Tier-1 Global EPC Contractor, Malaysia Greenfield Gas Processing Plant
10,247 tags extracted across 600 P&IDs. 99.5% measured accuracy. Independent QA audit zero rejections.
10,247
Tags Extracted
600
P&IDs Processed
99.5%
Measured Accuracy
Zero
QA Rejections
A tier-1 global EPC contractor delivering a greenfield gas processing plant in Malaysia needed structured engineering data extracted from 600 P&IDs across eight process units. The contractor's internal estimate was 9-12 months to build the tag register manually at 95% accuracy, with significant residual risk on safety-critical fields. The owner-operator handover specification required CFIHOS 2.0-compliant data delivery within an 18-month window from mechanical completion, with contractual liquidated damages for late handover.
600 P&IDs spanning eight process units (gas sweetening, dehydration, NGL recovery, LNG liquefaction, regasification, send-out compression, utilities, flare), each with different licensor tag conventions and revision histories.
Internal team estimated 9-12 months of manual tag extraction at 95% accuracy. Safety-critical fields (SIL classification, instrument loop SIF reference, design pressure / temperature on IBR-scoped equipment) required 99.5%+ accuracy for owner acceptance.
Owner-operator handover specification required CFIHOS 2.0-compliant structured data with ISO 15926 Part 4 RDL class mapping, RDL relationship class population for asset-to-document linkages, and asset hierarchy population from enterprise through control-module levels.
Aggressive 18-month handover window from mechanical completion with contractual liquidated damages tied to a fixed milestone date. Missing the milestone meant deferred revenue plus penalty exposure.
Ingested all 600 P&IDs in PDF (modern revision) and DWG (legacy reference baseline) format. Pre-processing handled multi-sheet drawing continuation, large-format A0 / A1 sheets, and licensor-specific border block conventions.
Our extraction models parsed every instrument tag using ISA 5.1 native symbology recognition (instrument bubble, succeeding-letter function, loop number), every line number using project Piping Material Specification convention, every equipment tag using ASME equipment-numbering pattern, and every valve specification including type / size / class / end connection.
Cross-validated extracted P&ID data against 1,840 instrument datasheets, 320 equipment datasheets, the project line list, and the C&E matrix. Surfaced 412 cross-document inconsistencies across the 8 process units before owner-operator submission.
Domain engineers reviewed every safety-critical field: SIL 2 / SIL 3 classification on Safety Instrumented Functions, design pressure and temperature on IBR-scoped pressure equipment, instrument loop SIF reference for IEC 61511 compliance, hazardous area classification per IEC 60079 boundary.
Independent QA audit conducted on a randomised 5% sample of extracted tags. Measured field-level accuracy: 99.5% across the full sample with zero safety-critical errors.
Output delivered in CFIHOS 2.0 XML and ISO 15926 Part 8 RDF formats for ingestion into the operator's AVEVA AIM platform. Pre-certified data feed configured for SAP PM S/4HANA functional location and equipment master loading at commissioning.
10,247 instrument and equipment tags extracted across 600 P&IDs at 99.5% measured field-level accuracy.
412 cross-document inconsistencies (P&ID vs instrument index, P&ID vs line list, P&ID vs C&E matrix) surfaced and resolved before owner-operator submission.
CFIHOS 2.0-compliant structured data delivered, ingested cleanly into AVEVA AIM platform with no remapping required.
Engagement completed in 11 weeks against the contractor's 9-12 month manual baseline estimate.
Zero safety-critical errors in the independent QA audit sample.
Contractor handover milestone hit with 14 weeks of buffer against the contractual liquidated damages deadline.
“We had run internal trials on AI extraction before. The accuracy was not where we needed it for owner-operator handover. Pathnovo delivered measured 99.5% accuracy across 10,247 tags. That is the first time we have seen production-grade results on an engagement of this scale.”
EDEngineering Director
Tier-1 Global EPC Contractor, Malaysia Greenfield Gas Processing Plant
Documents Processed
P&IDs (600)
Instrument Datasheets (1,840)
Equipment Datasheets (320)
Line List
C&E Matrix
Capabilities Used
Target System
AVEVA AIM + SAP PM S/4HANA
Measured field-level accuracy across the full 10,247-tag extracted set, verified by an independent QA audit on a randomised 5% sample. Each sampled tag was manually re-extracted by an independent reviewer and compared field by field against the Pathnovo output. The 99.5% figure is the verified rate, not a marketing benchmark. Crucially, zero safety-critical errors (SIL classification, design pressure / temperature, hazardous area boundary) appeared in the sample.
CFIHOS (Capital Facilities Information Handover Specification) 2.0 is the global tier-1 owner-operator standard for engineering data handover. It defines RDL (Reference Data Library) class mapping, asset hierarchy structure, and the asset-to-document relationship model. Without CFIHOS compliance, the operator's information system (AVEVA AIM in this case) cannot ingest the data without manual remapping, typically costing 3 to 6 months of post-handover reconciliation. Pathnovo's CFIHOS 2.0 XML output ingested cleanly with no remapping required.
Pre-processing handled both formats natively: DWG for the legacy revision baseline (1990s and 2000s commissioning baseline drawings preserved as reference) and PDF for the modern revision set issued during the current project. Extraction ran on the modern PDFs as the authoritative source; the DWG baseline was used for cross-validation where revision history was ambiguous. Multi-format ingestion eliminated the conventional practice of converting all DWGs to PDFs as a pre-processing step.
P&ID-to-instrument-index conflicts (tag exists on P&ID but missing from index, or vice versa), P&ID-to-line-list mismatches (line number convention drift between disciplines), and P&ID-to-C&E-matrix conflicts (SIS-relevant tag absent from C&E matrix, SIL classification disagreement between datasheet and C&E entry). All 412 were resolved by the EPC engineering team before owner-operator submission, eliminating the standard handover-QA rejection cycle.
The contractor faced contractual liquidated damages tied to a fixed handover milestone. The 11-week internal estimate was for manual extraction at 95% accuracy with no buffer for QA-driven rework. Pathnovo's 11-week delivery at measured 99.5% accuracy gave the contractor 14 weeks of buffer to absorb downstream commissioning slippage, owner-operator review cycles, and final handover QA without LD exposure. On a tier-1 EPC contract, 14 weeks of buffer at typical daily LD rates is material to project profitability.
Have a similar project? Let's talk.
Send us 10 documents from your current project. We extract, reconcile, and show you exactly what we find in 48 hours, before any contract.
If the accuracy isn't what we promised, you owe us nothing.
Connect with Pathnovo to discuss your engineering document intelligence needs.
Email: hello@pathnovo.com
Send us a message, and we'll get back to you shortly.
You can also stay connected through our official social media channels.
Our Offices
Bangalore Office
Unit 101, OXFORD TOWERS 139, Old HAL Airport Rd, Kodihalli, Bengaluru, Karnataka 560008