Skip to main content

Hitachi
Contact UsContact Us

Consumer Packaged Goods

Find candidates of toothpaste composition with an optimized blending ratio

THE OBJECTIVES

Get blending ratio to synthesize the compounds with high-quality.

THE CHALLENGES

  • Too many blending patterns to search a good blending ratio.
  • Relationship between material properties and fluidity is not clear.

[image]Customer’s existing experimental data

THE APPROACH

  • Create machine learning model using the patterns of existing experimental data.
  • Clarify the raw material that affect material properties.

[image]Clarification

  • Execute inverse analysis with optimization algorithms.
  • Search blending patterns in virtual space to get target properties.
  • Suggest efficient experimental plan by using Bayesian optimization.
  • Performed constrained optimization such as raw materials costs.

[image]Optimization

  • Output experiment candidates that satisfy the target fluidity.

[image]Experiment candidates

THE RESULTS

  • Reduction of the experiment trials and costs.
  • Efficient production by optimized process.

Related pages