Icon

Semiconductor_​FAB_​KNIME_​v1.0

Semiconductor FAB KNIME -CMOS 90nm · 500 wafers · 13 Front-End phases Tools: ASML PAS5500 · AMAT Mirra · LAM 9400 · KLA 5200 · TEL Lithius Author: [tuo nome] · GitHub: github.com/[tuo-user]/semiconductor-fab-knime License: MIT · 2026

BAY 0 - DATA SOURCE & QUALITY CONTROL Fab Virtual · 500 wafers · 13-phase CMOS 90nm Front-End Real tool specs: ASML, AMAT, LAM, KLA, TEL · APC + tool aging

BAY 1 - EXPLORATORY DATA ANALYSIS (EDA) Multi-phase distribution - scrap root cause - correlation matrix

ANALYTICS - FAULT DETECTION & ML MODELS Scrap prediction - 70/30 train-test split - stratified - seed=42

ANALYTICS - WHAT-IF PROCESS OPTIMIZATION Etch temperature scenarios - predicted SCRAP reduction per tool

BAY 0B - PROCESS DRIFT & SPC MONITORING Time-series SCRAP trend per tool - EWMA drift detection

F00_Statistics - descriptive stats - baseline QC check
Statistics
F00_ColFilter - select 87 numeric + 5 categorical - drop batch metadata
Column Filter
F00_MissingVal - median imputation <3% NaN · per-phase strategy
Missing Value
F00_Num2Str - encode categorical: Tool_ID, Batch, Shift
Number to String
F01_BarChart_Scrap — SCRAP rate by tool per phase · identify outliers
Box Plot
SCRAP rates for tool Annealing are similar (~20%) the problem is not the equipment but the process conditions (night shift and peak temp)
Bar Chart
F01_GroupBy_Scrap — SCRAP% by phase and tool · cross-tab analysis
GroupBy
SCRAP rate per tool-phase pair - used for SPC control chart
GroupBy
Gradient Boosted Trees Predictor
F00_LinePlot_Drift - SCRAP% trend over BatchIdx · detect tool aging
Line Plot (legacy)
WI_Results_C
GroupBy
WI_Scenario_C1 — Etch_Temp -5°C e WI_Scenario_C2 — HF_pct -1%
Math Formula
Etch_Temp
Math Formula
ML_RF_Predictor
Decision Tree Predictor
Scorer
F00_ColorMap - phase color encoding for trend visualization
Color Manager
CSV Reader
ML_Split - 70% train / 30% test - stratified on SCRAP_FINAL - seed=42
Table Partitioner
Scorer
ML_RF_Learner - 100 trees - Gini - seed=42
Random Forest Learner
Random Forest Predictor
ML_GBT_Predictor
Gradient Boosted Trees Predictor
Multi-stage root cause: final SCRAP is driven by defect propagationacross Lithography (F05), Oxidation (F03), and Wet Etch (F06).AUC=0.714 - GBT outperforms DT and RF on imbalanced classes.
Decision Tree View
Scorer
ML_GBT_Learner - 100 iter - lr=0.1 - best model AUC=0.714
Gradient Boosted Trees Learner
ML_DT_Learner - Gini - max depth 5 - min node 10
Decision Tree Learner
The correlation matrix shows physically consistent correlations: HF_Concentration and EtchRate show strong positive correlation (r≈0.9), while Temperature and Selectivity show negative correlation, confirming the critical role of TOOL_C.
Linear Correlation
WI_Results_A
GroupBy
ML_DT_Predictor
Decision Tree Predictor
Scorer
Row Filter
ML_ColFilter - drop non-numeric keep 87 process features
Column Filter
ML_RuleEngine - threshold=0.15 - TPR=49% - FPR=17%
Rule Engine
drift Thermal Ox
Box Plot
ML_ROC - GBT - AUC=0.714 - positive class=SCRAP
ROC Curve
WI_Results_B
GroupBy
Gradient Boosted Trees Predictor
Row Filter
WI_Scenario_B - Etch_Temp nominal - baseline
Math Formula
WI_Scenario_A - Etch_Temp +5°C - simulate overtemp
Math Formula
Gradient Boosted Trees Predictor

Nodes

Extensions

Links

Download

To use this workflow in KNIME, download it from the below URL and open it in KNIME:

Download Workflow
Created by: simone
Created at: 2026-05-18
On NodePit since: 2026-05-28
Last update: 2026-05-28
Created with KNIME version: v5.11.0

Deploy, schedule, execute, and monitor your KNIME workflows locally, in the cloud or on-premises – with our brand new NodePit Runner.

Try NodePit Runner!