Tilden Mine¶
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 Tilden Mine. 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¶
Tilden is not a steel mill, and the roadmap reflects that. It is the only mine in the engagement — an open-pit hematite operation producing approximately 7.7 million tons of pellets annually, using selective flocculation and amine flotation (historically the first mine to produce pellets from both hematite and magnetite ore, 1990-2009; magnetite reserves now exhausted). Tilden's defining constraint is not mechanical throughput but process chemistry: a non-homogeneous ore body that changes composition as the pit deepens, a reagent suite designed in 1974 for different ore, and a concentrator that achieves roughly 70% iron recovery versus a design benchmark of ~75% and an optimized hematite flotation target of 80%. The site spends $50 million per year on chemical reagents. The gap between what is known at the drill hole and what the concentrator does with the ore when it arrives 12 hours later is Tilden's version of the information flow problem that appeared at every site — same thesis, different industry. Tilden produced the strongest readout of the entire engagement. Ryan Korpela's reaction — "really impressed," repeated twice, with no budget objections raised and an explicit request to "move up the priority list" — sets the stage for rapid execution. Chad Asgaard's March 24 directive was clear: concentrator optimization first. Make money.
The First 90 Days (Quick Wins)¶
1. HPGR Knowledge Base and PdM Pilot (TLD-P10) — Month 1 start. Adam Bingham is the strongest grassroots AI champion at any site — already using Copilot independently, nominated this project himself, and has 1,200 pages of unread HPGR manuals ready for ingestion. Phase 1 turns those manuals into a searchable AI knowledge base in weeks. Phase 2 adds feed rate investigation and PdM anomaly detection baseline by Month 3. This is Palmer's knowledge capture theme proven with the best-scoped pilot in the engagement. The HPGR is the testing ground; the methodology scales to every knowledge-dependent asset at every site.
2. Concentrator Desliming & Recovery Optimization (TLD-P01) — ACTIVE. Charter submitted, $312K Phase 1 firm. This is the concentrator flagship, launched per Chad Asgaard's March 24 directive. Charter v3 defines a 7-week Proof of Value ($312K) deploying four integrated AI components around the desliming circuit: (1) desliming optimization model — supervised regression on historical process variables predicting optimal PAA dispersant dose per section, backtested on ore variability events, (2) flotation recovery correlation linking desliming to downstream stability, (3) beaker test vision replacing subjective visual assessment with camera-based standardized measurement (Phase 2), (4) live CRP engine with rule-based bottleneck identification + LLM recommendations drawing on Keith's captured process knowledge (Phase 2). Phase 1 proves the core thesis for $312K with a hard go/no-go gate at Week 7. Phase 2 ($1.3-1.56M indicative) builds the full system including beaker vision, operator advisory dashboard, and operational integration. Phase 3 extends upstream to mine-to-mill optimization (geostatistical stockpile composition modeling, blast pattern optimization). Keith Holmgren is the primary SME; Ryan Korpela is executive sponsor.
3. Mine-to-Dock Logistics Optimization (TLD-P08, Phase 1) — Month 2 start. Kevin spends 3-4 hours every day manually replanning the vessel loading schedule because BCS (the vessel tracking system) data is not integrated with production and rail scheduling. A daily schedule optimizer replacing Kevin's manual process is a bounded, high-visibility quick win. Ryan ranked logistics as his number three priority, and it aligns directly with Palmer's cross-site logistics priority (PRJ-07).
Project Sequence¶
| ID | Project | Horizon | Corporate PRJ | Value ($/yr) | Champion |
|---|---|---|---|---|---|
| TLD-P10 | HPGR Knowledge Base + PdM Pilot | H1 | PRJ-06/03 | $2-5M | Adam Bingham |
| TLD-P09 | Ops-Maintenance Data Integration | H1 | PRJ-01 | $4-9M | Pete Austin, George Harmon |
| TLD-P11 | Maintenance Workflow & Inventory | H1 | PRJ-06 | $3-10M | Adam Bingham, warehouse team |
| TLD-P14 | Safety & Inspection Digitization | H1 | site-specific | $1-4M | Dan Clarendon |
| TLD-P12 | Mining Knowledge Capture / Virtual SME | H1-H2 | PRJ-09 | $2-7M | Adam Bingham, Lynn Casco |
| TLD-P08 | Mine-to-Dock Logistics | H1-H2 | PRJ-07 | $2-6M | Kevin |
| TLD-P01 | Concentrator Feed-Forward & Ore Intel | H1-H2 | PRJ-10 | $8-18M | Todd Davis, Sean Halston |
| TLD-P04 | Mining Fleet PdM & Lifecycle Intel | H1-H3 | PRJ-03 | $9-24M | Pete Austin, Chase Lincoln |
| TLD-P03 | Pellet Plant Quality & Control | H1-H2 | PRJ-04 | $3-9M | Process engineering team |
| TLD-P05 | Fixed Plant PdM & Failure Analytics | H1-H2 | PRJ-03/01 | $4-12M | George Harmon, Gary |
| TLD-P13 | Maintenance Planning & Scheduling | H1-H2 | site-specific | $2-7M | Gary, Pete Austin |
| TLD-P02 | Concentrator Operations & Recovery | H2 | PRJ-04 | $8-21M | Dan McGrath, Sean Halston |
| TLD-P06 | Drill & Blast Intelligence | H2 | site-specific | $1-4M | Jeff Domann |
| TLD-P07 | Mine Operations & Dispatch Intel | H2-H3 | PRJ-07/02 | $6-20M | Tyler Craig, Brad Koski |
| TLD-P15 | Environmental, Utilities & Geotech | H2 | site-specific | $1-4M | Brent |
| TLD-P16 | HR & Administrative Operations | H1 | site-specific | $0.3-1M | Lynn Casco |
Horizon 1: Bridge the Gap (Months 0-6)¶
Tilden has more H1-ready projects than any steel site because its operational domains — mine, concentrator, pellet plant, logistics, maintenance — run in parallel rather than in series. Twelve of sixteen projects are root nodes with no upstream dependencies. The risk is not dependency bottlenecks but capacity spread. The roadmap manages this through staggered starts: five small, bounded deliverables launch in Month 1 (HPGR knowledge base, Modular-ELLIPS integration, ELLIPS inventory cleanup, Take-5 digitization, BLA contract assistant), followed by higher-value projects in Months 2-3 (stockpile model, tire lifecycle model, logistics optimizer, calcium control automation).
Key H1 milestones: - Month 2: HPGR knowledge base searchable. Dispatch auto-correction deployed (eliminating Molly's manual scanning). ELLIPS inventory duplicates flagged. - Month 3: Daily logistics schedule optimizer replacing Kevin's 3-4 hour manual process. Stockpile distribution model prototype showing per-section feed quality predictions. Tire lifecycle model v1. - Month 6: Eleven projects showing measurable results. Foundation data for H2 in place.
Horizon 2: Build the Foundation (Months 7-12)¶
H2 is where the concentrator flagship reaches full power. Phase 2 of TLD-P01 links drill hole assay data and the Vulcan geological model to concentrator parameter prediction — the 12-hour ore transit becomes an advantage when prediction works. TLD-P02 (concentrator operations) adds filter instrumentation across all 42 units (approximately $125K investment) and G2 augmentation for grinding optimization. The realistic target is closing the gap from ~70% toward the 75-80% range achievable with optimized hematite flotation. Each 1% recovery improvement equals roughly 77,000 additional tons at over $100 per ton.
TLD-P04 (fleet PdM) transitions from tire prediction to duty-cycle maintenance models for shovels and truck engines. Pete Austin's insight — that hours are a meaningless maintenance metric for mining equipment, and duty cycle (what the machine was actually doing) is what matters — drives the model design. TLD-P06 (drill and blast) connects Jeff Domann's drill data to Dyno Nobel's auto-density trucks; both ends of the data pipeline exist but the bridge between them does not.
Horizon 3: Predict & Optimize (Months 13-24)¶
The full ore-to-concentrator closed-loop optimization (TLD-P01 Phase 3) is the terminal play: drill data feeds blend optimization, which feeds concentrator parameter adjustment, which feeds reagent optimization, with the model learning from the changing ore body. This is the highest-value AI play at any mining operation in CLF's portfolio. TLD-P04 Phase 4 migrates Pete Austin's Excel-based fleet lifecycle intelligence into a system that generates multi-year CAPEX plans and repair-versus-replace recommendations at each major rebuild decision point.
Investment & Return¶
Tilden-Relevant Corporate Projects¶
| Project | Tilden Role | Tilden Phases | Project Investment | Tilden Value |
|---|---|---|---|---|
| PRJ-10 Process Chemistry | Entry site | Ph1 ($312K firm), Ph2 ($1.3-1.56M indicative), Ph3 ($648K) | $2.26-2.52M | $8-18M/yr |
| PRJ-09 Knowledge Capture | Co-entry site | Ph1 ($684K HPGR pilot, shared with BH) | $2.35M | $1-5M/yr |
| PRJ-03 PdM Platform | Ph2 fleet PdM | Ph2 (Tilden fleet, shared with CLV expansion) | $2.08M | $9-24M/yr (fleet) |
| PRJ-01 Ops-Maint Integration | Ph2 extension | Ph2 (Modular-ELLIPS ETL) | $1.84M | $2-5M/yr |
| PRJ-07 Logistics | Ph3 mine-to-dock | Ph3 (scheduling optimization) | $2.59M | $2-6M/yr |
H1 investment at Tilden: ~$996K (PRJ-10 Phase 1 $312K + PRJ-09 Phase 1 portion). PRJ-10 charter submitted, pending contract execution.
Key value drivers: - Concentrator optimization (PRJ-10) — $8-18M/yr. Strongest single-site business case in the portfolio. $312K PoV proves whether the $1.3-1.56M full system build is justified. Chad's direct directive: "focus high on process optimization." - Fleet lifecycle intelligence (PRJ-03 Ph2) — $9-24M/yr. $70K per tire times 108 tires per year equals $7.5M in tire spend alone. $300K per wheel motor (4th rebuild territory). Pete Austin's institutional knowledge is the single greatest knowledge-loss risk at Tilden. - HPGR knowledge capture (PRJ-09) — Best-scoped pilot across all sites. Adam Bingham already using Copilot. 1,200 pages of unread manuals. The methodology proven here scales to every knowledge-dependent asset.
Champions & Accountability¶
| Project | Champion | Role | Why This Person |
|---|---|---|---|
| TLD-P01 | Todd Davis + Sean Halston | Lead Process Eng + ECSM Concentrator Ops | Own the concentrator data and decision-making process |
| TLD-P04 | Pete Austin | Mine Maint Section Mgr (30 yrs) | Built the Excel fleet lifecycle models, irreplaceable institutional knowledge |
| TLD-P10 | Adam Bingham | HPGR technician / grassroots champion | Already using Copilot independently, nominated the HPGR pilot, best-scoped proposal at any site |
| TLD-P07 | Tyler Craig | Mine operations | Proposed pit flow optimization at readout, Ryan elevated it to number two priority |
| TLD-P08 | Kevin | Train scheduling | Spends 3-4 hrs/day on manual replanning — most direct personal ROI of any champion |
| TLD-P06 | Jeff Domann | Drill & Blast | Owns both sides of the drill-to-blast data gap |
| TLD-P09 | Pete Austin + George Harmon | Mine Maint + Reliability Eng | Cross-functional pair bridging dispatch (Modular) and maintenance (ELLIPS) |
Executive sponsor: Ryan Korpela (Operations Manager). The most enthusiastic site leader in the engagement. Validated all four presented projects, ranked them explicitly (mine-to-mill first, pit optimization second, logistics third, tires and HPGR as stepping stones), and asked to "move up the priority list." No budget objections raised.
How This Site Feeds the Corporate Story¶
Tilden is the entry site for two corporate projects and the proof that the engagement thesis is universal across industries:
-
PRJ-10 (Process Chemistry Optimization) — This is Tilden's flagship and Chad Asgaard's March 24 directive: concentrator optimization first. Scoping and planning are underway — Erico is aligning with Ryan on data access and site coordination. The $50M reagent anchor and 70% iron recovery baseline — with realistic upside to 75-80% through ore-adaptive reagent optimization — create the largest single-site value opportunity in the engagement. The methodology — using upstream data to predict downstream process response — is the same principle as BOF endpoint prediction and burden mix optimization at steel sites, just applied to mineral processing. (Note: CLF's Minnesota taconite operations achieve 90%+ recovery through magnetic separation of magnetite, a fundamentally different process from Tilden's hematite flotation.)
-
PRJ-09 (Knowledge Capture / Virtual SME) — The HPGR pilot at Tilden is the best-scoped entry point for the cross-site knowledge capture program that Palmer explicitly named. Adam Bingham is the ideal grassroots champion: technically literate, already using AI tools, self-motivated, and working on a new asset with 1,200 pages of unread documentation. The methodology proven on the HPGR scales to every knowledge-dependent asset across all five sites.
Tilden also makes the corporate thesis impossible to dismiss as a steel-only problem. When the presentation shows that the same information flow thesis — data exists on both sides of a gap, but nobody connects them — appears in open-pit mining (ore quality to concentrator response) just as it does in integrated steelmaking (operations to maintenance), the argument shifts from "we found site-specific problems" to "we found a systemic pattern." Four sites, two industries, one thesis. That is the headline Palmer has not heard yet.
Ryan Korpela's readout — the strongest of the engagement — gives the corporate story its most compelling proof of site-level buy-in. When an operations manager says "really impressed" twice, validates all four projects, ranks them without hesitation, and asks to accelerate, it demonstrates that the roadmap is not a consulting exercise imposed from above. It was built from the evidence practitioners provided and validated by the leader accountable for results.