Middletown Works¶
This appendix is part of the AI Innovation Sprint engagement with Cleveland-Cliffs Inc. It covers the site-specific roadmap, project sequencing, investment rationale, and champion accountability for Middletown Works. It is designed to be read as a standalone document or alongside the main report.
| Term | Definition |
|---|---|
| H1 | Horizon 1: Bridge the Gap (Months 0-6) — quick wins, data integration, immediate pain relief |
| H2 | Horizon 2: Build the Foundation (Months 7-12) — production-grade systems, cross-process linkage |
| H3 | Horizon 3: Predict & Optimize (Months 13-24) — enterprise optimization, predictive intelligence |
| PRJ-XX | Corporate project ID — cross-site programs that bundle site-level projects. See Chapter 5 of the main report for full detail. |
Site Summary¶
Middletown is the best steel shop in Cleveland-Cliffs. Near-zero unplanned turnarounds (versus 5-12 at other sites), less than 1% scrap rate (versus over 10% at Cleveland), and R&D describes an "order of magnitude" operational difference. This is not a problem child — it is a benchmark. The roadmap for Middletown captures and scales what already works, closes the gaps that competent individuals currently paper over with spreadsheets and tribal knowledge, and builds the finishing-line intelligence that only Middletown can prove. Middletown has CLF's longest finishing chain — vacuum degasser through caster, hot strip mill, pickling, cold mill, and five coating technologies — which makes it the natural proving ground for through-process quality traceability. One hard constraint shapes everything: Dave Reinhold has zero discretionary AI budget. Every project must self-fund or it does not happen.
The First 90 Days (Quick Wins)¶
1. Procurement & Inventory Intelligence (MDT-P04) — The self-funding starter. Dave Reinhold endorsed this at the Day 5 readout as the entry point that generates savings to fund everything else. Middletown carries $104M in MRO inventory ($150M total) with 32,000 parts, 10% estimated duplicates, and a procurement process where buyers manually hunt across catalogs. Sean is the champion. Phase 1 is semantic deduplication of the item master and automated reorder point optimization — a proven recipe that delivered at other CLF sites. Target: 3-5% inventory reduction ($3-5M freed capital) within 90 days, generating the savings cascade that unlocks subsequent projects.
2. Ops-Maintenance Data Integration (MDT-P09) — Foundation layer. Steve Longbottom (RIT) and John Houston (maintenance technology) are the corrected champions — this was clarified at the Day 5 readout. Middletown's version of the universal CLF pattern: Teams/SWAMI (the legacy Armco-era CMMS with no vendor support) holds maintenance data, while production tracking lives in separate systems. Three independent stakeholders validated this as "the biggest problem facing the plant" — Dave on Day 1, Dean on Day 2, Brian Benning on Day 3. Different surface than Cleveland's Tabware/SQL gap, same underlying fracture.
3. QA Knowledge & Investigation (MDT-P03) — Bounded quality quick win. Chuck and the quality team spend days tracing defects backward through 6+ process steps when a customer claim arrives. The SAS-based quality reporting pipeline generates static PDFs and the main programmer retired. Replacing the SAS pipeline with modern SPC tooling and automated drift detection is an urgent, bounded deliverable that Chuck can champion. This is the "laser strike" that Steve Palmer's criteria demand: cross-site scalable, quick-ROI, concrete business case.
Project Sequence¶
| ID | Project | Horizon | Corporate PRJ | Value ($/yr) | Champion |
|---|---|---|---|---|---|
| MDT-P04 | Procurement & Inventory Intelligence | H1 | PRJ-06 | $3-8M | Dave Reinhold, Sean |
| MDT-P09 | Ops-Maintenance Data Integration | H1 | PRJ-01 | $3-8M | Steve Longbottom, John Houston |
| MDT-P03 | QA Knowledge & Investigation | H1 | PRJ-04 | $1.5-5M | Chuck, Alex |
| MDT-P05 | Fleet Maintenance & Copilot | H1-H2 | PRJ-03/06 | $1-6M | Dave, West |
| MDT-P07 | Safety Incident Analytics | H1 | site-specific | low direct $ | Dave |
| MDT-P16 | Process Control Knowledge / Virtual SME | H1-H2 | PRJ-09 | $1-4M + risk | Brian Benning |
| MDT-P15 | Cross-Site Caster Reliability Analytics | H1 | PRJ-01 adj. | $3-8M | Matt |
| MDT-P01 | Surface Inspection Enhancement | H2 | PRJ-04 | $9-27M | Chuck |
| MDT-P02 | Through-Process Quality & Traceability | H2 | PRJ-04 | $11-29M | Chuck, Chris |
| MDT-P06 | Intra-Plant Coil Logistics | H2 | PRJ-07 | $2-5M | Chris, West |
| MDT-P08 | BF Optimization & Raw Material Intelligence | H2 | PRJ-05 | $8-25M | Palmer, Matt |
| MDT-P10 | Finishing Line Scheduling | H2 | PRJ-02 | $3-10M | TBD |
| MDT-P11 | Steelmaking Process Optimization | H2 | PRJ-08 | $4-13M | Matt, R&D |
| MDT-P13 | HSM Rolling Model Replacement | H2 | site-specific | $3-10M | Brian Benning |
| MDT-P14 | Cobble Prediction & HSM Process Risk | H2 | PRJ-05 | $2-8M | Matt, R&D |
| MDT-P12 | Energy & Utility Optimization | H2 | site-specific | $2-5M | TBD |
Horizon 1: Bridge the Gap (Months 0-6)¶
Seven projects can start immediately because Middletown's operational maturity means fewer hard dependencies. MDT-P04 (procurement) is first because it self-funds. MDT-P09 (ops-maint integration) builds the data foundation. MDT-P03 (QA knowledge) delivers quick quality wins. MDT-P16 (Virtual SME) was leadership's number one priority at the Day 5 readout — Paul expanded the scope in real-time to cover L0/L1/L2 per department, and Brian Benning is the champion. Scope management is critical here: the temptation to make Virtual SME everything for everyone must be resisted in favor of a bounded first deployment (one department, one process area).
MDT-P07 (safety analytics) is Dave's trust-building move — he proposed it himself, it demonstrates engagement with the data, and it costs almost nothing.
Horizon 2: Build the Foundation (Months 7-12)¶
This is where Middletown's unique value emerges. MDT-P01 (surface inspection) is Palmer's explicit priority: Ametek cameras on 4+ lines run at approximately 60% accuracy on critical defect types. Retraining the classifier on lab-confirmed samples is a bounded ML problem with existing hardware and measurable accuracy improvement. MDT-P02 (through-process quality) leverages the longest finishing chain in CLF to build end-to-end genealogy linking heat to slab to coil to finished product. When a defect is found at the coating line, the system traces backward through every step automatically instead of requiring manual investigation.
MDT-P06 (coil logistics) is Palmer's number one cross-site priority: 40 trips per day of coil movement, 2 hours per day of manual planning, and Chris and West as champions. MDT-P08 (BF optimization) includes burden mix optimization starting at IH7 — the best data, as identified by R&D — and BF stove optimization, which Palmer explicitly named.
Horizon 3: Predict & Optimize (Months 13-24)¶
Finishing line scheduling (MDT-P10) becomes feasible only after through-process traceability and quality data are in place. Five or more finishing lines competing for cold-rolled substrate from one 5-stand mill is a scheduling problem unique to Middletown's process chain depth.
Investment & Return¶
Middletown-Relevant Corporate Projects¶
| Project | Middletown Role | Middletown Phases | Project Investment | Middletown Value |
|---|---|---|---|---|
| PRJ-06 Maintenance Workflow | Entry site | Ph1 ($612K procurement) | $2.48M | $2-19M/yr |
| PRJ-04 Quality & Yield | Entry site | Ph1 ($756K Ametek+SPC), Ph2 ($1,260K quality linkage) | $2.70M | $15-43M (portfolio, MDT entry) |
| PRJ-07 Logistics | Co-entry site | Ph1 (coil planning, shared with BH) | $2.59M | $2-5M/yr (MDT portion) |
| PRJ-09 Knowledge Capture | Ph2 Virtual SME | Ph2 ($1,218K framework) | $2.35M | $1-5M/yr |
| PRJ-01 Ops-Maint Integration | Ph2 extension | Ph2 (IBM Mainframe adapter) | $1.84M | $2-5M/yr |
| PRJ-05 Cobble & Process Risk | Ph1 MDT stoves + Ph2 cobble | Ph1 + Ph2 (shared) | $2.63M | Portion of $8-35M |
H1 investment at Middletown: ~$612K (PRJ-06 Phase 1 — procurement fast-track). PRJ-04 and PRJ-07 Phase 1 scopes also touch Middletown.
Key value drivers: - Self-funding cascade — procurement savings ($3-8M) fund quality projects, quality savings fund process optimization. No discretionary budget means every dollar must trace back to a previous project's ROI. - Through-process quality — Middletown's 1% quality loss is double the historical 0.5%. The Ametek classifier at 60% accuracy is the most concrete "laser strike" in the engagement. $2.70M investment against $15-43M portfolio-wide annual value (Middletown as entry site captures primary share). - Benchmark, not problem child — Middletown projects are about capturing and scaling what works, not fixing what's broken
Champions & Accountability¶
| Project | Champion | Role | Why This Person |
|---|---|---|---|
| MDT-P04 | Sean | Inventory Management | Owns the 32K-part item master, understands duplicate patterns |
| MDT-P09 | Steve Longbottom, John Houston | RIT + Maintenance Technology | Corrected at Day 5 readout — they own the system integration layer |
| MDT-P01 | Chuck | Quality Manager | Ametek accuracy is his daily pain, automotive claims his exposure |
| MDT-P02 | Chuck + Chris | Quality + Finishing/Automation | Chris controls the finishing line data, Chuck controls quality systems |
| MDT-P06 | Chris + West | Finishing/Automation + Trucks | Chris sees the coil flow from above, West's team moves them |
| MDT-P16 | Brian Benning | Process Control (27 yrs) | Already building knowledge tools, expanded scope at readout, strongest grassroots champion |
| MDT-P08 | Matt + Eric Bridge | R&D | Cross-site BF knowledge, already building BOF endpoint model |
Executive sponsor: Dave Reinhold. Protective of his people, budget-constrained, but will support projects that make Middletown look good relative to its peers. The key is framing every project as "help you do what you already do better" — not "fix what's broken."
How This Site Feeds the Corporate Story¶
Middletown is the entry site for two of Palmer's four named priorities:
-
PRJ-04 (Through-Process Quality & Yield) — Middletown has the longest finishing chain, the most coating technologies, and the Ametek surface inspection cameras Palmer explicitly wants improved. The classifier retraining pattern proven here scales to every CLF site with vision systems. Surface inspection is Palmer's top technical priority.
-
PRJ-07 (Intra-Plant Logistics) — Palmer named coil logistics as his number one cross-site priority. Middletown's 40 trips/day coil movement problem is well-characterized, and the coil tracking system exists. The pattern proven here directly informs Burns Harbor's much larger shipping velocity challenge.
Middletown also gives the corporate story its sharpest proof point. When the presentation shows that CLF's best steel shop still has a 60% defect classification accuracy rate, a retired SAS programmer blocking quality reporting, and $104M in MRO inventory with 10% duplicates, the message is clear: these are systemic problems that even operational excellence cannot compensate for. The information flow gap exists at every site — it is just better masked at Middletown by competent individuals who will eventually retire.
The Day 5 readout revealed a gap between what site leadership endorsed (Virtual SME as number one, Procurement as number two) and what Palmer wants (coil logistics, surface inspection). The corporate presentation bridges this by showing that both perspectives are correct: site leadership identified the enabling projects (Virtual SME captures knowledge before it walks out the door), while Palmer identified the value-delivery projects (logistics and quality generate measurable ROI). The roadmap sequences them so the enabling work feeds the value delivery.