From Changeover Variability To Recovered Capacity

How a Steel Manufacturer Unlocked $330K in Annual Throughput from a Single Changeover Improvement and $1.3M Across Four Lines

A North American steel manufacturer operating a capital-intensive pickling line partnered with Arch to improve the performance of high-frequency planned changeovers. In this process manufacturing environment, where up to 180 changeovers occur daily, small inefficiencies compounded into meaningful lost capacity.

Arch applied micro-step visibility and AI-driven root cause analysis to break down the process, isolate sources of variability, and distinguish operator-driven delays from other factors. This analysis identified flattener jog retries as a key source of delay, enabling a ~12-second reduction per changeover.

At scale, this translated to ~35 minutes of recovered production time per day, or ~5 additional coils daily. On a constraint line with an estimated downtime cost of $1,600 per hour, this represents approximately $330,000 in annual throughput opportunity, with total potential exceeding $1.3M when replicated across four similar lines.

HIGHLIGHTS:

35

minutes recovered per day,
enabling ~5 additional coils

~$330K

annual throughput
from a single improvement

~$1.3M

total opportunity
when scaled across four lines

12-second reduction per changeover,
across ~180 daily events

The Challenge

Limited visibility into high-frequency planned downtime constrained performance on a capital-intensive continuous process. On the steel pickling line, operators execute up to 180 coil changeovers per day, manually stitching new coils to maintain flow through cleaning and downstream galvanizing operations. While total changeover time was tracked, the team lacked insight into what drove variation.

No visibility into micro-steps masked root causes: Only total average duration was measured, obscuring which stages introduced delays.
Operator performance gaps were not actionable: Top operators consistently outperformed peers, but best practices were not measurable or repeatable.
Manual, switch-based signals limited insight: The process relied on PLC switch signals rather than continuous parameters, making it difficult to interpret and analyze performance.

With downtime estimated at $1,600 per hour and the pickling line feeding downstream operations, even small inefficiencies compounded quickly and constrained downstream throughput.

The Arch Approach

Arch deployed its Factory Intelligence Platform to digitize and optimize planned changeovers.

Micro-step digitalization created structured visibility: Arch Data Twin reconstructed the full changeover sequence from PLC signals, aligning each step to standard operating procedures with defined targets.
AI root cause analysis identified true bottlenecks: Arch analyzed hundreds of changeovers across operators and shifts, comparing fast and slow executions to isolate the specific steps driving delays.
Multi-level root analysis distinguished operator vs material drivers: AI determined that delays were driven by operator execution rather than material variation, enabling targeted operator skilling improvements instead of unnecessary process adjustments.
Real-time alerts and weekly summaries enabled action: Threshold based alerts flagged extended changeovers in real time, while trend-based summaries highlighted recurring patterns for
continuous improvement.

This approach transformed a manual, opaque process into a measurable system for identifying and acting on high-impact improvements.

Key Outcomes

By combining micro-step visibility with AI-driven root cause analysis, the manufacturer translated small, repeatable improvements into measurable capacity and financial impact.

Identified primary source of delay:

Flattener jog retries were the dominant contributor to extended changeovers, with uncoiler jog retries identified as a secondary improvement opportunity.

Converted seconds into meaningful capacity gains:

Addressing flattener jog retries reduced changeover time by ~12 seconds per event, compounding across ~180 daily changeovers to recover ~35 minutes of production time per day and enable ~5 additional coils without added labor or assets.

Quantified high-impact ROI on a constraint line:

On a pickling line with an estimated downtime cost of $1,600 per hour, this improvement represents approximately $330,000 in annual throughput value from a single issue.

Demonstrated scalability across similar assets:

Replicating this improvement across four comparable lines represents over $1.3M in total annual throughput opportunity.

Delivered multi-level root cause insight with clear actions:

Level 1 (step-level): Identified which phase of the changeover introduced delays.
Level 2 (switch-level): Isolated specific equipment actions, including flattener and uncoiler jog retries.
Level 3 (operator vs material): Determined that variability was driven by operator execution, not material conditions.

Enabled targeted improvement actions:

Operator: Focused training on shifts and scenarios with inconsistent execution.
Material: Ruled out material as a primary driver, avoiding unnecessary process or supplier changes.

Strategic Lessons

Planned downtime is a capacity lever: High-frequency changeovers create continuous opportunities to unlock throughput.
Micro-downtime can rival major downtime in impact: Small delays measured in seconds can exceed the impact of larger, less frequent events when repeated hundreds of times per day.
Process manufacturing can be digitized effectively: Even switch-based inputs can be structured into actionable, step-level intelligence.
AI reveals actionable root causes, not just symptoms: Multi-level analysis distinguishes operator, material, and process drivers to guide the right actions.
Small improvements scale quickly in continuous processes: Seconds saved per event translate into meaningful capacity gains.
Transition points are high-leverage control points: Continuous lines are most vulnerable during changeovers, making them the highest-leverage opportunities for improvement.