Predictive maintenance: vibration, thermography and oil analysis
The three most widely used predictive techniques measure a real physical condition of the equipment to anticipate failure, instead of intervening on a calendar. A case with a diesel engine shows what this looks like in practice.
- Topic
- Reliability and TPM
- Reading time
- 11 minutes
- Sources
- 3 articles, 1 paper
- Tool
- Reading only
In one line
Predictive maintenance does not intervene on a calendar or wait for the breakdown: it measures a physical variable of the equipment (vibration, temperature, oil condition) and acts when that variable shows that something is degrading.
What it is
It is a preventive maintenance tool, not a separate type of maintenance: it continuously tracks variables related to how the machine operates in order to predict failures and decide on the corrective action at the right moment, neither too early (wasting the service life of a healthy component) nor too late (with the equipment already stopped). Its implementation cost is high compared with corrective or calendar-based preventive maintenance, but today it is the most widely used approach in industry because it shows the real condition of every machine in the plant at any time.
Vibration analysis
It is applied to rotating machinery. All machines vibrate even when running well — what matters is when the level or the shape of that vibration changes. The measurement is taken with sensors placed on the bearing supports (which is where most vibrations are transmitted), taking readings in three directions: two radial (vertical and horizontal) and one axial.
The raw signal over time is hard to interpret because it mixes the vibrations of all the machine's parts at once. That is why the Fast Fourier Transform (FFT) is applied to move the signal into the frequency domain: in the resulting spectrum, the frequency tells you what is causing the vibration and the amplitude tells you how serious the problem is.
Typical failures this technique detects:
- Unbalance and shaft misalignment
- Bearing defects and bent shafts
- Mechanical looseness
- Defects in belt drives and gears
- Electrical problems associated with motors
A requirement that is often underestimated: for the diagnosis to make sense you need to know the machine's technical data, its maintenance history and its reference spectra beforehand — without that baseline, an isolated spectrum says little.
Thermography
It is based on a simple physical principle: every body with a temperature above absolute zero emits thermal radiation, and that radiation is proportional to the object's temperature. A thermographic camera captures that infrared radiation (in the range of 8 to 15 µm) and converts it into a visible image called a thermogram.
The cameras deliver two types of reading:
- Qualitative: a color image that lets you locate at a glance the hottest points of a panel, motor or installation
- Quantitative: the exact temperature value at those points, which lets you decide how serious the problem is and how urgent the intervention is
It is used mainly on high- and low-voltage electrical installations, motors and windings, furnaces, boilers and heat exchangers, steam traps and HVAC systems — any point where an incipient problem (a loose connection, a bearing starting to fail) shows up first as a localized temperature rise. Its operational advantage is that it requires no physical contact and no need to stop the equipment to take the measurement, which makes it safe even on energized electrical panels.
Oil analysis
Lubricating oil performs three functions — protecting against wear, controlling temperature, carrying away impurities — and when it degrades or becomes contaminated it stops performing them, and that is when the machine starts to fail. Oil analysis quantifies that condition with laboratory tests on a sample:
- Contamination: metallic wear particles, fuel, water, carbonaceous matter
- Degradation: viscosity, detergency, basicity (the capacity to neutralize acids)
Real example
A case published at the LACCEI 2013 conference applied oil analysis to an Isuzu 4BD1 diesel engine (4 cylinders, compression ratio 17.5:1) in the UNEXPO laboratory in Barquisimeto, Venezuela. The engine was run without load for 150 continuous hours, with oil samples taken every 15 hours.
The results show exactly the kind of early warning this technique is meant to provide:
- The viscosity dropped progressively and crossed the lower control limit (12.66 cSt) at around 45 hours — 33% of the test period
- From those same 45 hours on, fuel dilution appeared and kept increasing without exceeding the maximum allowed (5%)
- The soot content increased considerably, a sign of incomplete combustion
- The wear metals (iron, chromium, lead, copper, tin, aluminum) rose progressively but stayed within the permissible limits over the 150 hours of the test
The study's conclusion was specific: the problem was not mechanical wear (the metals were still under control) but combustion — probably poorly calibrated or leaking injectors — and the recommendation was to calibrate the injectors and the injection pump. It is the typical result of a good oil analysis program: it does not just say "the oil is bad", it points to which system of the equipment you need to look at.
Benefits
- It lets you schedule the repair without interrupting production, because detection happens with the equipment running under normal conditions
- The total cost (predictive maintenance + scheduled repair) is lower than that of an unplanned failure with production downtime included
- It is not an expense: it is an investment that avoids the economic loss of an unexpected stoppage
Limitations to keep in mind
- The initial cost is high: specialized equipment (vibration analyzer, thermographic camera) and trained personnel to interpret the results, not just to collect them — many companies end up outsourcing thermography for this reason
- Each technique needs its own baseline for the equipment (reference spectrum, normal temperature range, virgin oil for comparison); without that starting point, a single isolated reading does not allow you to diagnose anything with certainty
- No technique covers all failure modes by itself: in the diesel engine case, the wear metals showed no alert while the viscosity and the soot did — you need to look at several parameters together, not just one, before drawing a conclusion
In summary
Vibration, thermography and oil analysis attack three different physical symptoms of the same idea: the machine warns you before it fails, if someone is measuring the right variable. The diesel engine case shows it in concrete numbers — 45 hours of advance warning on a combustion problem, detected from an oil sample without opening the engine.
More on Reliability and TPM
Risk-based inspection (RBI): inspect more where the risk is higher
RBI prioritizes and plans equipment inspections according to risk, which combines the probability of failure with its consequences. It uses the same logic as criticality analysis, but to decide what, when and how to inspect.
TPM: the pillars of Total Productive Maintenance
TPM moves basic maintenance tasks to the operator and organizes everything else into eight pillars, with the stated goal of zero failures, zero defects and zero accidents.
What is RCM (Reliability Centered Maintenance)
RCM is not just another type of maintenance: it is a method for deciding which maintenance strategy fits each failure mode, by answering seven questions.