3.2 Machinery Monitoring and Protection

Protection Versus Condition Monitoring

Measurement spec 1.D is machinery monitoring and protection. The PE exam wants the sensor that actually sees the failure mode, then a clear statement of whether the signal is a protection trip or a condition-monitoring advisory.

Industry practice is framed by API 670-style machinery protection systems: dedicated monitors, fast response, alarm and danger setpoints, and a trip path that must not depend on a historian or a vibration analyst being online. Condition monitoring—spectra, enveloping, oil debris, trend portals—diagnoses developing defects and schedules work. It is not the shutdown path.

Do not memorize or quote a specific API 670 edition as an NCEES supplied standard for the 2027 PE Control Systems exam. It is not on that supplied-standards list. Use the architecture ideas (relative versus absolute vibration, X–Y probes, Keyphasor, alarm versus danger, voted overspeed). Do not spend exam minutes recalling clause numbers.

Relative Shaft Vibration Versus Casing Vibration

Eddy-current proximity probes measure relative shaft vibration: the instantaneous gap between the probe tip, mounted on the bearing housing, and the shaft journal. A high-frequency oscillator in the probe driver produces a voltage proportional to gap. A common scale is about 200 mV/mil (approximately 7.87 V/mm), with a typical installed gap on the order of 1 mm (40 mils), vendor-specific. The measurement is shaft-to-probe, which is shaft-to-bearing if the probe is rigidly mounted on the bearing housing.

On hydrodynamic sleeve (fluid-film) bearings, the shaft can orbit within the bearing clearance. Unbalance, misalignment, oil whirl, oil whip, and rotor rub all show up as journal motion inside that clearance. The casing of a well-seated compressor or turbine may barely move. If you only bolt an accelerometer to the case, you are measuring casing absolute vibration. That is the right primary sensor for rolling-element bearings and for gear mesh, where defect pulses are high-frequency and the housing does move. It is the wrong primary sensor for shaft orbits on a sleeve-bearing machine.

Velocity transducers (moving-coil seismometers) and integrated accelerometers remain useful for casing velocity against general machinery guidelines. They still do not produce a shaft centerline orbit.

For orbits you need two proximity probes in the same radial plane, 90° apart (X and Y), plus a Keyphasor—a once-per-revolution phase reference, usually a proximity probe or magnetic pickup seeing a keyway or a painted mark. The Keyphasor lets you freeze the orbit, compute phase lag, and distinguish synchronous (1×) unbalance from subsynchronous whirl. Many API-style installations place the radial probes 45° off vertical so they miss the horizontal split line; they are still an orthogonal pair.

Probe faults matter. A failed oscillator, a shorted extension cable, or a gap driven outside the linear range should be a channel fault, not a silent “zero vibration” reading. Protection logic often inhibits a trip from a faulted channel and relies on the redundant probe or on voting, which is why X and Y are not optional decoration.

Other Protection Measurements

Bearing temperature uses resistance temperature detectors (RTDs) or thermocouples embedded in the babbitt, in the bearing drain, or in the housing. Temperature is a backup to vibration, not a substitute. A thin oil film can wipe a bearing with only a modest temperature rise if the element sits in the wrong pocket.

Thrust (axial position) uses proximity probes looking at a thrust collar or shaft end. You are measuring position, not force. A step change is often a failed thrust shoe or a process load reversal (compressor surge, pump seizure). Dual or triple thrust probes with voting appear on critical machines because a single probe gap error looks like a thrust failure.

Lube-oil pressure and filter differential pressure are ordinary process pressure transmitters. Low header pressure is a classic trip. Do not use a vibration probe to infer lubrication, and do not use oil pressure as proof that the rotor is centered.

Speed and overspeed use magnetic pickups, proximity probes, or dedicated overspeed wheels. Overspeed protection is typically voted (for example 2oo3) and kept independent of the general vibration rack so a monitor reboot cannot disable the trip.

Rod drop on lubricated reciprocating compressors uses a proximity probe looking at the piston rod. As rider bands (wear bands) lose thickness, the rod sags and the gap trend shifts. The intent is to alarm before the piston or packing rides the liner. Rod drop is not a centrifugal-machine measurement and it is not a substitute for frame vibration on a reciprocating unit.

Alarm, Danger, Voting, and Shutdown

Protection channels almost always have two setpoints. Alert/alarm is advisory: notify operations, perhaps start a controlled load reduction or bring a spare online. Danger is the shutdown setpoint: trip the driver, close the suction, or take the defined ESD action. On a high vibration or high temperature trip, danger is higher in engineering units than alarm. On a low oil-pressure trip, danger is lower than alarm. Mixing those directions is an easy exam miss.

Voting examples:

• 1oo2 is sensitive (either channel trips) and more vulnerable to a single probe spike unless probe-fault inhibit is engineered.
• 2oo2 requires both, which cuts spurious trips and can fail dangerously if one probe is stuck “healthy.”
• 2oo3 is the usual compromise for overspeed and for many high-integrity trips.

A PE stem that says “the spectrum analyzer shows increasing 1× and sidebands” is condition monitoring. A stem that says “danger relay energized, stop valve closed in 200 ms” is protection. Connecting the protection rack’s 4–20 mA monitor output into a historian does not turn the historian into the protection system. Conversely, a portable analyzer cannot be credited as the API 670-style trip path.

Shutdown versus advisory also drives response time. Protection monitors are specified in tens of milliseconds to a few shaft revolutions. Oil-debris or weekly spectral routes are hours-to-weeks tools. If the stem asks what prevents a catastrophic rub, the answer is the danger trip, not the monthly report.

Worked Example: Why the Proximity Probe Wins on a Sleeve-Bearing Compressor

A 20 MW process-gas centrifugal compressor has tilting-pad sleeve bearings. During a high 1× event, the rotating-equipment engineer wants shaft orbits to tell unbalance from a developing rub. A casing accelerometer at the bearing cap measures housing motion. The pedestal is stiff, the rotor is heavy, and most of the dynamic displacement is shaft-in-bearing, not case-in-space. The accelerometer time waveform does not give X–Y shaft centerline, so you cannot see a flattened orbit (rub) or a forward-whirl loop. Two proximity probes at 45° off vertical plus a Keyphasor give gap in mils, phase, and the orbit.

Use casing accelerometers in addition if an accessory gearbox has rolling-element bearings or if you need high-frequency gear-mesh condition monitoring. Do not use them instead of shaft probes on the compressor journals. If the same train has a rolling-element-bearing oil pump, the pump’s casing accelerometer and the compressor’s proximity probes are answering different questions on the same lube console.

Exam Traps

• Calling a casing accelerometer “shaft vibration” on a sleeve-bearing journal.
• Treating a portable analyzer or a DCS trend as the machinery protection system.
• Confusing alarm (advisory) with danger (trip), or reversing high versus low setpoint order.
• Claiming a named API 670 edition is in the PE supplied-standards package.
• Using rod-drop probes on a centrifugal machine, or treating thrust probes as shaft horsepower meters.