Luke Hillery
+ LAH Studio


Commercial
  1. Adizero Control
  2. Techfit Control
  3. RHEON™ Armour
  4. BOLT™ Impact Armor  
  5. Ruroc Atlas 4.0
  6. Rapha Trail Knee Pad


Personal
  1. ECOUR


Student
  1. Making a Pointe
  2. Axowear
  3. Ghostwriter



Bio
Email
Instagram
Pinterest



©2024 LAH STUDIO


Making a Pointe:  The Automated Design of Customised Ballet Pointe ShoesMaster’s Thesis 2021




// The Why

 Ballet pointe shoes, often requiring challenging "breaking in" routines, are among the most personalized products in the footwear industry. This stems from the unique materials and manufacturing techniques used in their creation. But there's a new avenue for innovation: automated design and production, informed by 3D scanning and parametric design, can replace the traditional one-off, custom last model needed for bespoke footwear. 

Instead, we can create a "digital last" that serves as the foundation for customizable footwear components.
In this study, the potential of using user measurements and preferences to create fully customized components tailored to individual dimensions and abilities is explored. Three data-driven parametric modeling approaches based on 1D, 2D, and 3D user data are investigated. 

The research demonstrates the feasibility of cost-effective 3D scanning and photography-based methods to craft personalized components based on dancers' anthropometric data, ultimately enhancing footwear fit and reducing injury risks in ballet.
// Methodology

3D scan data was gathered from SATRA Technology Centre, including 50 female subjects aged 18-34, with an average age of 25.5 and shoe size UK4.5. These subjects represent a significant demographic for ballet dancers. Additionally, two pointe-trained ballet dancers we scanned locally for physical prototyping and qualitative fit validation. 


 Mesh manipulation software was used to repair imperfections in the scan data, ensuring manifold surfaces, correct orientation, and no self-intersections. To reduce computational complexity, the triangle count of each mesh was reduced and noisy surfaces above the ankle were removed consistently across all meshes.

This methodology ensured the acquisition and preparation of accurate 3D scan data for subsequent parametric modeling and fit analysis.

// Automating Parametric Design

PARAMETRIC WORKFLOW, LUKE HILLERY 2021

The research employs parametric modeling with McNeel Rhinoceros® Version 7.0 and the Grasshopper 3D plugin. The study evaluates pointe shoe fit quality with varying data accessibility through three different approaches:

  • One-Dimensional Approach: This method uses linear foot dimensions (e.g., instep height, instep width, foot length) to non-uniformly scale the toebox and shank components. Manual measurements are automatically extracted from the 3D mesh within the Grasshopper script, ensuring measurement accuracy. Most measurements exhibit high correlation, with a few exceptions.

  • Two-Dimensional Approach: Utilizing 2D digital photographs of the foot, this approach extracts curves, profiles, and linear dimensions. Researchers project the 3D mesh in Cartesian directions (x, y, z) to obtain foot outlines. This method minimizes data variations when comparing 2D and 3D results, eliminating the need for separate 3D foot mesh alignment. It's highly accessible, requiring only a digital camera and alignment software.

  • Three-Dimensional Approach: The advanced approach utilizes the refined mesh from raw scan data. Researchers reconstruct the foot and ankle's manifold mesh as a polysurface to closely match the foot's topology when generating the toebox geometry.

These approaches enable the assessment of pointe shoe fit based on different data complexities, ranging from simple linear measurements to detailed 3D scans.

// Results


In the quantitative fit analysis, the research evaluated the "fit" of each component design process by calculating the RMS Euclidean distance between the foot and toebox surfaces. 

Here are the key findings:

  • Using linear foot dimensions alone resulted in an RMS Euclidean distance with a mean of 5.447 mm.

  • Two-dimensional input data produced a mean of 2.070 mm, with a narrower spread than one-dimensional input.

  • Three-dimensional input data achieved a mean of 0.082 mm and showed a much narrower spread than one- and two-dimensional input.

 For data visualization, a "fit heatmap" was generated for each foot-component pairing, using a color gradient to highlight areas with significant deviation from the true foot form. Notably:

  • One-dimensional input data closely fit around the Metatarsal-Phalangeal Joint (MPJ) and exhibited the largest deviation (up to 20mm) on the dorsal side of the foot above the 3rd and 4th MPJ.

  • Two-dimensional input data showed a more uniform fit across the dorsal side of the foot but had significant deviations on the dorsal side of the 4th and 5th MPJs.

  • Three-dimensional input data provided the most uniform fit, with some minor deviations at the tips of the big and 2nd toes.
FIT HEATMAP GENERATION SNIPPET, LUKE HILLERY 2021

Index   Next →