Technical Study Summary · Process Engineering

Improving Packaging Seal Strength
Through Design of Experiments

A 2³ Full Factorial DoE study identifying the optimal machine settings to maximise seal strength on food wrapping equipment.

Study Period: January – March 2013 Design: 2³ Full Factorial · 24 Randomised Runs Application: Food Wrapping Machinery Analysis Tool: Minitab 16 · ANOVA

Section 01

Study Background & Objective

Business Driver

Recurring customer complaints regarding package seal failures during transit had led to an estimated 3–4% product return rate in the quarter preceding the study. The QA team identified inconsistent sealing as the root cause, prompting a structured DoE investigation rather than continued OVAT (one-factor-at-a-time) testing.

Primary Objective

Identify the optimal combination of sealing temperature, sealing pressure, and machine speed to maximise seal strength and reduce field returns, while building statistical capability within the engineering team.

Context note: This study moved the team from OVAT testing to a statistically rigorous full factorial framework. Analysis ran in Minitab 16. Seal strength was measured using a calibrated digital tensile tester per ASTM F88 Standard Peel Test, 5 samples per run.

Section 02

Experimental Design

Sealing Temperature

Factor A · Positive Effect

Drives polymer chain diffusion to sealing initiation temperature. Largest positive lever in the study.

150°C (Low) 170°C (High)

Sealing Pressure

Factor B · Positive Effect

Acts as a thermal conductor — compresses film layers, removes air pockets and enables heat penetration to the sealant core.

40 psi (Low) 60 psi (High)

Machine Speed

Factor C · Negative Effect

Higher speed reduces dwell time between jaws — film core cannot reach sealing initiation temperature, producing a shallow bond.

80 ppm (Low) 100 ppm (High)

Response Variable

Seal Strength — N/m

Force required to peel apart sealed edge per ASTM F88. 5 samples per run, averaged. Minimum specification: 650 N/m.

Min spec: 650 Target: 887+

Design structure: 2³ full factorial · 8 unique treatment combinations · 3 replicates = 24 total randomised runs · Resolution V (no confounding between main effects or 2-factor interactions) · Statistical power >90% for effect sizes ≥30 N/m · Software: Minitab 16

Section 03

Theoretical Foundations

✓ Full Factorial 2³ — Used Here

✓Orthogonality: Each factor's effect is estimated independently — no cross-contamination of results.
✓Interaction Detection: Captures synergies like Temp × Pressure that OVAT permanently misses.
✓Efficiency: 8 unique runs cover the complete experimental space. With replication: 24 total runs.
✓Resolution V: No confounding between main effects or 2-factor interactions — all terms estimable.

✗ OVAT — Avoided in This Study

✗Interaction Blind: Changes one variable while holding others fixed — cannot detect synergies between factors.
✗Runs Inefficient: More total experiments for less information; poor signal-to-noise ratio.
✗Optimum Missed: The true optimal is often at an interaction point invisible to OVAT.
✗No Error Estimate: Without replication structure, experimental error cannot be properly quantified.

ANOVA — Key Statistical Concepts

Variance Decomposition

SS_Total = SS_Model + SS_Error

Separates the variation explained by controlled factors from random experimental noise.

F-Statistic & p-value

F = MS_Factor / MS_Error. If p < 0.05, reject H₀ — factor has a real effect with 95% confidence.

R² & Adequate Precision

R² = proportion of variance explained (0–1). Adequate Precision = signal-to-noise; must exceed 4.0 for a navigable model.

Section 04

2013 Analytical Dataset

Run	Temp A (°C)	Pressure B (psi)	Speed C (ppm)	Rep 1 (N/m)	Rep 2 (N/m)	Rep 3 (N/m)	Average (N/m)
1	150 (−)	40 (−)	80 (−)	610	625	618	617.7
2	170 (+)	40 (−)	80 (−)	755	740	762	752.3
3	150 (−)	60 (+)	80 (−)	690	705	698	697.7
4	170 (+)	60 (+)	80 (−)	880	895	887	887.3 ★
5	150 (−)	40 (−)	100 (+)	550	565	558	557.7 ▼
6	170 (+)	40 (−)	100 (+)	640	655	647	647.3
7	150 (−)	60 (+)	100 (+)	615	630	622	622.3
8	170 (+)	60 (+)	100 (+)	710	725	718	717.7

Optimal (Run 4)

887.3

N/m · 170°C / 60 psi / 80 ppm

Worst (Run 5)

557.7

N/m · 150°C / 40 psi / 100 ppm

Range

329.6

N/m spread · Min spec: 650 N/m

★ Optimal row highlighted blue ▼ Worst row highlighted amber · Runs 5 & 6 below the 650 N/m minimum specification at original settings.

Section 05

ANOVA Results & Factor Coefficients

A POSITIVE · p < 0.001

+65

Sealing Temperature

Highest sensitivity. Drives polymer diffusion; dominant single factor in the model.

C NEGATIVE · p < 0.001

−55

Machine Speed

Reduces jaw dwell time. Film core cannot reach sealing initiation temperature — shallow bond results.

B POSITIVE · p < 0.001

+40

Sealing Pressure

Thermal conductor — flattens film layers, removes air pockets, enables heat to reach sealant core.

AB INTERACTION · p = 0.015

+20

Temp × Pressure

Significant synergy. Benefit of increasing temperature is much greater at 60 psi than at 40 psi.

Seal Strength (N/m) = 715 +65·A +40·B −55·C +20·A×B
Coded factors: A=(T−160)/10 · B=(P−50)/10 · C=(S−90)/10

⚠ Important gap: The model as reported does not include the ABC three-factor interaction term, and R², Adjusted R², Predicted R², and Adequate Precision were not formally presented. Published comparable 2³ studies have found the three-factor interaction highly significant. This term should be estimated and tested before deploying the regression model for process control.

Section 06

Temperature × Pressure Interaction (AB)

The Physics Behind the Interaction

→Thermal Conductivity: High pressure (60 psi) compresses film layers, removing microscopic air pockets that otherwise act as thermal insulators.
→Sealing Initiation Temp: At 40 psi, only the film surface reaches sealing temperature. The polymer core remains unactivated, producing a shallow bond.
→Synergistic Gain: At 60 psi + 170°C, both film surfaces achieve full polymer diffusion — the mechanism behind the non-parallel interaction plot lines.
→Practical Implication: Optimising temperature or pressure independently misses the true optimum. Both must be set jointly.

INTERACTION PLOT — Temp × Pressure

Non-parallel lines confirm significant interaction · p = 0.015

Key implication: The gain from raising temperature is +190 N/m at 60 psi but only +135 N/m at 40 psi. Operating at high temperature with low pressure wastes energy and underperforms. The two parameters must always be set in concert.

Section 07

Identified Optimal Process Settings

Temperature

170°C

Factor A — High Level

Pressure

60 psi

Factor B — High Level

Machine Speed

80 ppm

Factor C — Low Level

Section 08

Study Gaps vs. Modern DoE Best Practice

Gap Area	2013 Original Approach	Recommended Improvement	Priority
ABC 3-Factor Interaction	Not tested in ANOVA table	Published 2³ full factorial studies have found the three-factor interaction significant (F>100, p<0.0001) in comparable processes. This term must be explicitly estimated — omitting it leaves the model underspecified.	High
Model Diagnostics	R², Adj R², Pred R² not reported	Report R², Adj R², Pred R², and Adequate Precision (must exceed 4.0). A well-specified 2³ model should achieve R²>0.98 and Adequate Precision>30 for reliable process navigation.	High
Gage R&R	No measurement system study conducted	Pre-study Gage R&R required per AIAG MSA 4th Edition. Measurement variation must be <10% of total process variation before committing to a DoE.	High
Centre Points	Not included — curvature undetectable	Add 3–5 centre-point runs (160°C, 50 psi, 90 ppm) in Phase 2 to detect non-linearity in the response surface.	Medium
Residual Analysis	No normality / variance checks shown	Generate: normal probability plot of residuals, residuals vs. fitted, and run-order plot. Non-constant variance is common in heat-sealing processes.	Medium
Response Surface & Contour Plots	Not generated in either document	Phase 2 RSM study (Central Composite Design) to map the full seal strength response surface — identify exact optimal ridge and throughput trade-off zones.	Phase 2

Section 09

Implementation Roadmap

Four immediate actions before full plant deployment, followed by a structured Phase 2 programme:

Parameter Locking — SOP Update

Lock 170°C / 60 psi / 80 ppm as the standard SOP for the food wrapping machine. Update machine setup card and Andon board immediately. Maintain 60 psi jaw pressure via bi-weekly mechanical audit.

Jaw Hardware Audit

Inspect seal jaw heating elements for cold spots using a thermal camera. Temperature uniformity across the jaw face matters: localised cold zones directly undermine the large positive effect of Factor A.

Sensor & PID Upgrade

Upgrade thermocouple precision to ±1°C. Install jaw pressure transducers for real-time PSI monitoring. Consider PID loop auto-tuning to eliminate thermal drift between production shifts.

Cpk Validation Run

Execute 50-pack stress run at optimal settings. Compute Cpk (Process Capability Index). Target Cpk ≥ 1.33 before full plant rollout. Document prediction interval compliance and confirm model predictions.

Phase 2 Roadmap

→Conduct RSM study (Central Composite Design) with centre points to explore non-linear response and generate full 3D contour maps of the seal strength surface.
→Investigate film type and thickness as additional factors — significant cost levers in the FMCG packaging market.
→Explore whether intermediate speed (90 ppm) with compensated temperature/pressure settings can recover throughput without sacrificing seal quality.
→Consider pre-heating film as a mechanism to unlock higher throughput (100–120 ppm) without dwell-time trade-off.

Section 10

Key Takeaways

✓

Optimal settings confirmed: 170°C / 60 psi / 80 ppm delivers 887.3 N/m seal strength — 59% above the worst-case combination and well above the 650 N/m minimum specification.

✓

Factor hierarchy established: Temperature is the primary lever (+65), Machine Speed the greatest threat (−55), and Pressure the structural foundation (+40). Speed should never be increased without compensating adjustments to the other two.

✓

Temperature × Pressure interaction is real and significant (p = 0.015): The synergy is not additive — temperature and pressure must be jointly optimised. Setting them independently means missing the optimum.

✓

Full Factorial design was the right call: the AB interaction would have been permanently invisible under OVAT. That decision enabled the most important finding of the study.

Three open items before production rollout: (1) ABC three-factor interaction not tested, (2) model diagnostics (R², Adequate Precision) not reported, (3) pre-study Gage R&R on the tensile tester not conducted. All three are classified High Priority.

→

Phase 2 path is clear: RSM study with centre points to map the full response surface, detect curvature, and unlock throughput at 90–100 ppm without sacrificing seal integrity.