Hit enter to search or esc to close

Knitted Wire Mesh & Neural Networks

KnitMesh Technologies is working with the University of Liverpool Department of Engineering to develop a design tool that will model the performance of knitted mesh components.

Typical Analysis of a Knitted Mesh Product

Knitted mesh products are complex in structure and performance influenced by controllable and uncontrollable factors. The combination of factors with different design variables increase the complexity of product performance, the design tool will model the performance based on the maximum load a given part would withstand.

The component to the right is a typical mesh ring used as a gasket or sealing element in an assembly. The part is subjected to compressive force in application, so the maximum load capacity defines product performance. The maximum load capacity is determined by compression loading at various load limits.

  • The mesh ring is subjected to compressive loading. (Rig shown on left)
  • The Load-Deformation data is recorded and exported for data analysis
  • The fundamental principle of conservation energy is employed for understanding the load carrying capacity
  • The energies obtained for different load limits are plotted to identify maximum load capacity which are termed as “Breakeven Loads”

For the sample used in this test the plot of energy versus compressive load is shown on the left. The point at which the elastic and plastic compressive energies diverge is determined to be the Breakeven Load. In this particular case the breakeven load is 300N. Factors which affect breakeven load include:

  • Wire diameter
  • Component mass
  • Stitch width and height
  • Number of wires
  • Wire material
  • Crimp height

Research & Development

In an attempt to predict the performance characteristics of a compressed knitted mesh component before it is made, and hence reduce development times, KnitMesh are developing a theoretical system based on neural networking.

  • Breakeven load data is stored in a database along with the parameters which contributed to each particular test result.
  • Neural network designed to predict unknown data which can be Breakeven load, maximum deflection, wire diameter or part mass for given input
  • This output is then used to produce a prototype design

A Neural Network is a complex arrangement of interconnecting nodes where an output from one node (or nodes) triggers a response in another node (or nodes).

The network can be ‘taught’ to solve complex engineering problems where a fundamental analysis is not possible – for example in the complicated ‘random’ arrangement for knitted mesh.

Once established, the Neural Network can be used to optimise designs. For example, for a required load / deflection characteristic, a design could be selected that minimised cost or weight, or used a particular wire grade.