Predictive Maintenance for Refineries and Petrochemical Plants: Predictive Maintenance Indonesia

Direct answer (AEO): A refinery or petrochemical complex is not one plant but a set of linked process units with different economics, different failure consequences, and different tolerance for downtime. That structure makes the central predictive maintenance question a question of allocation: which units justify continuous monitoring, which justify periodic surveillance, and which should be allowed to run to failure because their failure is cheap and contained. In Indonesian refining and petrochemical operations, the plants that manage this well typically run a tiered programme — online coverage on the critical compressors and rotating trains, disciplined periodic routes on balance-of-plant equipment, and explicit run-to-failure decisions with documented justification. The plants that struggle usually apply one uniform approach across everything and end up with either excessive cost or unnoticed degradation.

Operating a refinery or petrochemical plant in Indonesia? Contact Tiaravib — email info@tiaravib.com or WhatsApp +62 821-1378-3727.

Why a Uniform Programme Always Fails in a Complex

The instinct when building a reliability programme across a refinery is to standardise: same technique, same interval, same coverage everywhere. It is administratively tidy and it is economically wrong, because the assets in a complex are not comparable. A platformer recycle compressor and a cooling water circulation pump both rotate, both have bearings, and both can be monitored with vibration — but their failure consequences differ by orders of magnitude, and treating them identically wastes effort on one while under-protecting the other.

The practical consequence of a uniform programme is predictable. Effort spreads thinly across thousands of measurement points, the analyst’s attention is consumed by high-volume low-consequence findings, and the genuinely dangerous trends — the ones on the machines that stop a unit — get buried in a long report. Meanwhile the plant believes it has comprehensive coverage. This is one of the most common and most damaging patterns in large Indonesian complexes, and it usually survives until the first serious event exposes it.

The corrective is a tiered structure built directly on a criticality assessment. Tier one assets justify continuous online monitoring because their failure stops a unit and their degradation can progress within days. Tier two assets justify a disciplined periodic route because their failure is expensive but slower-moving, or because the unit has partial redundancy. Tier three assets justify run-to-failure — with the decision documented, so that the eventual failure is a planned event rather than a surprise. The discipline of writing down the run-to-failure decisions is what makes the tier structure real; without it, tier three is indistinguishable from neglect. The framework for this allocation is the core of any predictive maintenance framework, and in a complex it is the only part of the framework that genuinely matters.

Where Continuous Coverage Earns Its Cost

Online monitoring is expensive relative to a periodic route — sensors, cabling, acquisition hardware, communications, and the analyst time to interpret continuous data all carry cost. It earns that cost only where three conditions coincide: the consequence of failure is severe, the degradation can develop faster than the periodic interval, and the failure mode produces a detectable signature.

In refining and petrochemical plants, the assets that satisfy all three are consistent across configurations. Critical compressors on the process path — recycle, make-up, refrigeration, and hydrogen compressors — combine severe consequence with fast-developing failure modes and clear vibration signatures. Large steam and gas turbines driving those compressors carry the same profile plus efficiency degradation that is only visible in performance trending. High-energy pumps in critical service, particularly where there is no installed spare, justify online coverage because a bearing failure can escalate to a catastrophic event. And the lube and seal oil systems of the above belong in the same tier, because they cause a large proportion of the trips.

Note that the last row is not an asset tier at all — it is a data layer that sits across the tiers and usually gets neglected. In a refinery, the most valuable early warning on a compressor is frequently its efficiency against flow at constant conditions, which reveals fouling, internal leakage, and impeller erosion long before any mechanical symptom. A programme that monitors vibration excellently and ignores performance is monitoring half the failure modes.

Turnaround Coupling: Where Refinery Programmes Succeed or Fail

In general industry, a predictive finding can usually be acted upon within weeks. In a refinery, many findings can only be acted upon during the unit turnaround, because the machine cannot be opened, the spare cannot be installed, or the isolation required is only available during an outage. This changes the entire value calculation: a finding that arrives eight weeks before a turnaround is worth substantially more than the same finding arriving two weeks after it, because the latter means waiting another full cycle.

This has a direct structural implication for the programme. Surveillance intensity on turnaround-scope assets should increase in the months before an outage, not remain constant. The objective is to find every defect that will need the outage while the outage can still be planned around it, and to establish with confidence which assets genuinely do not need to be opened. That is precisely the logic described in the shutdown and turnaround approach, and in a refinery it is not an optimisation but a fundamental requirement, because the turnaround is the only maintenance window that exists for a large part of the asset base.

The recurring failure is a mismatch of calendars. The reliability team runs a twelve-month monitoring cycle; the turnaround team plans on an eighteen-to-twenty-four month cycle tied to catalyst life and statutory inspection. Unless the two are explicitly synchronised, the monitoring data will arrive at a point in the planning cycle where it cannot influence scope, and the knowledge is wasted. Fixing this costs nothing but a meeting and a written agreement on cut-off dates.

Operating Constraints That Shape the Programme

Refineries and petrochemical plants impose constraints on condition monitoring that are more severe than in most industries, and a programme designed without them will not survive contact with operations. Hazardous area classification restricts instrumentation to certified equipment, which affects both sensor selection and the cost of any retrofit. Hot work and access permits add time to every field activity, so route planning must account for permit availability rather than assuming access. Flare, relief, and emergency systems impose operating limits that constrain when machines can be tested or loaded.

Two further constraints are specific and frequently underestimated. First, many critical machines cannot be safely approached while running, which means measurement points must be installed during outages and protected afterwards — a programme cannot simply be extended by walking up to a machine. Second, hydrogen and light hydrocarbon service introduces seal integrity concerns that make oil and seal system monitoring disproportionately important, because a seal failure on a hydrogen compressor is both a reliability event and a safety event.

The practical result is that a refinery programme is designed at the turnaround and executed between turnarounds, exactly as in a fertilizer complex. Sensor installation, cable routing, junction box placement, and measurement point definition are outage activities. A plant that tries to build coverage incrementally during operation will find itself limited to whatever happens to be accessible, which is rarely where the risk sits.

Making the Commercial Case to Operations

Refinery reliability proposals frequently win or lose on how they are framed. Presented as a maintenance initiative, they compete with every other maintenance priority for budget and attention. Presented as a unit-availability measure with an explicit margin calculation, they compete on different terms entirely.

The calculation is straightforward once the units are separated. For each process unit, determine the margin lost per day of unplanned shutdown, then determine the historical frequency of unplanned stops attributable to rotating equipment. The product is the annual exposure. Against that, set the cost of a tiered programme. In most Indonesian complexes the ratio is decisive, and the honest version of the calculation is what persuades operations leadership — including the cases where a specific unit’s exposure turns out to be too small to justify intensive monitoring, which builds credibility for the units where it is large.

This is also where the programme connects to the wider asset management agenda. A tiered monitoring structure built on criticality feeds directly into capital planning, because the same criticality assessment that allocates monitoring effort also identifies the assets where replacement or upgrade is economically justified. The relationship between monitoring, capital planning, and lifecycle cost is direct rather than incidental, and refineries that treat them as separate exercises duplicate the analysis and often reach inconsistent conclusions.

Related practice for Indonesian operators is set out in LNG facility monitoring.

Key Takeaways

• A refinery is a set of linked units with different economics, so a uniform monitoring programme always misallocates effort.
• Three conditions must coincide to justify online coverage: severe consequence, fast-developing failure, and a detectable signature.
• Lube and seal oil systems belong in the top tier because they cause a large share of turbomachinery trips.
• Performance trending across the tiers is routinely neglected and detects failure modes that vibration cannot see.
• Surveillance intensity should rise before a turnaround, because most findings can only be acted upon during an outage.
• Hazardous area, permit, and access constraints mean the programme is designed during outages and executed between them.

Frequently Asked Questions

How do we decide which machines get online monitoring in a refinery?

Apply three tests together: does failure stop or severely constrain a unit, can the degradation develop faster than the periodic interval, and does the failure mode produce a detectable signature. Machines that pass all three go online. Machines that fail any one of them are better served by a periodic route at an appropriate interval.

Is run-to-failure ever the right decision?

Yes, for contained failures on assets with installed spare where the cost of monitoring exceeds the cost of the failure. The essential discipline is documenting the decision explicitly, so the eventual failure is an anticipated event rather than a surprise. Undocumented run-to-failure is indistinguishable from neglect and will be judged that way afterwards.

Why should monitoring increase before a turnaround?

Because most rotating equipment findings can only be acted upon during an outage. A finding that arrives eight weeks before a turnaround shapes the scope; the identical finding arriving two weeks after the turnaround means waiting a full cycle. Surveillance intensity on turnaround-scope assets should therefore rise in the months preceding the outage.

What is the most neglected data layer in refinery programmes?

Process performance. Compressor efficiency against flow, heater duty, and exchanger fouling reveal degradation that vibration cannot detect, and they often provide the earliest warning of a developing problem. Most plants monitor vibration well and performance poorly, which covers roughly half the relevant failure modes.

How much of a refinery’s trips come from auxiliary systems?

A substantial share, and consistently more than plants expect. Lube oil contamination, filter blockage, seal gas supply faults, and control valve malfunction account for a large proportion of turbomachinery trips even though the assets involved are inexpensive. This is why the oil console belongs in the same monitoring tier as the machine it protects.

Can a refinery programme be built without an outage to install sensors?

Only partially. Hazardous area requirements and access restrictions mean permanent measurement points are best installed during turnarounds. Between outages the programme is limited to equipment that is genuinely accessible, which is rarely the equipment carrying the most risk.

To discuss a tiered reliability programme for a refinery or petrochemical complex, contact Tiaravib at info@tiaravib.com or WhatsApp +62 821-1378-3727.

Refinery and petrochemical reliability is fundamentally an allocation problem before it is a technical one. Tiaravib has supported Indonesian oil, gas, and process industry sites since 1995, working across rotating equipment monitoring, oil analysis, and electrical assessment, and the consistent finding is that the plants which monitor selectively and deliberately outperform those which monitor everything. Standards that inform the work include ISO 17359 for programme design, API standards for turbomachinery and refinery equipment practice, and ISO 20816-3 for vibration severity assessment.