Predictive Maintenance — 1SP Foundation¶


Project	PRJ-03: Predictive Maintenance Platform
Site	Cleveland Works — 1SP (BOF, LMF, Caster)
Champion	Jamie Betts (Maintenance Manager)
Key SMEs	John Messi (ops/reliability), Brian Thompson (Division Mgr)
Vooban Lead	Alexis Fortin-Côté
IE Lead	Erico Lemos
Duration	8 weeks
Status	Phase 1 Active

Charter Snapshot¶

An 8-week Proof of Value delivering three outcomes:

Multi-asset data assessment across 1SP — predictive readiness scorecard for every asset reachable through ingested data subsystems
Predictive deep-dive on three critical assets — BOF bag house ID fans (primary), BOF scrubbing system (secondary), Crane 300 (tertiary)
Data foundation design (IE-led) — L0-L1 to cloud architecture, asset onboarding playbook, cost trajectory for scaling

Asset	Role	Why
BOF Bag House ID Fans	Primary	Richest data, clearest degradation, environmental compliance threshold
BOF Scrubbing System	Secondary	3-week degradation cycle, -5 heats/day impact, quantifiable
Crane 300	Tertiary	Single point of failure, near-zero instrumentation — proves methodology on hardest case

Phase	Weeks	Status	Key Activities
Data Ingestion & Readiness	1–2	Not started	Ingest Pi/Tabware/Wonderware, build data quality scorecard, identify failure events
Predictive Deep-Dive	3–5	Not started	Bag house degradation model, scrubbing cycle prediction, Crane 300 MCSA analysis
Dashboard Prototype	6–7	Not started	1SP health dashboard with live analytics on deep-dive assets
Final Demo & Deliverables	8	Not started	Leadership demo, readiness report, Phase 2 recommendation

Phase	Weeks	Status	Key Activities
L0-L1 Discovery	1–2	Not started	Catalog all 1SP data sources, assess connectivity
Foundation Architecture	3–5	Not started	Integration platform design, asset onboarding playbook
Blueprint & Cost Model	6–8	Not started	Data foundation blueprint, Phase 2/3 cost trajectory

Metric	Baseline	Target	Current
Maintenance ratio (reactive/planned)	70/30	Shift toward planned	—
Bag house detection lead time	None (reactive)	Days before failure	—
Scrubbing degradation prediction	None	Predict 3-week cycle	—
Assets with data scorecard	0	All 1SP assets	—
Heats/day at 1SP	28 target (often missed)	Reduce unplanned losses	—

#	Decision	Owner	Deadline	Status
1	Data access mechanism (API vs. CSV export)	CLF IT / Andrew Mullen	Week 0	Pending
2	Starnet vibration data access for 1SP cranes	CLF IT	Week 1	Pending
3	Phase 2 go/no-go	Chad Asgaard / Palmer	Week 8	Future

What Cleveland-Cliffs owns. These items gate project progress — delays here delay delivery.

#	Item	Owner	Deadline	Status
D1	Pi historian export — differential pressures, motor currents, temperatures, flows (all 1SP)	Phil Thorman (OT)	Week 1	Pending
D2	Tabware CMMS export — work orders, PM history, failure records for all 1SP assets	IT / Maintenance	Week 1	Pending
D3	Substation motor current / power draw data for Crane 300 area	Phil Thorman	Week 1	Pending
D4	Third-party vibration data access (Starnet?) for 1SP crane sensors	Vendor coordination / Phil Thorman	Week 1	Pending
D5	Viz route-based vibration data for 1SP area equipment	Vendor export	Week 2	Pending
D6	Crane 300 paper maintenance binders	Maintenance team	Week 2	Pending
D7	Environmental logs — compliance records, emission events, bag house inspections	Environmental / Ops	Week 2	Pending
D8	Production data — heats/day, vessel availability, downtime events	Operations / L2	Week 2	Pending
D9	Thermography reports (subcontractor)	Maintenance / Hoyong Hwang	Week 2	Pending
D10	Oil analysis reports (subcontractor)	Maintenance / Hoyong Hwang	Week 2	Pending

Item	Owner	When
Sensor tag documentation (tag → physical measurement mapping)	Phil Thorman	Weeks 1–2
Asset hierarchy confirmation in Tabware for all 1SP critical assets	Maintenance / IT	Week 1
Failure mode context — what breaks, how often, warning signs	John Messi, Evan, Richie	Weeks 1–3
Scrubbing system degradation cycle specifics	John Messi	Weeks 1–2

Stubna confirmed this committee during the Apr 20 tie-out meeting. Two mandates:

Validate that AI-generated work orders are relevant — not noise added to an already full maintenance queue
Integrate work order outputs into actual work processes so they don't become "another report nobody reads"

Role	Person	Why
Chair / maintenance gatekeeper	Jamie Betts	Schedules work, allocates resources — Stubna's pick
Mechanical SME	Evan	Bag house mechanical detail, judges technical relevance
Electrical SME	Richie	Motor data interpretation, validates electrical findings
Operations rep	TBD	Junkin: "having somebody from the operations team is important" — integrate into shift reports

Stubna committed to sending committee names

Confirmed during Apr 20 call. Operations representative still TBD — Junkin flagged this as critical to avoid "just another report."

Role	Person	Accountability
Executive sponsor	John Stubna	Go/no-go decisions, executive updates
AI/Innovation coordinator	Andrew Mullen	Cross-site alignment, progress tracking, EAM coordination
1SP Division Manager	Brian Thompson	Operational constraints, scheduling
1SP Maintenance Manager	Jamie Betts	Work order validation lead, maintenance planning
Primary ops / reliability	John Messi (18 yrs)	Day-to-day coordination, failure context, scrubbing expertise
OT / Process Automation	Phil Thorman	Pi data, sensor tag mapping, Starnet coordination
Mechanical SME	Evan	Bag house mechanical detail, crane systems
Electrical SME	Richie	Motor data interpretation, substation context
Process data	Hoyong Hwang	Pi data, production data, daily operation reports
Day-to-day coordinator	TBD (proposed: John Messi)	Data export logistics, scheduling

Day-to-day coordinator not yet confirmed

Charter proposes John Messi. Needs formal confirmation before kickoff.

Work order integration has two parallel tracks:

Track A — Tabware (current): SQL read access confirmed. Write-back feasibility being assessed. System is old tech — integration may be limited.
Track B — EAM 2.0: Cleveland Works targeted as first site, fall 2026. Modern API available. Vooban has access to API documentation and is pre-building integration.

If Tabware write-back proves too costly, recommendation shifts to a dedicated AI interface or direct EAM 2.0 integration once migration completes.

#	Blocker	Since	Owner	Impact
	No active blockers

Date	Update
Apr 20	Tie-out meeting with Stubna, Macino, Junkin, Hwang. Stubna confirmed Jamie Betts as work order validation lead. Macino added Evan + Richie. Junkin pushed for ops rep on committee. Stubna committed to sending committee names same day. EAM/Tabware dual-track discussed. Hwang confirmed Pi data + subcontractor reports (thermography, oil analysis) as key data sources. Team "ready to get started."
Mar 24	Chad Asgaard greenlit PdM PoV at corporate readout. "Get started right away."

Touchpoint	Timing	Audience
Weekly status	Every Friday	Project team
Week 2 checkpoint	End of Week 2	Data reality determines modeling approach
Week 4 midpoint	Mid-engagement	Leadership preview
Week 8 final demo	End of POV	Full leadership team