⚠ This machine has fewer than 10 runs. Shown for reference only — excluded from predictive recommendations.
ΔE tolerance
Customer spec3.0
Set a target color.
Awaiting target
—
—
ΔE to closest historical match
—
Recommended process recipe—
Bias voltage
—
C₂H₂ flow
—
N₂ flow
—
Ar flow
—
Temperature (°C)
—
Cathode config
—
Pumps speed
—
Expected output
—
Data coverage
—
—
Match confidence
Based on data density around the target in the training set.
KNN vs GP model comparison
KNN Historical best match
—
ΔE to target
Nearest historical run. Reliable within dataset. Cannot extrapolate.
GP Gaussian Process prediction
—
ΔE · 95% CI
——
—
Available pricing tiers
Standard
ΔE ≤ 3.0
Global model · always available
Precision
ΔE ≤ 1.5
≥ 15 runs in color region
Guaranteed
ΔE ≤ 0.8
≥ 8 runs near exact target
Nearest validated runs
Alternative process recipes — ranked by ΔE
Each alternative is a distinct historical recipe. ΔE bar shows match quality — narrower is closer. Temperature shown where recorded; otherwise per-machine average.
Recommend → Execute → Measure → Learn → Refine
Experiment Plan
Target unexplored regions first. Each suggestion represents a gap where additional runs would most improve model confidence.
Suggested next experiments — prioritised by coverage gap
Process window stability — repeated recipes
Recipe
Runs
Mean ΔE
Status
Machine
High ΔE on repeated recipes = unstable process window. Submit additional runs via Run Intake.
Active learning — model requests
Low confidence — dark near-neutral region
Only 2 runs exist with L* < 50 and a* < 3. 3 additional experiments at Bias 60, C₂H₂ 80–100, N₂ 1800–2000 would improve model confidence by an estimated 40% in this zone.
Machine 77 — Ar flow behaviour unconfirmed
Machine 77 shows an unusually strong Ar → L* correlation (r = 0.90). Based on only 33 runs. 5 targeted experiments varying Ar at fixed C₂H₂ and N₂ would confirm or correct this pattern.
Unstable recipe detected — B25C44N300A350
This recipe has been run twice with ΔE = 13.3 between runs. The process window may be near a transition boundary. A structured 3-point experiment is recommended.
Parameter space — C₂H₂ vs N₂
Dot colour = actual coating colour. Dot size = Ar flow. White space = unexplored process regions.
Submit a Run
Log actual process parameters and measured outcomes. Submissions pass through Gate 1–3 validation before reaching the model.
Machine
Actual process parameters
Bias V
C₂H₂
N₂
Ar
Measured color (L*a*b*)
L*
a*
b*
Enter L*a*b* to preview
Run outcome
Observations
Arcing observedTarget poisoningUnstable plasmaPoor adhesionInconsistent colorGood reproducibilityBetter than expected
Engineering notes (optional)
✓ Run logged. Passes through Gate 1–3 validation before reaching the model.
Run Log
Submitted runs this session. Each passes Gate 1–3 validation before reaching the model.
No runs submitted yet.
Integration status
MES / ERP connectorNot configured
PLC / machine exportNot configured
Quarterly update pipelineReady
Process Clusters
Six natural groupings of operating conditions identified from training data. Each represents a distinct process regime with characteristic color output.
Operating condition clusters — KMeans, k=6
Cluster interpretation guide
Tight clusters (low L* std) represent stable, reproducible process windows. These are your anchor points for reliable production recipes.
Wide clusters (high L* std) indicate parameter combinations that produce inconsistent color output. Avoid for production without additional DOE.
Small clusters (n < 10) are under-explored regions. The experiment suggestions on the Plan tab target these gaps.
Stability — repeated recipes
Recipe
Runs
Mean ΔE
Status
Machine
Diagnostics & Analytics
Machine performance comparison, color space coverage, and process trends for advanced process analysis.
Red = negative · Grey ≈ zero · Teal = positive Cols: L* · a* · b* × Global · M493 · M506 · M514 · M77
Simulate process parameters
Bias voltage (V)35 V
C₂H₂ flow45
N₂ flow350
Ar flow550
Temperature (°C)120
Cathode current (A)150
Num. cathodes2
Pumps speed (%)66
KNN from 382 real ZrCN runs. Temperature uses per-machine mean when not recorded. Cathode total = num × current.
Predicted coating output
—
L*
—
a*
—
b*
—
Prediction confidence
—
Parameter sensitivity at this operating point
Bias V
—
C₂H₂
—
N₂
—
Ar
—
Temp
—
Cathodes
—
Higher ΔE = more sensitive. Use high-sensitivity parameters for fine color tuning.
Model Insights
Statistical relationships between process parameters and L*a*b* color output — computed per machine. Some relationships flip direction depending on machine configuration.
ParameterChannel
Scatter — Bias V vs L*
Each dot = one run, coloured by machine. Dashed lines = per-machine trend.
Per-machine correlation r
Machine
r value
n
Direction
Heatmap — all parameters × channels
Bias VC₂H₂N₂Ar
Red = negative · Grey ≈ zero · Teal = positive Cols: L* · a* · b* × Global · M493 · M506 · M514 · M77
Global
C₂H₂ is the dominant color driver
Across all 381 runs, C₂H₂ has the strongest relationship: more acetylene = darker (L* r=−0.72), redder (a* r=+0.58). Direction consistent across all machines.
Global
N₂ is the second lever
Higher N₂ = higher L* (r=+0.37), lower a* (r=−0.44). Combined with C₂H₂, these two parameters explain most color variation.
Machine conflict
N₂ → L* flips sign on Machine 77
Globally more N₂ = brighter (r=+0.37). On M77 more N₂ = darker (r=−0.54). A recipe targeting high L* via N₂ on M493 may produce the opposite result on M77.
Machine-specific · M77
Ar is critical on M77 only
Ar has almost no global correlation (r=+0.07). On M77 it becomes the strongest predictor: Ar → L* r=+0.90. On M77, Ar is effectively the primary color control lever.