Robotics and machine learning helped researchers design packaging films with chosen combinations of strength and transparency. The author preprint posted on 26 May 2025 describes a search that first learns which mixtures form usable films.[1]
An automated pipetting robot prepared 2,420 recipes from 23 ingredients. A classifier screened for detachable, uniform films. Sixteen further learning rounds made and tested 343 films, shifting from broad exploration toward better mechanical properties.[1]
Strength alone does not describe a packaging film. A stiff material resists deformation, while energy-to-break describes how much energy a specimen absorbs per unit volume before rupturing. ASTM’s thin-film testing guidance treats these as distinct properties and explains that thickness, pulling speed and specimen preparation affect comparisons. For a designer, that means a ranking needs a specified property and a consistent test; a single adjective such as “strong” does not identify the job a film can do.[2]
One model-selected film measured 152 MPa ultimate tensile strength and 91% light transmission at 550 nanometres. Tensile tests used three replicates. The roughly billion recipes evaluated by the model were predictions, not fabricated samples.[1]
Sources & further reading
- Predictive Design of Sustainable Biobased Packaging via Machine Intelligence for Improved Postharvest Preservation Research Square · 26 May 2025 · Accessed 29 Oct 2026
- Standard Test Method for Tensile Properties of Thin Plastic Sheeting ASTM International · Accessed 29 Oct 2026