Mechanics-Consistent Adhesion in Soft Wearables
ACTIVE1. The Situation: The “False Positive” Trap
The Gap: Current industry standards (like ASTM D1876) assume that higher average peel force equals better adhesion.
The Problem: I identified that this assumption fails for soft wearables. In stretchable textiles, up to 90% of the “measured force” is actually just the fabric stretching (dissipation), not the glue holding. This creates a “False Positive Trap”: selecting adhesives that look strong in the lab but fail functionally on the human body due to stick-slip instability.
2. The Action: Rational Engineering
Instead of relying on misleading averages, I applied first-principles thinking to re-engineer how we define failure.
- Logic over Data: I deconstructed the force profile to distinguish between true adhesion (the peaks, representing crack resistance) and artifacts (the troughs, representing slip).
- The Metric: I introduced the Peel Stability Index (PSI). Unlike raw force, this metric quantifies consistency. A high PSI predicts that a sensor will perform reliably without electrical noise, preventing costly device failures downstream.
3. The Execution: Automated Python Pipeline
To remove human bias and speed up analysis, I built a custom Python pipeline that operationalizes this physics-based logic.
(Noisy & Unreliable)"] --> B["Preprocessing
Data Cleaning & Unit Conversion"] B --> C["Baseline Correction
Normalize Starting Force to 0N"] C --> D{"Identify Steady-State
Window"} D -- "Scan: Min. Std Dev" --> E["Peak & Trough Detection
Isolate Initiation Forces (Fci)"] E --> F["Calculate Stability Metrics
PSI = Fmean / SigmaF"] F --> H["Summary Report
Metrics: Gci, PSI"]
Key Contribution: This automation reduced data processing time by 80% while ensuring that only mechanically stable (Category II) adhesives are selected for the final medical device.
4. The Result
- Framework prevents “False Positive” selections
- Ensures mechanically stable adhesives for medical devices
- Validated using T-Peel testing (ASTM D2724 adapted) with custom Python signal analysis
