The challenge
The organisation faced a strategic maintenance problem as much as an operational one. A major petrochemical complex with 3,100 critical rotating and static assets across six production units was experiencing 21 unplanned production shutdowns annually — each averaging 19 hours at a per-hour cost exceeding $2.3M. The $340M annual fixed-interval maintenance programme was generating increasing emergency interventions despite the investment. Operations leadership needed a path to condition-based maintenance that would reduce unplanned shutdowns by 70% and total maintenance cost by 20% within three years — and the data to build that capability already existed in the facility's historian systems.
Discovery
A 90-day asset criticality audit revealed the strategic gap at the heart of the operation: the sensor historian contained 8 years of data with clear precursor signatures for the failure modes responsible for historic shutdowns — yet no systematic analytical programme had ever been applied to it. The maintenance organisation was spending $340M annually on time-based schedules while the data needed to move to condition-based maintenance was already sitting in their systems, unanalysed.
Our approach
Diagnostic and data audit: completed a 90-day asset criticality assessment and data quality review across all 3,100 assets and 14 historian and control system data sources to establish the analytical foundation
Solution architecture and integration design: designed the industrial intelligence platform architecture, integrating with the OSIsoft PI historian, Honeywell DCS, and SAP PM maintenance system through OT connectors without process disruption
Failure model development: built asset-specific failure prediction models for 312 critical rotating equipment assets across 47 failure mode categories, trained on 8 years of historical sensor, maintenance, and failure data
Condition-based maintenance programme implementation: designed and delivered the condition-based maintenance recommendation programme, replacing time-based schedules for covered asset classes with condition assessments and precision maintenance windows
Operational integration: connected work order generation with SAP PM to automate maintenance scheduling and parts ordering for analytically-triggered interventions
Safety and digital twin programme: designed real-time safety anomaly detection across high-risk process areas and established a digital twin model for 18 critical compressor trains — the assets responsible for 68% of historical unplanned shutdown cost
The Outcome
The engagement delivered a continuous condition intelligence layer across the facility's critical asset population, ingesting 2.4 million data points per hour from sensors, process data, and maintenance history to build a live condition picture of every monitored asset. The system identifies failure signatures weeks before they would manifest as operational problems — providing the maintenance organisation with precision intervention windows that reduce unnecessary maintenance activity while preventing unplanned failures. In the first full operational year, 47 imminent failures were identified in advance, preventing 19 of the 21 historical unplanned shutdown events.
The results
“The financial impact is clear in our numbers — but what matters equally is the confidence. Our maintenance teams now have real intelligence about what their assets need. We are not gambling on time-based schedules any more. We know.”
