Every power plant that has tried to run SPC on Excel has quit within a year. Every plant that has tried to run it directly out of PI ProcessBook or the DCS trend viewer has ended up with three charts nobody looks at anymore. The problem is not the analysts, the tools, or the intent — it's the arithmetic. A single 500 MW coal unit has 3,000 to 6,000 tags worth trending. A combined-cycle block can have 10,000. Doing SPC manually on that data means somebody calculating control limits from a rolling window, refreshing charts on a schedule, sorting real signals from load transients, and routing findings to work orders — for every tag, every shift, forever. It doesn't scale. What plants actually need is software built around the way generation data behaves: hundreds of tags streaming in from PI or AVEVA, load-band-aware baselines, calibrated rule sets that don't drown operators in false alarms, and the ability to close the loop from fired rule to work order without a human retyping anything. That is the difference between SPC as a spreadsheet exercise and SPC as a live control layer. iFactory is built for exactly that scale.
iFactory Automated SPC for Generation
SPC That Scales From 4 Units to 40,000 Tags — Without a Team of Statisticians
Historian-native, load-aware, and closed-loop. What automated SPC software must do to survive a power plant's tag count and shift schedule.
6,000+
tags on a typical 500 MW unit
10,000+
on a combined-cycle block
3
shifts, 24×7 monitoring
Zero
tolerance for manual retyping
Why Manual SPC Fails at Plant Scale
Manual SPC on Excel or in PI ProcessBook works for a proof of concept on ten tags. It breaks somewhere between 100 and 1,000 tags — long before the plant is fully covered. This is what actually goes wrong.
Refresh lag
Excel exports and manual chart refreshes lag hours behind the process. By the time the chart shows the drift, the drift has been running for a shift or two.
Static baselines
Control limits calculated once and never recalculated get stale. Load range changes, coal quality shifts, a new heater comes online — the old sigma stops being the truth.
Load transient noise
A single sigma across all load conditions fires on every ramp. Operators see three false alarms per shift and stop reading the charts within weeks.
Findings dead-end
A fired rule that reaches a chart, but not a work order, changes nothing. Manual SPC almost always dead-ends at the analyst's inbox.
Person dependency
The whole program lives on one analyst's laptop. They go on leave, change roles, or retire — and the SPC stops.
No audit trail
Excel charts don't track when a rule fired, who saw it, or what action was taken. Regulators and auditors want the trail — manual tools don't produce it.
The Ten Features That Actually Matter
Not every SPC tool marketed for manufacturing works for a power plant. These are the ten capabilities that separate software that survives past year one from software that gets shelved. Use it as your buyer's checklist.
01
Historian-native ingestion
Reads tags directly from PI, AVEVA, IP.21, GE Historian, and Honeywell without middleware. No CSV exports, no scheduled batch pulls.
02
Load-band-aware baselines
Control limits calculated per operating band, not as a single number. A 550 MW baseline is not a 300 MW baseline — the software should know that.
03
Ambient and fuel correction
Wet-bulb, coal GCV, and ambient temperature normalized out before rules run — so a hot day doesn't fire on the cooling tower signal.
04
Steady-state gating
Load ramps, startups, and shutdowns automatically excluded from baseline calculation. Transients don't muddy the SPC signal.
05
Calibrated rule recipes
Pre-set Western Electric and Nelson subsets per signal family — vibration, heat rate, emissions, cold-end — instead of "turn everything on and pray."
06
Multi-tag pattern recognition
Rules that look across related tags together — rising BP with rising TTD points at the condenser, rising BP with rising CW inlet points at the tower.
07
Work-order routing
A fired rule creates a specific inspection or CAPA task with an owner and deadline. No copy-paste from chart to CMMS.
08
Verify-and-close loop
After the action, the chart confirms the mean returned to baseline — CAPA closes only when statistical proof lands.
09
Role-aware dashboards
Operators, engineers, and station heads see different views of the same data — one dashboard doesn't fit three roles.
10
Auditable event log
Every fire, every acknowledgment, every action — logged with timestamp, user, and outcome. The record regulators and management audits both need.
Want to score your current setup against these ten? Book a demo and we'll walk your PI or historian tag list through the checklist.
Excel/PI Manual vs. Purpose-Built Automated SPC
This is what the real gap looks like on the ten capabilities. Same plant data, same rules — but the leverage of each column is completely different.
Capability
Excel / manual
PI ProcessBook trends
Purpose-built SPC (iFactory)
Historian ingestion
CSV export, manual
Native for PI only
Native for PI, AVEVA, IP.21, GE
Load-band baseline
Manual formulas
Single sigma
Auto per load band
Ambient correction
Not available
Not available
Wet-bulb & GCV normalized
Steady-state gating
Manual filter
Manual filter
Automatic exclusion
Rule recipes
Rule 1 only
Basic WE only
Per-signal calibrated set
Multi-tag patterns
Not supported
Not supported
Cross-tag rules native
Work-order routing
Manual copy-paste
Manual copy-paste
Automatic CMMS integration
Verify-and-close
Not tracked
Not tracked
Statistical CAPA closure
Role-aware dashboards
Single sheet
Per user, manual
Operator / engineer / head views
Auditable event log
No trail
Trend history only
Full fire & action log
How iFactory Sits on Your Existing Stack
The right automated SPC platform doesn't replace what you have — it makes the data you already collect actionable. This is where iFactory fits.
Field & DCS
Sensors, DCS, PLC, SCADA — no changes needed
Historian
PI, AVEVA, IP.21, GE Historian, Honeywell — read via native connector
iFactory SPC engine
Baselines, rules, cross-tag patterns, alerts, verification
CMMS & roles
Work orders and role dashboards to operator, engineer, station head
What Automated SPC Delivers
The value of automated SPC isn't just cleaner charts. It's the ability to keep running the program for years — and to convert real findings into real work.
10,000+
Tags handled
without a spreadsheet, without a team
Live
Baselines
recomputed as the plant changes
Auto
Work orders
from fired rule to CMMS in one step
Audit
Trail complete
every fire, ack, action, and closure logged
Curious how iFactory would map onto your PI or AVEVA setup? Talk to our integration team — we'll walk through your tag list and historian schema.
Frequently Asked Questions
Can't we just do SPC in PI ProcessBook or DCS trend?
You can plot the data — but that's a chart, not an SPC program. PI trends and DCS displays give you the reading; they don't compute load-band-aware baselines, run cross-tag pattern rules, exclude transients from the calculation, or route findings into work orders. On a proof of concept with ten tags they work. On 6,000 tags across 24×7 shifts, they turn into charts nobody looks at.
Does this replace our historian?
No. iFactory sits on top of your PI, AVEVA, IP.21, GE Historian, or Honeywell system — the historian keeps doing what it does today. iFactory reads the tags natively, runs the SPC engine on top, and pushes findings into your CMMS. Nothing on the field, DCS, or historian side changes.
How long does deployment actually take?
The pilot on one unit typically runs 6-12 weeks: historian connector setup, tag classification, baseline learning window, and calibrated rule recipes applied. After 30 days of live running, an analyst review tunes false alarms out. Plant-wide rollout scales from there — additional units usually take a fraction of the time once the recipes are proven.
What about integration with our CMMS?
iFactory routes fired rules into your existing CMMS — SAP PM, Maximo, or others — as work orders with the fired rule, affected tag, and recommended playbook attached. The verify-and-close step reads the CMMS status back so the SPC event only closes when the actual work does.
Can we see it running on our own plant before committing?
Yes. Bring one unit and 60-90 days of historian data. We'll connect, classify signals, build the baseline, apply the calibrated rule recipes, and show the fire log against actual history — real signals surfaced, false alarms suppressed. Book a demo and we'll walk it live.
Stop losing SPC in six months.
See Automated SPC on Your Plant's Real Data
Bring one unit, one historian, and 90 days of trend. We'll connect natively, classify every signal, apply the calibrated rule set, and show the fire log — with real drifts surfaced, false alarms suppressed, and work orders routed into your CMMS.
Native
historian connectors
