Skip to main content
Bonsai Software
All case studies

One traceable FMECA database for the Utrecht tram network

BAM is taking on the maintenance of the entire Utrecht tram network, the Uithoflijn and the SUNIJ line combined. But the knowledge about failures and failure modes sat scattered across loose, incompatible documents. Together with BAM we built a reproducible data pipeline that brings all those sources together into one automated, traceable FMECA database in the province's IHA format, as the foundation for the maintenance concept.

Client
BAM
Sector
Construction & Infrastructure · Rail
Network
Uithoflijn + SUNIJ line
Target format
IHA-FOR-038, 41 columns
BAM · One traceable FMECA database for the Utrecht tram network

The challenge

An FMECA (Failure Mode, Effects and Criticality Analysis) is the basis under any serious maintenance plan: which components can fail in which ways, what the effect is and what you do about it. For the Utrecht tram network that knowledge existed, but not as a single source. There was literally no place BAM could build a maintenance concept on.

The sources had completely different data structures: the Uithoflijn data in Excel with numbered codes, the SUNIJ data buried in a PDF, heavy-rail rows in ProRail text paths, plus a production FMECA as a PDF. No two sources spoke the same language.

On top of that, everything had to be mapped onto the province's object breakdown structure (OBS), eight levels deep, while that tree itself had gaps. And the target format set strict requirements: 41 columns, failure modes only on the lowest nodes in the tree, and every row traceable to its source and source row.

Our approach

  1. 01

    All sources onto one schema

    A reproducible pipeline reads the Excel and PDF sources and normalises everything onto one schema. Not a hand-merged Excel, but a machine that can build the FMECA from scratch.

  2. 02

    AI for the heavy lifting

    Mapping thousands of heavy-rail rows onto the object tree, deduplicating production rows and recognising synonyms across sources. Where the tree was too coarse, new nodes were added.

  3. 03

    Traceable down to the source row

    Every decision the pipeline makes is recorded in auditable files and every failure mode hangs on a leaf node in the object tree. Every row traces back to its source and source row.

  4. 04

    Corrections survive every rebuild

    BAM's reviews are read back cell by cell and stored durably. The team simply kept working in Excel; add a new source or correction and run the pipeline again.

The result

The database holds 1,067 rows, one hundred percent mapped onto the object tree, without a single unmapped row. The object tree itself grew from 133 to 222 nodes, all 152 maintenance activities are linked to planning codes, and the number of rows with a concrete maintenance task rose from 590 to 775.

The project was delivered for roughly half the estimated staffing cost. The collaboration ran in short cycles: review rounds often several times a week, feedback processed within a day or two. BAM kept control of the content, Bonsai guarded the technical execution.

Want this too?

Book a call now