed flow treated, and 30-day recurrence. Confirmation rate tests survey quality. Treated flow follows the scope that received correction. Recurrence tests durability. A single success percentage would hide temporary repairs, uncertain findings, or a campaign that generated many tags but few confirmed sources.
Audit route quality and backlog health
A monthly sample rechecks physical tags against their technical objects, tests whether photographs still identify the component, and verifies that pressure was present in estimates. It also reviews canceled work: duplicate notification, designed bleed, inaccessible source, changed production need, or poor location. Those reasons improve future route design. Quiet cancellation without a reason merely removes work from view.
Backlog age is split by readiness. A confirmed leak awaiting material is managed differently from an indication awaiting confirmation. Detection-to-decision time shows whether inspection produces action, while decision-to-repair time shows execution constraint. Combining both would blame the survey for a process-release delay or hide a slow technical review behind fast wrench time.
Know when a neighboring technique owns the next question
Predictive maintenance can combine technologies, but acoustic evidence retains its boundary. Thermography evaluates thermal distribution, vibration analysis characterizes dynamic components, oil analysis examines lubricant and debris, and a pressure-decay test can prove a defined boundary. Ultrasound location hands off a precise question instead of inheriting diagnostic authority the signal cannot support.
Comparability is lost when instrument, firmware, accessory, distance, pressure, or setup changes without a record. Strong background sources may mask leaks. Absorbing materials, barriers, and enclosed geometry alter propagation. Low differential pressure may produce insufficient turbulence. Very slow bearings and variable loads can make a single contact value misleading. These limitations create exclusions, new baselines, or confirmatory tests.
PM Run supports the chain around the instrument
PM Run is a planning, mobility, and execution layer over SAP PM. Planned operations can reach field teams, while notes, measurements, permitted images, PDFs, and confirmations return through the adopted flow. PM Run does not supply ultrasound collectors, permanent sensors, IoT, automatic flow estimation, acoustic diagnosis, or its own predictive engine. Qualified inspection, engineering, operations, and safety personnel retain technical authority.
Governance reviews route coverage, confirmed-finding ratio, detection-to-decision time, access-blocked points, value supported by validated assumptions, and recurrence by component. A long backlog is still weak if findings lack a technical object, work readiness, and field return. Explore maintenance software integrated with SAP PM to keep that operating chain visible.
Preserve location history after tags are removed
The campaign closes with a point register that survives physical tag removal. Component, position, contextual image, pressure, loss class, and disposition remain linked to the finding. The following route can then distinguish recurrence at the same joint, a new source on the same machine, and sound reflected near an old label. Overwriting the original evidence would erase the basis for durability review.
Route interval follows pressure, consequence, modification history, and the rate at which new leaks appear. Recently changed branches may receive an early check, while stable segments remain on a defensible cycle. This prevents repeated low-value scanning in one area and prolonged blindness in connections exposed to vibration or frequent handling.
Primary and neutral references
- U.S. Department of Energy, Improving Compressed Air System Performance, a neutral reference for compressed-air losses and operation.
- NASA Reliability-Centered Maintenance Guide, a public foundation for maintenance decisions tied to function and consequence.
- SAP Help Portal, Maintenance Notification, official documentation for the condition record.
