Predictive Maintenance for Plastic Machinery: Vibration Monitoring and Fault Early Warning

Predictive maintenance for plastic machinery has moved from a specialist discipline practised by a handful of large compounders into a baseline expectation for any plant running continuous extrusion, high-speed blow molding or multi-shift injection molding. The economics are simple and unforgiving: a compounding extruder that stops without warning does not merely lose the hours required to change a bearing, it loses the purge material, the thermal state of the barrel, the stabilised process window and, in many cases, the customer confidence attached to an on-time delivery. Vibration monitoring is the single most information-dense technique available for catching those failures while they are still weeks or months away from causing a stoppage, and when it is combined with lubricant analysis, thermal monitoring and process-side data it forms the backbone of a working condition-based maintenance programme.

This guide is written for maintenance engineers, plant managers and reliability specialists working with plastic processing equipment. It covers the physics of vibration measurement, the severity classification framework defined in ISO 10816 and its successor ISO 20816, practical sensor selection and mounting, the time-domain and frequency-domain indicators that actually discriminate between fault types, the calculation of rolling-element bearing defect frequencies, gear mesh diagnostics in extruder reduction gearboxes, machine-specific measurement point layouts, hydraulic condition monitoring on injection molding machines, alarm threshold design, data collection strategy, and integration with plant MES and SCADA systems. Wanplas, whose network of specialised factories builds compounding extruders, extrusion and injection blow molding machines, PET bottle blowing systems, pipe and profile lines, film and sheet lines and plastic recycling equipment, has assembled these practices from commissioning and service work across more than 100 exported regions, and this article distils that field experience into an implementable framework for 2026 and beyond.

Why Predictive Maintenance Now Defines Plastics Plant Competitiveness

Predictive maintenance is the practice of inferring the remaining useful life of a machine component from measured physical evidence of its degradation, then scheduling intervention at the point that minimises total cost. It differs fundamentally from time-based preventive maintenance, which replaces parts on a calendar regardless of their actual condition, and from reactive maintenance, which waits for functional failure. In plastics processing the case for predictive methods is unusually strong because the equipment population is dominated by rotating machinery with well-characterised failure modes and because unplanned stoppages carry disproportionate secondary costs.

Consider the cost structure of a single unplanned stop on a twin-screw compounding line. The direct repair cost of a failed gearbox output bearing is modest. The consequential costs are not: the melt in the barrel degrades and must be purged, the die and screen changer require cleaning, the downstream pelletiser and classifier must be emptied, the batch in progress is scrapped or downgraded, and restart typically consumes several hours of stabilisation before the product returns to specification. On a masterbatch or engineering-compound line running a high-value formulation, the scrap and restart burden regularly exceeds the repair burden by an order of magnitude. The same logic applies to a PET stretch blow molding line, where preform heating ovens must re-stabilise thermally, and to a recycling washing line, where partially processed material sits wet in friction washers and dewatering equipment.

There is also a quality dimension that pure availability metrics miss. Elevated vibration in an extruder gearbox output stage translates into screw speed modulation, which translates into melt pressure ripple at the die, which shows up as thickness variation in a cast film or dimensional variation in a pipe wall. Long before a bearing fails outright it degrades product uniformity. Plants that monitor vibration frequently discover that a persistent gauge-variation complaint traces back to a mechanical defect rather than a formulation or temperature issue. Predictive maintenance therefore protects yield as well as uptime, and this dual benefit is a significant part of why industry surveys of manufacturers implementing structured condition-monitoring programmes commonly report unplanned downtime reductions in the range of 20 to 50 percent, spare-parts inventory reductions of 10 to 30 percent, and overall equipment effectiveness gains of 5 to 15 percent once the programme reaches maturity.

Vibration Fundamentals: Displacement, Velocity and Acceleration

Every vibration measurement is one of three physically related quantities, and choosing the wrong one is the most common reason that a monitoring programme fails to detect a developing fault. Displacement, velocity and acceleration are successive time derivatives of one another, which means each emphasises a different part of the frequency range. Displacement emphasises low frequencies, acceleration emphasises high frequencies, and velocity sits in between and is therefore the most balanced single indicator of general machine health.

Displacement

Displacement is measured in micrometres, conventionally as peak-to-peak, and describes the physical distance the measured surface travels. It is the correct quantity below roughly 10 Hz, where velocity and acceleration values become vanishingly small even when the physical motion is large. In plastics machinery, displacement measurement is used mainly on journal-bearing machines such as large screw compressors serving central compressed-air systems, and on slow-turning shafts. Eddy-current proximity probes measure shaft displacement relative to the bearing housing without contact, typically with a sensitivity of 7.87 mV per micrometre and a nominal installed gap around 1.27 mm. They are indispensable for detecting oil whirl and oil whip in sleeve bearings but are rarely fitted to rolling-element machines, which represent the overwhelming majority of plastics equipment.

Velocity

Velocity, expressed in mm/s RMS, is the workhorse quantity for machine condition assessment because vibration velocity is approximately proportional to the fatigue-inducing energy in the structure across the 10 Hz to 1000 Hz band. This is precisely why the ISO severity standards are written in velocity terms. A velocity measurement gives roughly equal weight to a low-frequency imbalance component and a mid-frequency misalignment component, so a single broadband number provides a defensible summary of overall mechanical health. Velocity can be measured with an electrodynamic moving-coil transducer, typically 20 mV per mm/s with a usable band of 10 Hz to 1000 Hz, but in modern practice it is almost always obtained by electronically integrating an accelerometer signal, which avoids the moving parts and orientation sensitivity of the coil sensor.

Acceleration

Acceleration, expressed in g or m/s squared, dominates above roughly 1 kHz and is the only practical quantity for detecting the very short, very low-energy impacts produced by early rolling-element bearing damage and by gear tooth defects. A piezoelectric IEPE accelerometer with 100 mV/g sensitivity and a usable range from 0.5 Hz to 10 kHz is the standard general-purpose choice. High-frequency variants extending to 20 kHz or beyond are used for ultrasonic and acoustic-emission techniques that detect lubrication distress before any measurable geometric damage exists. The trade-off is that acceleration spectra are visually dominated by high-frequency content and are poor at revealing low-frequency imbalance, which is why a complete programme records acceleration for bearing analysis and derives velocity for severity assessment from the same signal.

Selecting the Right Quantity

A practical rule for plastic machinery is to use velocity as the primary trended severity indicator on every point, to record and analyse acceleration whenever rolling-element bearings or gear meshes are within the measurement path, and to reserve displacement for the small number of journal-bearing assets in the plant. On very slow shafts, such as the output stage of a high-reduction extruder gearbox turning a screw at 300 rpm, the shaft frequency is only 5 Hz and standard accelerometers with a 0.5 Hz low corner must be selected deliberately, because a general-purpose sensor rolling off below 2 Hz will simply not see the fundamental.

ISO 10816 and ISO 20816: Vibration Severity Zones Explained

ISO 10816 and its successor series ISO 20816 provide the internationally recognised framework for judging whether a measured vibration level is acceptable. The standards divide measured broadband velocity into four evaluation zones, labelled A through D, and give numerical boundaries that depend on machine power, shaft height and whether the machine support is classified as rigid or flexible. ISO 20816-1, published in 2016, superseded ISO 10816-1, and ISO 20816-3 now covers industrial machines from 15 kW to 50 MW with shaft heights of 160 mm and above. Many plants still refer to the older ISO 10816-3 numbers, and because the zone boundary values were largely carried across, the two are used interchangeably in day-to-day practice.

Zone A corresponds to newly commissioned machines. Zone B is acceptable for unrestricted long-term operation. Zone C is unsatisfactory for continuous long-term operation and the machine should be run only until a suitable repair opportunity arises. Zone D is of sufficient severity to cause damage to the machine. The critical point for programme design is that these zones map directly onto a four-level alarm scheme, which is the structure most condition-monitoring software expects.

Zone Boundary Values

The table below consolidates the boundary values most relevant to plastic machinery. Group 2 covers medium-size machines from 15 kW to 300 kW, which includes the large majority of extruder main drives, blow molding hydraulic power units and recycling line drives. Group 1 covers large machines from 300 kW to 50 MW, which in plastics plants generally means very large compounding extruder drives and central utility equipment. Support is classified as rigid when the first natural frequency of the machine-support system is above the main excitation frequency, and flexible when it is below.

Two practical cautions apply. First, ISO values are broadband velocity limits for the specified frequency band and must not be compared against a narrowband peak or against an acceleration reading. Second, the standards explicitly state that a change in level is often more significant than the absolute level. A machine that has run for two years at 1.6 mm/s RMS and then moves to 2.4 mm/s RMS is still nominally in Zone B, but a 50 percent step change demands investigation. Wanplas engineers who commission compounding and extrusion lines routinely record a baseline vibration signature during factory acceptance testing precisely so that this comparison is available to the customer from day one. On pipe and profile extrusion lines from the Faygo factory, for example, the mandatory 72-hour continuous operation test before shipment provides a stabilised baseline under realistic thermal and load conditions rather than a cold no-load reading.

Sensor Selection and Mounting for Plastic Machinery

Sensor selection determines the upper limit of what a monitoring programme can detect, and mounting method determines whether that limit is actually achieved. The most sophisticated analysis algorithm cannot recover information that was filtered out by a magnet mount resonating at 2 kHz. In plastics plants the additional complication is heat: barrel zones, die heads and heating ovens create surface and ambient temperatures that exceed the rating of general-purpose sensors, so thermal specification must be treated as a first-class selection criterion.

Sensor Type Selection Matrix

Mounting Method and Achievable Bandwidth

Mounting stiffness sets the mounted resonance frequency, and usable bandwidth is conventionally taken as one third of that resonance. Stud mounting into a spot-faced, tapped hole gives the highest stiffness and supports analysis up to 10 to 15 kHz. Adhesive mounting with a cyanoacrylate or epoxy pad reaches roughly 8 kHz. A rare-earth magnet on a flat machined surface is limited to about 2 to 3 kHz, and a two-pole magnet on a curved surface is worse still. A hand-held probe tip should never be trusted above 1 kHz and is unsuitable for any bearing envelope work. Because bearing defect detection requires clean data well above 3 kHz, permanent stud-mounted pads at every route point are the single highest-value investment a plastics plant can make when starting a programme.

Orientation conventions matter for repeatability. Each measurement location should be recorded in three axes where access allows: horizontal, vertical and axial, defined relative to the shaft. Horizontal readings on a typical machine are higher than vertical because horizontal support stiffness is lower. Axial readings are normally the lowest, so an axial level approaching or exceeding the radial levels is itself a diagnostic indicator pointing toward misalignment, a bent shaft, or thrust bearing distress. Measurement points should be marked physically on the machine and given permanent identifiers in the software database so that trend continuity is preserved across different technicians.

Time-Domain Indicators: RMS, Peak, Crest Factor and Kurtosis

Time-domain indicators are scalar values computed directly from the sampled waveform, and they form the first analysis layer of any predictive maintenance system because they compress a large waveform into a handful of trendable numbers. The four that matter most are RMS, peak, crest factor and kurtosis, and each responds to a different aspect of machine condition. Used together they distinguish a smoothly worsening imbalance from a sharp, impulsive bearing defect long before spectral analysis is required.

RMS is the root mean square of the waveform and represents the energy content of the vibration. Velocity RMS computed over 10 Hz to 1000 Hz is the quantity compared against the ISO zone boundaries, and it is the indicator that best reflects overall severity. Its weakness is that a small number of very large, very short impacts contribute little to RMS, so early bearing damage barely moves the RMS needle.

Peak, and its companion peak-to-peak, captures the largest excursion in the record. Because early bearing and gear defects express themselves as isolated impacts, acceleration peak rises much earlier than acceleration RMS. Peak alone is unreliable because a single electrical spike or a technician bumping the sensor produces a false alarm, so peak is best used as a numerator rather than as a standalone alarm.

Crest factor is the ratio of peak to RMS. A pure sine wave has a crest factor of 1.414; a healthy machine with a mix of harmonic and random content typically sits between 3 and 4. As isolated impacts begin, peak rises while RMS stays flat, so crest factor climbs into the 5 to 8 range. This is the classic early-warning window. The important subtlety is that crest factor is non-monotonic: as damage spreads and the impacts become continuous, RMS rises rapidly and crest factor falls back toward 3 to 4 even though the bearing is now severely damaged. A falling crest factor combined with a rising RMS is a late-stage alarm, not a recovery.

Kurtosis is the normalised fourth statistical moment of the waveform and quantifies the peakedness of the amplitude distribution. A random Gaussian signal has a kurtosis of approximately 3. Impulsive damage drives kurtosis above 4 to 5, and values above 7 to 8 usually indicate advanced localised defects. Kurtosis shares the non-monotonic behaviour of crest factor for the same reason. Because kurtosis is dimensionless and independent of overall amplitude, it is particularly useful on variable-load machines such as recycling shredders and granulators, where RMS naturally fluctuates with material feed rate.

Frequency-Domain Analysis: FFT, Envelope Demodulation and Order Tracking

Frequency-domain analysis converts a time waveform into a spectrum of amplitude against frequency, which is what allows a diagnosis rather than merely an alarm. A rising RMS tells the engineer that something is wrong; an FFT spectrum showing a dominant peak at exactly twice shaft speed with elevated axial content tells the engineer that a coupling needs realignment. Getting useful spectra from plastics machinery requires deliberate choices about frequency span, resolution, windowing and averaging.

Setting Up the FFT

The maximum analysis frequency, conventionally called Fmax, must be high enough to include the diagnostic content of interest. For general velocity severity work, 1000 Hz is sufficient. For a machine with a gearbox, Fmax must reach at least 3.25 times the highest gear mesh frequency so that the third harmonic and its sidebands are visible. For bearing analysis in acceleration, 5 kHz to 10 kHz is typical. The sampling rate must be at least 2.56 times Fmax to satisfy anti-aliasing requirements with a practical filter, so a 10 kHz Fmax implies 25.6 kilosamples per second.

Frequency resolution is Fmax divided by the number of spectral lines. A 1000 Hz span with 3200 lines yields 0.3125 Hz resolution and requires an acquisition time of 3.2 seconds, since acquisition time is the reciprocal of resolution. Resolution must be fine enough to separate sidebands spaced at shaft frequency from the carrier. On an extruder gearbox output shaft turning at 300 rpm, shaft frequency is 5 Hz, so a resolution of 0.5 Hz or better is needed to resolve those sidebands cleanly, and 1600 to 6400 lines is the usual choice.

Windowing prevents leakage caused by analysing a non-integer number of cycles. Hanning is the default for continuous machine vibration. Flat-top is used when amplitude accuracy matters more than frequency accuracy, such as during a calibration check. A uniform or rectangular window is used for transient capture such as a bump test. Averaging reduces the variance of the estimate: four to eight linear averages with 50 to 67 percent overlap is standard practice and gives a stable spectrum without excessive acquisition time.

Envelope Demodulation

Envelope demodulation, also called envelope analysis, amplitude demodulation or the high-frequency resonance technique, is the single most powerful tool for early bearing diagnosis. The principle is that a small raceway defect produces a train of very short impacts, each of which excites the structural resonances of the bearing and housing in the 1 kHz to 20 kHz region. The energy in these resonances is tiny compared with the 1x shaft component, so it is invisible in a normal spectrum. Envelope processing isolates it.

The processing chain has four steps. First, band-pass filter the raw acceleration signal to a band containing an impact-excited resonance, commonly 500 Hz to 10 kHz, or a narrower band chosen by spectral kurtosis. Second, rectify the filtered signal to obtain its magnitude. Third, low-pass filter to extract the slowly varying envelope. Fourth, compute the FFT of that envelope. The result is a low-frequency spectrum in which the bearing defect repetition rate appears as a clean discrete peak, typically between 20 Hz and 300 Hz for plastics machinery shaft speeds, often accompanied by harmonics and by sidebands spaced at shaft frequency.

Band selection is the main practical skill. A band that includes strong gear mesh energy will produce an envelope spectrum dominated by mesh-related content rather than bearing content. The spectral kurtosis method, visualised as a kurtogram, automatically searches the time-frequency plane for the band with the highest impulsiveness and is now standard in most analysis packages. Cepstrum analysis is a useful complement because it converts families of uniformly spaced sidebands into single peaks at the corresponding quefrency, which simplifies the identification of modulation caused by a rotating inner-race defect.

Order Tracking for Variable-Speed Machinery

Plastics machinery is increasingly driven by variable-frequency drives, and screw speed on a compounding extruder is a process variable that changes with recipe and throughput. A conventional FFT of a machine whose speed drifts during acquisition produces smeared peaks, because a component at 1x moves across several spectral lines. Order tracking solves this by resampling the signal in the angular domain using a tachometer or encoder reference, so that the horizontal axis becomes orders of shaft rotation rather than Hz. A 1x component stays at order 1.00 regardless of speed. For extruder drives that operate anywhere from 150 to 600 rpm depending on the compound being produced, order tracking is essential for meaningful trending; without it, a comparison between two measurements taken at different speeds is meaningless.

Rolling-Element Bearing Fault Frequencies: BPFO, BPFI, BSF and FTF

Rolling-element bearing defects generate vibration at four characteristic frequencies determined entirely by the bearing geometry and the shaft speed. Because these frequencies are non-integer multiples of shaft speed, they are readily distinguished from imbalance, misalignment and gear mesh components, which makes bearing diagnosis one of the most reliable applications of vibration analysis. Every predictive maintenance database should carry the calculated defect frequencies for each installed bearing.

The Four Defect Frequencies

Let fr be the shaft rotational frequency in Hz, n the number of rolling elements, d the rolling element diameter, D the bearing pitch diameter and alpha the contact angle. The four frequencies are calculated as follows.

• Ball Pass Frequency Outer race, BPFO equals n divided by 2, multiplied by fr, multiplied by the quantity one minus the ratio d over D times the cosine of alpha. This is the rate at which rolling elements pass a defect on the stationary outer race, and it is the most commonly observed defect frequency because the outer race carries the load zone.
• Ball Pass Frequency Inner race, BPFI equals n divided by 2, multiplied by fr, multiplied by the quantity one plus the ratio d over D times the cosine of alpha. Because the inner race rotates, an inner-race defect passes in and out of the load zone once per revolution, which amplitude-modulates the signal and produces sidebands spaced at exactly 1x shaft frequency around BPFI. Those sidebands are the definitive confirmation of an inner-race defect.
• Ball Spin Frequency, BSF equals D divided by twice d, multiplied by fr, multiplied by the quantity one minus the square of the ratio d over D times the cosine of alpha. A defect on a rolling element normally contacts both races per rotation, so 2 times BSF is often more prominent than BSF itself, with sidebands at the cage frequency.
• Fundamental Train Frequency, FTF equals fr divided by 2, multiplied by the quantity one minus the ratio d over D times the cosine of alpha. This is the cage rotation rate and always falls below 0.5x shaft speed, typically between 0.38x and 0.48x. A discrete FTF peak indicates cage damage, which is usually a late-stage and high-risk condition.

Useful approximations when bearing geometry is unavailable are BPFO equals about 0.4 times n times fr, and BPFI equals about 0.6 times n times fr. These are accurate to within roughly 10 percent for standard deep-groove and cylindrical roller bearings and are adequate for initial screening, though exact geometry should be entered before setting narrowband alarms.

Worked Example on an Extruder Motor

Consider a 6312 deep-groove ball bearing on an extruder main motor running at 1480 rpm, giving fr equal to 24.67 Hz. The bearing has 8 rolling elements, a ball diameter of 22.23 mm, a pitch diameter of 81.0 mm and a contact angle of zero. The ratio d over D is 0.2744. The calculated frequencies are FTF at 8.95 Hz, BPFO at 71.6 Hz, BSF at 41.6 Hz and BPFI at 125.8 Hz. An envelope spectrum showing a clear 71.6 Hz peak with harmonics at 143 Hz and 215 Hz is an unambiguous outer-race defect. A peak at 125.8 Hz flanked by sidebands at 101.1 Hz and 150.5 Hz, spaced 24.67 Hz apart, is an equally unambiguous inner-race defect.

The Four Stages of Bearing Failure

Bearing degradation follows a well-documented four-stage progression, and knowing which stage a bearing has reached determines how urgently it must be replaced.

Stage 1 produces energy only in the ultrasonic region, roughly 20 kHz to 60 kHz, detectable by acoustic emission or spike energy measurement. There is no change in the velocity spectrum and no audible or thermal symptom. Remaining life is commonly cited as 10 to 20 percent of the bearing rating life. Stage 2 excites bearing and housing natural frequencies in the 500 Hz to 2000 Hz region, and the envelope spectrum now shows a clear defect frequency. Remaining life is typically 5 to 10 percent of rating life. This is the ideal intervention window for planned replacement. Stage 3 brings the defect frequency and its harmonics into the conventional velocity spectrum, sidebands multiply, and temperature may begin to rise. Remaining life falls to roughly 1 to 5 percent. Stage 4 shows a rising broadband noise floor as the discrete defect frequencies smear into random content, 1x amplitude increases due to induced clearance, and audible noise and heat become obvious. Failure is imminent and the machine should be taken out of service at the first opportunity.

Gear Mesh Frequency and Extruder Gearbox Diagnostics

The reduction gearbox is the most failure-critical mechanical assembly on a twin-screw compounding extruder, because it must simultaneously transmit high torque, absorb the axial thrust generated by melt pressure, and maintain precise screw-to-screw timing. Gear mesh frequency analysis is therefore a core competence for anyone maintaining compounding lines such as the KTE series parallel co-rotating twin-screw extruders built by Kerke, a Wanplas factory, which range from the laboratory-scale KTE-16B up to the high-output KTE-135D.

Calculating Gear Mesh Frequency

Gear mesh frequency, GMF, equals the number of teeth on a gear multiplied by the rotational frequency of the shaft carrying that gear, expressed in Hz. Because both gears in a mesh share the same GMF, the calculation can be performed from either side and should give the same answer, which is a useful arithmetic check. A multi-stage gearbox has one GMF per stage. For a motor running at 1480 rpm, that is 24.67 Hz, driving a first-stage pinion with 23 teeth, GMF for stage one is 567 Hz. If that stage reduces to a shaft at 6.17 Hz carrying a 31-tooth second-stage pinion, GMF for stage two is 191 Hz. An analysis span must reach at least 3.25 times the highest GMF, so for this example an Fmax of at least 1850 Hz is required, and 2000 Hz would be selected in practice.

Interpreting Gear Spectra

A healthy gearbox shows a modest GMF peak with low, symmetric sidebands at 1x of each shaft. Diagnosis comes from how that pattern changes. A uniform increase in GMF amplitude across all stages usually indicates a load change rather than a fault, which is why gear spectra must always be compared at similar torque. An increase in GMF amplitude at one stage only, with growing 1x sidebands, indicates eccentricity or a mounting problem on that gear. High-amplitude sidebands spaced at the rotational frequency of a particular shaft localise the fault to the gear on that shaft. The appearance of 2x GMF and 3x GMF with rising sidebands indicates increasing backlash or wear. A single cracked or chipped tooth produces a once-per-revolution impact that shows in the time waveform as a periodic spike and in the spectrum as a broad family of shaft-speed harmonics; time synchronous averaging referenced to that shaft is the definitive confirmation technique because it removes everything not synchronous with the gear of interest.

Two additional frequencies are worth entering into the database. The gear assembly phase frequency, GAPF, equals GMF divided by the number of common factors between the tooth counts of the two gears and identifies faults affecting a group of teeth. The hunting tooth frequency, HTF, equals GMF multiplied by the number of common factors and divided by the product of the two tooth counts; it is very low, often below 1 Hz, and indicates the rate at which a specific pinion tooth re-engages a specific wheel tooth. A defect on one tooth of each gear produces a strong beat at HTF that is audible as a periodic growl and is a classic sign of a run-in problem after a rebuild.

Thrust Bearing Monitoring

In a twin-screw extruder gearbox the thrust bearing block is a distinct monitoring priority. Melt pressure acting on the screw tips generates a continuous axial load that is reacted by a stack of tapered roller or angular contact bearings in a very compact space. Because the load direction is axial, degradation appears first in the axial measurement direction, and a programme that only records horizontal and vertical readings on the gearbox will miss it. The recommended practice is a stud-mounted accelerometer on the thrust block housing oriented axially, sampled at the same interval as the main bearing points, with envelope analysis configured for the specific thrust bearing geometry. A rising axial acceleration RMS with a corresponding BPFO peak in the envelope spectrum is the earliest reliable indicator that a thrust stack rebuild should be scheduled, and catching it at that point avoids the far more expensive consequence of screw contact damage inside the barrel.

Spectral Signatures of Imbalance, Misalignment and Looseness

Imbalance, misalignment and mechanical looseness account for the majority of vibration faults on plastics machinery, and each has a distinctive spectral and phase signature that allows confident diagnosis without disassembly. The distinguishing information lies in three places: the ratio of 1x to 2x and higher harmonics, the relative amplitude in the axial direction, and the phase relationship between measurement points.

Imbalance produces a dominant 1x radial peak, low axial vibration, a stable and repeatable phase reading, and amplitude that increases approximately with the square of rotational speed. Horizontal and vertical phase at the same bearing differ by approximately 90 degrees. In plastics plants, imbalance most often develops on granulator and shredder rotors as blades wear unevenly, on centrifugal dryer rotors in recycling washing lines as material builds up on one side, and on winder and cooling fan assemblies. Overhung rotors are a special case producing high 1x in both radial and axial directions.

Misalignment produces a strong 2x component, frequently exceeding 1x, together with elevated axial vibration and a phase difference of approximately 180 degrees measured across the coupling. Angular misalignment loads the coupling in bending and dominates the axial direction; parallel offset misalignment loads it in shear and dominates the radial direction. Severe misalignment adds a 3x component and can generate significant energy at higher harmonics. Because thermal growth changes alignment between cold and running conditions, alignment should be verified at operating temperature on machines with substantial thermal gradients, which describes nearly every extruder drive train.

Mechanical looseness appears in three recognised types. Type A is structural looseness at the foundation or machine feet, showing a directional 1x with high phase variation between the base and the foot. Type B is looseness at the joint between components, such as loose bolts or a cracked bearing pedestal, showing 1x, 2x and 3x. Type C is fit looseness, such as an excessive clearance between a bearing outer ring and its housing, and it produces a distinctive pattern of many harmonics of 1x extending to ten orders or beyond, often with half-order subharmonics at 0.5x, 1.5x and 2.5x, and a raised noise floor.

Measurement Point Layout by Machine Type

A measurement point layout defines where sensors are placed, in which axes, at what frequency span and against which thresholds. Getting this right at the design stage matters more than any subsequent analytical refinement, because a fault that generates no signal at any monitored location will never be detected regardless of how the data is processed. The governing principle is that vibration must be measured on the load-carrying structure, as close to the bearing as physically possible, on machined metal with a direct mechanical path to the bearing outer ring.

Compounding and Pipe Extrusion Lines

An extrusion line requires monitoring on the main drive motor at both drive end and non-drive end, on the gearbox input, intermediate and output bearing housings, on the thrust bearing block in the axial direction, on any melt pump, and on downstream rotating equipment such as pelletiser cutter heads, haul-off units and winders. On the twin-screw compounding extruders and single-screw SE-series machines supplied by Wanplas factories, the gearbox housing is generously proportioned and provides good stud-mounting locations, but the thrust block is often shrouded by guarding and the mounting pad must be planned before installation rather than retrofitted. On pipe and profile lines from the Faygo factory, the vacuum calibration tank pumps and the haul-off caterpillar drives are frequently overlooked yet cause a meaningful share of unplanned stops.

Injection Molding and Blow Molding Machines

Hydraulic injection molding machines concentrate their rotating machinery in the power unit. The pump motor, the pump itself and the oil cooler circulation pump are the essential points, supplemented by reservoir oil temperature and an in-line particle counter. Fully electric and hybrid machines shift the emphasis to servo motor bearings, ball screw assemblies and toothed belt drives. Extrusion blow molding machines such as the ABLB and ABLD series from Apollo, a Wanplas factory, combine an extruder drive train with a hydraulic clamping unit and an accumulator head, so both monitoring philosophies apply on the same machine. On injection blow molding machines such as the IBM75, IBM65 and IBM55 Hybrid from the Aibim factory, the variable-displacement pump used in the PREFILL hydraulic system is the highest-value monitoring point because its condition directly governs parison wall consistency.

PET Blow Molding and Recycling Lines

On high-speed PET stretch blow molding machines such as the FGX series built by YuDa, a Wanplas factory, at outputs of 8000 to 15000 bottles per hour, the critical rotating assets are the main cam drive, the servo drive motors and, most importantly, the high-pressure air compressor supplying blowing air at up to 40 bar. Compressor condition monitoring should include crank-frequency and valve-related components as well as standard bearing analysis. Recycling lines from the Polyretec factory present the opposite challenge: shredders, granulators, friction washers and centrifugal dryers operate under highly variable, impact-rich loads where RMS fluctuates naturally with feed rate, so kurtosis and envelope indicators carry more diagnostic weight than overall level.

Injection Molding: Hydraulic Pump Vibration, Oil Temperature and Oil Quality

On a hydraulic injection molding machine the hydraulic power unit is both the largest energy consumer and the most common source of unplanned stoppage, and its condition is best assessed by combining vibration measurement with oil temperature and oil quality analysis. Vibration alone identifies mechanical degradation of the pump and motor; oil analysis identifies the contamination and oxidation processes that cause most of that degradation in the first place. Running the two techniques together typically detects developing problems several months earlier than either alone.

Pump Vibration Signatures

A variable-displacement axial piston pump generates a characteristic piston pass frequency equal to the number of pistons multiplied by shaft frequency. A nine-piston pump on a 1480 rpm motor produces a piston pass component at 222 Hz. A rising piston pass amplitude with harmonics indicates wear of the swash plate, slipper pads or valve plate. Cavitation is distinguishable because it produces broadband random energy from roughly 2 kHz to 10 kHz rather than discrete peaks, and it is almost always caused by suction-side restriction, a clogged suction strainer, a low reservoir level, or air entrainment from a leaking suction joint. Aeration produces a similar broadband signature accompanied by visible foam in the reservoir sight glass and by erratic pressure control. Because cavitation erodes internal surfaces rapidly, it should be treated as an urgent finding even when overall velocity RMS remains within Zone B.

Oil Temperature Control

Hydraulic oil temperature should be held between 40 and 50 degrees C for typical mineral-based ISO VG 46 fluids. Below 30 degrees C viscosity is too high and pump inlet starvation becomes a risk during cold starts. Above 55 degrees C an alert should be raised, and above 60 degrees C the machine should be tripped or unloaded. Oxidation rate follows an Arrhenius relationship, and the widely used engineering approximation is that oil service life halves for every 10 degrees C of sustained temperature above 60 degrees C. Sustained high temperature also softens seals, reduces film thickness, and accelerates varnish formation on servo valve spools, which manifests as sluggish or erratic position control long before any mechanical component fails outright.

Oil Cleanliness and Chemistry

Oil cleanliness is specified using the ISO 4406 three-number code, which reports particle counts greater than 4, 6 and 14 micrometres. General industrial hydraulic circuits are usually targeted at 18/16/13. Machines with proportional valves should be held at 17/15/12, and high-response servo valve systems on precision and hybrid electric machines require 16/14/11 or cleaner. Each single-digit reduction in an ISO code represents halving the particle count, and moving from 20/18/15 to 17/15/12 typically extends component life several times over. Wear metal analysis by spectrometry provides the complementary diagnosis: iron indicates pump and cylinder wear, copper indicates bushing and cooler tube wear, chromium points to plated rod surfaces, aluminium to pump housings and pistons, and silicon to ingressed dust.

Tie bar strain monitoring deserves mention alongside hydraulic condition. Uneven clamp force distribution accelerates platen wear and produces flash on one side of the mold. Strain gauges applied to each tie bar should show clamping loads balanced within roughly 5 percent of one another; a persistent imbalance beyond that indicates platen parallelism error or a worn toggle linkage and correlates strongly with subsequent tie bar fatigue failure.

Setting Alarm Thresholds: Good, Satisfactory, Unsatisfactory, Unacceptable

Alarm thresholds convert measurements into decisions, and poorly set thresholds are the most common reason predictive maintenance programmes are abandoned. Thresholds that are too tight generate nuisance alarms until the maintenance team stops trusting the system; thresholds that are too loose allow failures to progress undetected. The correct approach combines three methods: absolute limits from ISO 20816, statistical limits derived from each machine’s own history, and narrowband limits on specific diagnostic frequencies.

Absolute limits come directly from the zone boundary table. They require no historical data and can be applied from the first measurement, which makes them the right starting point for a new installation. Their weakness is that they are generic; a machine that inherently runs at 2.6 mm/s RMS because of an unusual support structure will sit permanently near the B/C boundary without any developing fault.

Statistical limits solve this by referencing the machine to itself. After ten to twenty measurements under comparable operating conditions, compute the mean and standard deviation of each trended indicator and set the alert at mean plus three standard deviations, or where variance is low, at 1.5 to 2 times the established baseline. The danger level is then set at approximately 2.5 to 3 times baseline, capped so that it never exceeds the ISO Zone C/D boundary. This combination catches both machines that degrade from a low baseline and machines whose absolute level is genuinely excessive.

Narrowband limits are the diagnostic layer. Rather than alarming on the overall level, they alarm on the energy within a defined frequency window, such as a window centred on BPFO plus or minus 5 percent, or a window spanning 0.9 to 1.1 times GMF. Narrowband alarms detect faults that are invisible in the overall reading because the defect energy is small relative to total energy, and they carry diagnostic meaning automatically, since an alarm on the BPFO band is a bearing alarm rather than a generic vibration alarm.

Rate-of-change alarms complete the scheme. Any indicator that rises by more than roughly 20 percent between consecutive measurements should trigger investigation regardless of its absolute value, because a step change in a stable machine almost always has a physical cause.

The P-F Curve and Condition-Based Maintenance Planning

The P-F curve is the conceptual foundation of condition-based maintenance. It plots equipment condition against time and identifies two points: P, the point at which a developing failure first becomes detectable by some technique, and F, the point of functional failure. The interval between them, the P-F interval, is the window available for planned intervention, and the whole purpose of selecting monitoring techniques and intervals is to make that window as long and as reliably observed as possible.

Different techniques detect the same failure at different points on the curve. Acoustic emission and ultrasonic methods detect bearing lubrication distress and micro-scale surface damage earliest, giving a P-F interval commonly cited as 6 to 12 months. Vibration envelope analysis detects the resulting localised defect at 3 to 6 months. Oil analysis detects the wear debris generated by that defect at 2 to 6 months depending on sampling frequency. Infrared thermography detects the friction-generated heat at 1 to 3 months. Audible noise gives 2 to 4 weeks, and heat detectable by hand gives days. The critical planning rule is that the data collection interval must be no longer than one third to one half of the P-F interval, because a single measurement taken just before P and the next taken just after F provides no warning at all.

This rule immediately determines a monitoring schedule. If the objective is to catch bearing defects using envelope analysis with a P-F interval of 3 to 6 months, then monthly data collection is the maximum acceptable interval, and fortnightly is safer. If the objective is to catch lubrication distress using ultrasonic measurement with a 6 to 12 month P-F interval, quarterly collection suffices. Because the shortest relevant P-F interval governs, a plant that wants comprehensive coverage of bearing failures will settle on a monthly route supplemented by continuous online monitoring on the small number of assets whose failure would stop the whole plant.

Data Acquisition Intervals and Wireless Sensor Networks

The architecture of data collection determines both the cost and the effectiveness of a predictive maintenance programme, and the right answer is almost always a hybrid of permanently installed online sensors on a small number of critical assets and portable route-based collection on the wider population. Deciding which assets belong in which category is a criticality assessment, not a technical question, and it should be based on the consequence of failure rather than on the cost of the machine.

Route-Based Collection

Route-based collection uses a portable data collector and a defined route of measurement points, typically walked monthly. It offers the highest data quality per measurement because the analyser has wide dynamic range, can capture long waveforms and can be reconfigured on the spot to investigate an anomaly. It also puts a trained human in front of the machine, which catches leaks, loose guards and abnormal noise that no sensor reports. Its limitations are labour intensity, the inability to see anything that happens between routes, and dependence on consistent operating conditions at the moment of measurement, which is difficult on machines whose load varies with the product being run.

Online and Wireless Systems

Permanently installed systems remove the sampling gap. A wired online system with a multi-channel monitor is the appropriate choice for the highest-criticality assets, such as the main gearbox of a continuously operating compounding line, because it provides simultaneous multi-channel acquisition with phase reference. Wireless nodes have become the pragmatic choice for the broad middle tier. A modern industrial wireless node combines a MEMS or piezoelectric sensing element with edge processing, transmitting computed indicators rather than raw waveforms in order to conserve battery and bandwidth. Typical configurations report overall velocity RMS, acceleration RMS, peak, crest factor, kurtosis, several band energies and temperature once per hour, plus one full spectrum and time waveform per day, and achieve three to five years of battery life on that duty cycle. When an indicator crosses an alert threshold, the node automatically increases its reporting rate and uploads full raw waveforms for diagnosis.

Network selection depends on plant layout. IEEE 802.15.4 mesh protocols including WirelessHART and ISA100 offer robust self-healing meshes suited to dense equipment areas. Bluetooth 5 Low Energy is well matched to short-range, high-density deployments around a single machine line. LoRaWAN suits sprawling sites with long distances and low data rates, such as an outdoor recycling yard with washing and drying equipment spread across a large area. In every case, the wireless segment should be treated as part of the operational technology network and secured accordingly, with IEC 62443 providing the applicable framework for zones, conduits and device hardening.

Acquisition Settings That Matter

Three acquisition settings determine whether the collected data is diagnostically useful. The sampling rate must be at least 2.56 times Fmax so that anti-alias filtering works. The waveform length must span at least six to ten shaft revolutions for general analysis and twenty or more revolutions for envelope analysis, so that the defect frequency has enough repetitions to resolve; on a 300 rpm extruder output shaft, twenty revolutions is four seconds of data. Finally, operating conditions must be recorded with every measurement. Screw speed, motor load, melt pressure, barrel zone temperatures and product code should be tagged onto each record, because a vibration trend that ignores load is not a trend at all. The remote monitoring capability built into the FGX-series PET blowing machines from the YuDa factory illustrates the value of this pairing: because engineers at the China headquarters can review live PLC data alongside machine behaviour, an anomaly can be interpreted in the context of the actual production settings rather than in isolation.

Integrating Condition Monitoring with MES and SCADA

Condition monitoring delivers value only when its output reaches the people and systems that schedule work, which means integration with the plant SCADA layer and the manufacturing execution system is not an optional refinement but a core requirement. An isolated vibration database that only the reliability engineer opens will not change maintenance behaviour; a health state that automatically appears on the production supervisor’s dashboard and raises a work order in the CMMS will.

ISO 13374 defines the reference architecture for this integration through the six-block OSA-CBM model: data acquisition, data manipulation, state detection, health assessment, prognostic assessment and advisory generation. The value of the model is that it separates the layers cleanly, so that sensors and analytics from different vendors can be combined without rewriting the whole system. ISO 17359 provides the general procedural guidelines for a condition monitoring programme, ISO 13379 covers data interpretation and diagnostic techniques, and ISO 13381 addresses prognostics and remaining useful life estimation. Personnel qualification is addressed by ISO 18436-2, which defines four categories of vibration analyst certification; Category II is generally the minimum for anyone setting alarms and issuing diagnoses.

The practical data path uses an edge gateway that aggregates sensor data, applies the state detection and health assessment logic, and publishes results northbound. OPC UA is the preferred protocol where a structured information model is wanted, because condition monitoring companion specifications allow health states, alarms and asset hierarchies to be exposed in a standard way. MQTT with a Sparkplug B payload is lighter and better suited to large numbers of wireless nodes reporting infrequently. Modbus TCP remains common for reading process context from older machine PLCs. Whichever transport is used, the essential design decision is to combine condition data with process context in a single time-aligned record, so that analysis can distinguish a genuine mechanical change from a normal response to a recipe change.

Downstream, three integrations pay for themselves quickly. Automatic work order generation from a confirmed Zone C alarm ensures that a diagnosis becomes a scheduled job rather than an email. Spare part reservation linked to the diagnosed component means that the correct bearing is on site before the planned intervention, which is what allows spare-parts inventory to be reduced overall rather than simply reshuffled. Feeding equipment health states into the OEE calculation lets management see availability losses attributed to specific assets and failure modes, which in turn directs the next round of reliability improvement to where it will have the greatest effect.

Implementation Roadmap and Return on Investment

A predictive maintenance programme should be implemented in phases, because attempting to instrument an entire plant at once almost always produces more data than the organisation can act upon. A staged approach builds analytical capability and organisational trust at the same rate as it builds sensor coverage.

Phase one, criticality assessment and baseline. Rank every rotating asset by the consequence of its failure, considering production loss, safety, quality impact and repair lead time. Select the top 10 to 15 percent as critical. Establish measurement points, install permanent mounting pads and record a baseline signature under known, documented operating conditions. Enter bearing and gear geometry into the database so that defect frequencies are calculated rather than guessed.

Phase two, route-based monitoring. Deploy monthly portable collection across the critical assets and the next tier. Use absolute ISO thresholds initially and replace them with statistical thresholds as history accumulates. Train at least one analyst to ISO 18436-2 Category II. Expect the first six months to reveal a backlog of pre-existing defects; resolving these produces the initial and usually the largest step change in reliability.

Phase three, online and wireless coverage. Install permanent monitoring on assets whose failure stops an entire line, and wireless nodes on assets that are inaccessible, guarded or hot. Configure narrowband alarms on calculated defect frequencies. Add oil analysis on hydraulic systems and gearboxes.

Phase four, integration and prognostics. Connect the condition monitoring layer to SCADA, MES and CMMS. Add process context to every record. Begin remaining-useful-life estimation on the failure modes for which sufficient run-to-failure history now exists.

Return on investment should be evaluated in relative rather than absolute terms, because the numbers depend entirely on the value of the product being made and the cost structure of the plant. The commonly reported outcomes from mature programmes are a 20 to 50 percent reduction in unplanned downtime, a 10 to 30 percent reduction in spare-parts inventory value, a 5 to 15 percent improvement in overall equipment effectiveness, and a 20 to 40 percent reduction in total maintenance labour hours as emergency work is converted into planned work. Secondary benefits include extended component life from earlier correction of misalignment and lubrication problems, and reduced energy consumption, since a misaligned or unbalanced drive train converts a measurable fraction of its input power into heat and vibration rather than useful work.

Equipment specification also influences the economics. Machines delivered with pre-planned sensor mounting locations, accessible bearing housings, documented bearing and gear geometry and a commissioning vibration baseline are substantially cheaper to bring into a monitoring programme than machines that require retrofit. Wanplas positions this as part of its equipment documentation package across its factory network, alongside brand-level commitments such as the annual free spare parts allowance, the transportation guarantee, the production capacity guarantee and the open factory policy that allows customers to witness testing before shipment. Buyers evaluating equipment from any supplier, whether Wanplas, Coperion, ENGEL, Haitian, KraussMaffei, Sidel or EREMA, should treat monitoring readiness as a specification line item rather than an afterthought.

Frequently Asked Questions

What vibration level is acceptable for an extruder main drive motor?

For a typical extruder main motor in the 15 kW to 300 kW range mounted on a rigid foundation, ISO 20816-3 places the Zone A/B boundary at 1.4 mm/s RMS, the B/C boundary at 2.8 mm/s RMS and the C/D boundary at 4.5 mm/s RMS, measured as broadband velocity over 10 Hz to 1000 Hz. Readings below 2.8 mm/s RMS are acceptable for unrestricted long-term operation. Above 4.5 mm/s RMS the machine is in the unacceptable zone and damage accumulation should be assumed. Remember that a 50 percent increase from an established baseline warrants investigation even when the absolute value is still within Zone B.

Why does envelope demodulation detect bearing faults earlier than a normal FFT spectrum?

A small spall in a bearing raceway produces very low-energy, very short impacts that are buried under the far larger 1x shaft and gear mesh components in a standard velocity spectrum. Envelope demodulation first band-pass filters the acceleration signal in a high-frequency band, typically 500 Hz to 10 kHz, where only the impact-excited bearing and housing resonances live, then rectifies and low-pass filters the signal to extract the repetition rate of those impacts. The resulting envelope spectrum shows BPFO, BPFI, BSF or FTF as clear discrete peaks months before they appear in a conventional velocity FFT.

How often should vibration data be collected on plastic processing equipment?

The collection interval should be no longer than one third to one half of the shortest expected P-F interval for the failure modes being monitored. For rolling-element bearings monitored by envelope analysis the P-F interval is typically three to six months, so a monthly route or an online system reporting daily is appropriate. Critical continuous assets such as compounding extruder gearboxes and high-pressure PET blowing compressors justify permanently installed sensors reporting overall values every 5 to 15 minutes with a full spectrum daily.

What is the difference between crest factor and kurtosis for bearing diagnosis?

Crest factor is the ratio of peak amplitude to RMS amplitude and rises from around 3 to 4 on a healthy machine to 5 to 8 when isolated impacts appear. Kurtosis is the normalised fourth statistical moment of the waveform and sits near 3 for a healthy Gaussian signal, rising above 4 to 5 when impulsive damage develops. Both indicators are non-monotonic: in late-stage failure the impacts become continuous, RMS rises sharply and both indicators fall back toward their healthy values, which is why they must always be trended alongside overall RMS rather than interpreted alone.

Which oil cleanliness code should an injection molding machine hydraulic system meet?

General industrial hydraulic circuits on injection molding machines are usually targeted at ISO 4406 code 18/16/13. Machines with proportional valves should be held to 17/15/12, and high-response servo valve systems used on precision and hybrid electric machines require 16/14/11 or cleaner. Water content should stay below 500 ppm, ideally below 200 ppm, and a total acid number increase greater than 0.3 mg KOH per gram over the new-oil baseline indicates that oxidation has progressed far enough to justify a fluid change and a review of operating temperature.

Can wireless vibration sensors replace route-based data collection in a plastics plant?

Wireless sensors are well suited to assets that are hard to reach, guarded, or too hot to approach safely, and modern nodes achieve three to five years of battery life while reporting overall values hourly and one full spectrum per day. They do not fully replace route-based collection, because a portable analyser still provides higher dynamic range, longer waveforms and the ability to capture transients on demand, and because a technician walking the route notices leaks and loose guards that no sensor reports. Most successful programmes combine permanent wireless nodes on critical assets with periodic routes on the remaining population.

What spectral pattern indicates shaft misalignment rather than imbalance?

Pure imbalance shows a dominant 1x radial peak with low axial vibration and a stable phase reading, with amplitude rising roughly with the square of speed. Misalignment produces a strong 2x component, often exceeding 1x, together with elevated axial vibration and a phase difference of approximately 180 degrees measured across the coupling. Angular misalignment dominates the axial direction while parallel offset dominates the radial direction, and severe cases add a 3x component. Because thermal growth changes alignment, verification should be performed at operating temperature.

How are bearing defect frequencies calculated without the manufacturer data sheet?

Exact calculation requires the number of rolling elements, the ball or roller diameter, the pitch diameter and the contact angle. When these are unavailable, the approximations BPFO equals about 0.4 times the number of rolling elements times shaft frequency, and BPFI equals about 0.6 times the number of rolling elements times shaft frequency, are accurate to within roughly 10 percent for standard deep-groove and cylindrical roller bearings. FTF always falls between 0.38 and 0.48 times shaft frequency. These estimates are adequate for screening, but exact geometry should be obtained before configuring narrowband alarms.

Why does an extruder gearbox thrust bearing need a dedicated axial measurement?

The thrust bearing stack in a twin-screw extruder gearbox reacts the axial load generated by melt pressure acting on the screw tips, so degradation develops primarily in the axial direction. A monitoring scheme that records only horizontal and vertical readings on the gearbox housing will show little change until the fault is advanced. A stud-mounted accelerometer on the thrust block oriented axially, with envelope analysis configured for that specific bearing geometry, is the only reliable way to catch the fault early enough to schedule a rebuild before screw contact damage occurs inside the barrel.

How is condition monitoring data integrated with MES and SCADA systems?

The usual architecture uses an edge gateway that publishes processed condition indicators over OPC UA or MQTT with a Sparkplug B payload, while the machine PLC exposes process context such as screw speed, melt pressure, cycle count and oil temperature over the same protocol or Modbus TCP. ISO 13374 defines the six-block OSA-CBM data flow from data acquisition through to advisory generation, which maps cleanly onto this architecture. Health states are then written back to the MES so that a confirmed alarm automatically raises a CMMS work order carrying the diagnosed fault and the recommended action, and so that availability losses can be attributed to specific assets in the OEE calculation.

Does variable screw speed make vibration trending unreliable on compounding lines?

It does if the data is analysed only in the frequency domain in Hz, because a 1x component moves across spectral lines as speed changes and comparisons between measurements become meaningless. Order tracking solves this by resampling the signal in the angular domain using a tachometer or encoder reference, so the horizontal axis becomes orders of shaft rotation and a 1x component stays at order 1.00 regardless of speed. Recording motor load and screw speed with every measurement is equally important, because amplitude also varies with torque even when the fault condition has not changed.

Conclusion

Predictive maintenance for plastic machinery is a disciplined engineering practice rather than a technology purchase. Its foundations are physical: measure velocity for severity and acceleration for bearing and gear diagnosis, judge severity against the ISO 20816 zone boundaries appropriate to the machine class and support condition, trend RMS alongside crest factor and kurtosis so that both gradual and impulsive degradation are visible, and use envelope demodulation with calculated BPFO, BPFI, BSF and FTF values to identify the specific damaged component months before it fails. Layer gear mesh frequency analysis onto extruder gearboxes, add a dedicated axial measurement on the thrust bearing block, and pair vibration with oil temperature and ISO 4406 cleanliness monitoring on every hydraulic system.

Around that technical core, the organisational elements determine whether the programme survives. Thresholds must combine absolute ISO limits, machine-specific statistical limits and narrowband diagnostic limits, escalating through Good, Satisfactory, Unsatisfactory and Unacceptable with a defined owner and response time at each level. Collection intervals must be set from the P-F interval of the failure modes being targeted, not from convenience. Data must reach the MES and CMMS so that a diagnosis becomes a scheduled job with the right spare part already reserved. Implemented in this way, a condition monitoring programme reliably delivers the outcomes reported across the industry: substantially lower unplanned downtime, leaner spare-parts inventory, higher overall equipment effectiveness and longer component life.

Wanplas builds and supports the full spectrum of plastic processing equipment through its network of specialised factories, from Kerke twin-screw compounding extruders and Faygo pipe and profile lines to Apollo extrusion blow molding machines, Aibim injection blow molding systems, YuDa PET bottle blowing machines, YuanSu film, sheet and board lines and Polyretec recycling equipment. Every line is a candidate for the monitoring framework described here, and specifying sensor mounting pads, documented bearing and gear geometry and a commissioning vibration baseline at the time of order costs very little and removes most of the friction from a later programme rollout. For plants planning new capacity or upgrading existing lines in 2026, the most valuable step is to treat condition monitoring readiness as part of the machine specification, and to talk to the Wanplas engineering team about how each line should be instrumented before it ships rather than after it has already been installed.