What can I do with MI Machine Learning?

MI Machine Learning allows users to quickly and easily:

  • Create machine learning models from data stored in Granta MI (in the Training app).
  • Understand which material properties and process parameters have the most impact on your final product or design, using the Visualize app.
  • Apply constraints and targets to a machine learning model and improve processes with the Optimize app.

The easy-to-use apps integrate seamlessly with other Granta MI applications, and can be used by anyone: you don’t need to be a data scientist or a mathematician to gain insights from MI Machine Learning.

MI Machine Learning apps in 

MI Machine Learning is split into three task-oriented apps: Training, Visualize, and Optimize.

Training

Tip: Construct a feature matrix from data stored in Granta MI, and send a model for training.

Create machine learning models by training and testing the Alchemite neural network on data from Granta MI. Intellegens’ Alchemite engine has been developed specifically for materials and processing data, and unlike other machine learning methods, it can be trained on sparse, noisy data without the need for extensive clean-up.

Visualize

Tip: View the correlation matrix for a trained model, or visualize features on a 2D or parallel axis plot.

Thousands of data points can be captured in a simple correlation table, making it easy to see the influences on different material properties.

Optimize

Tip: Apply constraints to or set target values for a model to solve optimization problems.

Target values and constraints are sent to the Alchemite engine, and a set of input and output values is returned. The engine also returns an overall probability that the predicted values can be achieved. From the results table, you can navigate directly to plots in the Visualize app, pre-populated with these values.

Example: AM process optimization

MI Machine Learning is particularly suited to Additive Manufacturing process optimization:

  • Identify correlations between raw material properties, machine build parameters, post-build processing, experimental or simulation results and your final product. The Alchemite engine allows you to use all available data, even when all data isn’t available for every part.
  • Machine learning models capture insights that can be used to optimize process parameters and powders, improving the quality of AM parts whilst reducing time-to-market.
  • Use the Optimize app to understand what input parameters will deliver the best design outcomes. Target a specific yield strength or surface roughness to find out what process parameters are required, or set build parameters to predict likely properties of the final part.