Published September 9, 2026 | Version v1

Predictive Maintenance Metrics in the Minimisation of Non-Routine Flaring from Rotating Machinery Failure: A Review of Condition Monitoring, Prognostics and the Conditional Chain from Early Detection to Avoided Emergency Pressure Relief

  • 1. Independent Researcher, Nigeria
  • 2. Chevron Nigeria Limited, Nigeria
  • 3. Waltersmith Refining and Petrochemical Company Ltd, Lagos
  • 4. NNPC Energy Services Ltd (EnServ), Abuja, Nigeria

Description

Abstract : Non-routine flaring in a liquefaction facility is predominantly a consequence of equipment failure, and failure in rotating machinery is predominantly preceded by a detectable degradation signature. The two propositions are individually well established and are rarely connected, with the result that condition monitoring programmes are justified on maintenance cost and availability grounds while the flaring consequence, frequently the larger cost, is attributed to a separate account. This review examines the relationship between condition monitoring performance and non-routine flaring, drawing together the diagnostics and prognostics literature, the maintenance optimisation literature, and the emissions measurement literature. The scope is restricted to rotating machinery, principally the refrigerant compressors and their drivers. Instrumentation failures and control system faults are also a material cause of non-routine flaring, but they are outside this scope, because their degradation signatures and detection routes differ and the argument developed here does not transfer to them. No data are reported. The paper is analytical throughout, and the conditional probabilities used to illustrate the chain are stipulated for exposition rather than measured or estimated; they should not be read as figures characterising any facility. Five arguments are developed. First, the causal chain from degradation signature to avoided flare event runs through four conditional steps, and the probability that a detected anomaly prevents a flare event is the product of detection probability, lead time adequacy, intervention opportunity and intervention effectiveness, each separately measurable and none ordinarily measured. Second, lead time rather than detection sensitivity is the binding constraint in a liquefaction context, because the intervention for most machinery degradation requires a load reduction or an outage and the opportunity for either is determined by the operating plan rather than the maintenance schedule. Third, the rapid development of data-driven prognostics has improved the estimation of remaining useful life without addressing the opportunity constraint, so that its practical contribution in this setting is smaller than its methodological maturity suggests. Fourth, the standard predictive maintenance metrics measure the maintenance function and not the outcome the programme is claimed to produce. Fifth, the attribution of an avoided flare event is counterfactual and therefore contestable, and a measurement approach based on the observable chain avoids the difficulty entirely. The framework is offered as a measurement proposal requiring empirical validation, not as a validated instrument.

Files

11GJETR.pdf

Files (350.4 kB)

Name Size
md5:548a8efd5139504c4e054d5a3b3bf046
350.4 kB Preview Download