Scrap Reduction via DOE (Green Belt Project)

Overview

A high‑volume fabric style became the top contributor to department scrap, driven by weight, shade, and width variation. Using a full DMAIC approach, I integrated multi‑system data, validated the measurement system, identified statistically significant contributors, and executed a 7‑factor DOE to stabilize the process and reduce scrap.

Problem

This style generated significant annual scrap losses, driven by:

High variation in weight
Off‑spec shade and width
Frequent redries
Inconsistent machine settings
No standardized recipe

Baseline capability was poor (CpK = 0.135), and operators lacked a reliable recipe to consistently meet specifications.

Aproach

1. Data Integration & Baseline Analysis

I consolidated data from SAP, MES, Trend, and Lab systems into a unified dataset.

Key steps:

Stratified by material, speed, redry status, and production order
Identified top defect categories
Established 2022 as baseline
Conducted capability analysis (CpK < 1.33 → not capable)

2. Measurement System Analysis (MSA)

Performed a Gauge R&R to validate the weight measurement process.

%GRR = 26%
5 distinct categories
Measurement system acceptable for improvement work

3. Statistical Analysis

Using regression and ANOVA, I identified two contributors with statistically significant impact on weight:

Speed
Jet/Redry condition

Both null hypotheses were rejected.

4. Design of Experiments (DOE)

Performed an Impact vs. Effort analysis, which identified a DOE as the best path forward. Then executed a 7‑factor, 2‑level DOE to optimize:

Speed
Rail width
Pad pressure
Oven temperature
Water temperature
Fans
Overfeed

Created 16 experimental recipes and evaluated weight and width performance.

DOE Result:

R² = 94%
Identified optimal combination of speed, pad pressure, oven temperature, and rail width
Developed three progressively improved recipes that led to reduce the weight variation of the critical style.