How Manufacturing Engineers in Chemical Plants Use Asset Health Data to Optimize Process Reliability

The process engineering data gaps that limit your reliability analysis work are not problems that process engineering alone can solve. When you update a HAZOP node for a centrifugal pump in corrosive chemical service, the failure rate assumption you are validating came from an industry database that reflects a population of pumps across many facilities, many service conditions, and many maintenance programs. It does not reflect your pump, your service, your plant's operating history, or your maintenance execution quality.

When you are scoping a turnaround for your unit, the interval assumptions driving the scope were set on a calendar basis that assumes all assets degrade at the same rate. The pump that has been running in a demanding service with elevated temperature and periodic cavitation events has a different degradation trajectory than the pump in the adjacent benign service. The calendar treats them the same.

When you specify replacement equipment for a critical process service, the vendor MTBF data in the datasheet reflects laboratory test conditions and population averages. Whether that specification assumption holds in your specific process environment is a question that only post-installation operational data can answer, and in most plants that data is not systematically collected.

Continuous asset health monitoring addresses all three gaps. This guide examines each one from a process engineering perspective: the specific data the gap produces, why it matters for your engineering scope, and what the monitoring data adds.

Gap 1: HAZOP and PHA Failure Rate Assumptions for Rotating Equipment

A process hazard analysis is only as good as its input assumptions. For rotating equipment failure modes in chemical process environments, the critical input assumptions are failure rate frequency, severity of consequence if the failure reaches the process, and detection reliability.

Where generic data falls short:

Industry databases aggregate failure data across large populations and multiple decades. They do not distinguish between a centrifugal pump that has been monitored continuously and receives condition-based maintenance interventions and one that receives only scheduled inspection on a calendar basis. They do not distinguish between a specific corrosive service with known seal face materials degradation mechanisms and a general-purpose water service pump. And they do not capture the improvement in effective detection reliability that continuous monitoring provides versus periodic manual inspection.

The practical consequence for PHA quality:

A HAZOP facilitator updating a node for a centrifugal pump transfer from a reactor to a product column in a chlorinated solvent service will find generic pump failure rate data that may differ from the actual plant performance by a factor of two or more, in either direction. If the generic data understates the actual failure rate, the PHA underestimates process risk. If it overstates the actual rate for a well-monitored pump, the PHA may assign unnecessary safeguard requirements or misallocate process risk management resources.

What continuous monitoring provides:

Each monitoring event that is investigated and confirmed as a maintenance finding represents a data point in the plant-specific failure mode database: centrifugal pump, chlorinated service, seal face wear detected via vibration anomaly, confirmed at 34 days lead time before projected failure. Accumulated across multiple pump cycles and equipment classes, this creates the plant-specific failure frequency and detection lead time data that is currently absent from most chemical plant PHAs.

FMEA detection column:

The FMEA detection column lists the safeguard expected to identify the failure mode before it produces the listed consequence. If vibration monitoring is the listed detection method, the FMEA is asserting that continuous monitoring will detect the failure mode in time to intervene. That assertion needs to be validated from operational history, not assumed from vendor literature. A monitoring program that has been running for two or more operating cycles should be able to produce the detection event data that validates or challenges the FMEA assumption.

Protocol for incorporating monitoring data into PHA updates:

At the start of a scheduled PHA revalidation, compile the monitoring alert history for all covered equipment for the period since the previous revalidation. For each confirmed finding, record the failure mode, the equipment class and service, the lead time between alert and the point at which maintenance confirmed the developing fault, and the projected consequence if the fault had reached failure. This record is the plant-specific input to the failure rate and detection column review in the PHA update.

Gap 2: Turnaround Scope Determination Without In-Cycle Degradation Data

Turnaround planning in continuous chemical plants involves three intersecting engineering disciplines: process engineering (which equipment changes affected process performance during the operating cycle), reliability engineering (what are the current condition trends on rotating and static equipment), and operations (what process upsets or exceedances occurred that might indicate equipment-initiated events).

The calendar-based scope methodology handles reliability engineering through an assumed degradation rate: every pump in this service gets bearing inspection at every TAR because the assumed failure interval is shorter than the TAR interval. This assumption is conservative by design, but it is systematically inaccurate for individual assets that degrade at different rates than the assumption.

Two financial consequences of calendar-based scope:

Over-scoping. An asset that has been running without degradation, monitored continuously with stable vibration and temperature trends over the full inter-TAR period, does not need bearing replacement at the scheduled TAR if the condition data shows it has significant remaining useful life. Replacing it anyway wastes the remaining life of the replaced components and the labor cost of the unnecessary replacement. In a large continuous plant turnaround, over-scoped bearing and seal replacements across dozens of rotating assets represent a material avoidable cost.

Under-scoping. An asset that has been degrading faster than the calendar interval assumed, perhaps because of an operating period with elevated load or process fluid contamination, may be approaching failure before the next scheduled TAR. If the calendar-based scope treats it as a routine inspection item rather than a replacement candidate, the asset may fail mid-run and initiate an unplanned process shutdown with full emergency response cost.

What continuous monitoring adds to TAR scope engineering:

A 12 to 18-month continuous vibration and temperature trend for each monitored asset provides the degradation rate data that makes individual scope decisions defensible. An asset showing a stable, improving, or slowly degrading trend over the full inter-TAR period can be carried forward with confidence. An asset showing an accelerating degradation trend that projects failure before the next TAR interval is a scope addition candidate with engineering justification.

The process engineering contribution to TAR scope:

Process parameter data from the operating cycle contains additional scope-relevant information that the maintenance reliability assessment alone may not capture. A pump that operated in a cavitating condition for several weeks during a production exceedance event has a different accumulated stress history than one that operated within design parameters throughout. A compressor that was run at elevated discharge temperature during a cooling water availability event has experienced thermal stress cycles not reflected in its calendar age.

The manufacturing engineer who reviews process operating history in parallel with condition monitoring data during TAR scope engineering is applying both types of information to scope decisions. The result is a scope that reflects what actually happened to each asset during the operating cycle, not what was assumed to happen based on calendar time.

Gap 3: Equipment Specification Without Post-Installation Reliability Feedback

Equipment specification for chemical process service is an iterative engineering process. The first time a pump class is specified for a new service at a chemical plant, the specification is built from vendor data, industry standards, application engineering judgment, and peer plant experience. That specification may or may not hold in actual plant service.

The feedback loop problem:

Without systematic post-installation reliability data collection, the performance of an equipment specification in actual service is known only anecdotally. A maintenance technician knows that the API Plan 53B mechanical seals on the chlorinated solvent transfer pumps fail every 14 months. The reliability engineer knows the last two replacements were triggered by vibration alerts. The process engineer knows the pumps periodically operate above design flow during production campaigns. But none of this information has been systematically collected and analyzed against the original specification assumptions.

The result is that when the next procurement cycle comes and a replacement pump or a new installation in a similar service is specified, the specification does not reflect what was learned from the post-installation operating history of the previous specification. The same inadequate seal design, the same bearing life assumption, or the same missing monitoring provision gets written into the new specification.

Monitoring-readiness as a specification criterion:

The first category of specification improvement enabled by continuous monitoring is the monitoring-readiness provision itself. A pump specified without sensor mounting provisions requires retrofit to add them, which in a classified chemical process area involves pressure boundary management, classified area work procedures, and potentially a management of change review. Specifying the mounting provisions at procurement reduces retrofit cost to near zero.

Monitoring-readiness specification requirements for chemical process rotating equipment:

• Flat machined surface adjacent to each bearing housing, minimum 30mm diameter, perpendicular to bearing center axis within 5 degrees
• Cable routing provisions: drilled and tapped conduit attachment point at junction location accessible without process entry
• Area classification documentation: zone designation for sensor installation location (Zone 1, Zone 2, or equivalent NEC Division classification) included in equipment datasheet
• Process fluid compatibility: for any sensor in contact with process-wetted components, material compatibility certification for the specific process fluid

Using post-installation monitoring data to improve the next specification revision:

Once a monitoring program is operating on an equipment class, the operational data provides specification feedback that no vendor datasheet can supply. If the API Plan 53B seals on the pumps consistently show elevated vibration signatures 6 to 8 months before seal failure, and the condition monitoring data shows the vibration signature is associated with shaft deflection caused by off-design operating points, the specification feedback is clear: the hydraulic curve selection or the impeller clearance tolerance needs revision for this service.

That feedback is the engineering case for revising the specification. It is documented in the monitoring history and defensible in a specification change review.

How These Three Gaps Compound Each Other

Each gap is significant individually. The compounding effect occurs when they interact.

A HAZOP node for a centrifugal pump is updated using generic failure rate data because plant-specific monitoring history is absent. The risk classification is set based on that assumption. The safeguard requirements are sized for that risk level.

If the actual plant failure rate for that pump class in that service is higher than the generic assumption, the safeguard requirements may be insufficient. But without monitoring data, there is no mechanism to identify that the assumption is wrong until a failure event occurs that was not predicted by the PHA risk level.

The turnaround scope for that pump class is set on a calendar basis. If the actual degradation rate is faster than the calendar assumes (as it might be for the same pump operating under conditions that produce the higher-than-assumed failure rate), the under-scoped pump may fail mid-run, generating the failure event that the PHA analysis was designed to prevent.

The replacement pump is specified to the same specification that the original pump was specified to, because no post-installation reliability feedback loop exists to identify the specification gap. The under-performance of the seal design or the bearing life assumption is not captured.

Continuous monitoring interrupts this cycle at each stage: it creates the plant-specific failure mode data for PHA, it provides the degradation rate data for TAR scope, and it generates the specification feedback that improves future procurement.

Building the Plant-Specific Failure Mode Database

The plant-specific failure mode database is not a separate project. It is the structured accumulation of data from normal condition-based maintenance operations.

What to record for each monitoring event that produces a confirmed finding:

• Equipment tag and service description
• Failure mode category (bearing failure, seal failure, impeller wear, rotor unbalance, cavitation, misalignment)
• Monitoring indicator type that detected the event (vibration amplitude, vibration frequency pattern, temperature trend)
• Lead time: days between the first alert threshold crossing and the confirmed maintenance finding
• Consequence severity: what would have happened if the finding had not been addressed (process upset, process shutdown, environmental release, PSM recordable event)
• Contributing factors: any process operating conditions or maintenance history that contributed to the failure mode onset

Over two to three operating cycles, this record produces a failure mode frequency distribution by equipment class and service that is orders of magnitude more relevant to your plant's PHA than any generic database. It also produces the detection reliability data that validates the FMEA detection column assumptions for your specific monitoring application.

The manufacturing engineer who owns process reliability analysis should own this database, or at minimum co-own it with the reliability engineering function. The PHA and HAZOP update cycle is the mechanism for incorporating the data into formal process risk documentation.

The Hidden Factory: Invisible Process Losses in Chemical Operations

Continuous chemical process operations do not have discrete production runs, but they have an equivalent of the hidden factory: process throughput losses from rotating equipment that is running, but running degraded. A centrifugal pump with impeller wear that delivers 15% less flow than design spec does not trigger an alarm, but it reduces yield, extends batch times, and increases energy consumption per unit of output. A compressor with valve wear running at reduced efficiency produces the same pressure but consumes significantly more energy doing it.

This is the chemical process equivalent of the hidden factory: micro-losses and degraded throughput that are invisible without continuous equipment performance monitoring. These are not maintenance failures by the traditional definition. The equipment is operating. But the Manufacturing Engineer's OEE equivalent, process availability multiplied by throughput rate multiplied by quality yield, is below where it should be, and the cause is invisible without continuous equipment health monitoring that surfaces performance degradation before it reaches the threshold of a maintenance event.

Correlating machine health signatures, vibration trends, temperature, power draw, with process throughput and yield data gives the Manufacturing Engineer the objective record to see where the process is bleeding efficiency and why.

Finger-Pointing Between Maintenance and Process Engineering

In continuous chemical manufacturing, when a process unit underperforms, lower yield, higher energy consumption, off-spec product, the question of whether it is a maintenance issue (degraded rotating equipment) or a process issue (wrong operating parameters, feed variation, catalyst state) is not always straightforward. Without objective machine health data, the investigation defaults to qualitative debate between the maintenance team and the process engineering team.

Continuous machine health monitoring provides the sensor record that separates the two explanations: if the centrifugal pump handling the critical process stream shows bearing wear and reduced hydraulic efficiency, the throughput loss is a maintenance ownership. If the pump is mechanically healthy and the process flow is below target, the investigation moves to process parameters. The Manufacturing Engineer gets the data they need for a root cause analysis rather than a cross-functional blame session.

Process Instability Produces Off-Spec Product Before Equipment Fails

In chemical manufacturing, equipment degradation affects product quality before it produces a failure event. A reactor agitator with bearing wear produces inconsistent mixing that leads to batch uniformity issues. A heat exchanger with fouled tubes produces temperature excursions that affect reaction yield. A compressor with valve wear producing pressure fluctuations creates inconsistent process conditions.

A batch of off-spec chemical product is not just a quality failure, it is a disposal problem, a regulatory documentation problem, and potentially a customer consequence problem. The Manufacturing Engineer who receives quality data showing a batch deviation needs to understand whether the root cause is equipment health, process parameters, or raw material variation. Machine health data correlated with process and quality data is what makes that RCA possible.