Optimize an AM build using machine learning

In this example, whether the Border power can be reduced by 25% and still meet the part’s design requirements is investigated.

Design requirements:
  • Tensile strength of at least 750 MPa.
  • Surface roughness of less than 10 µm.
The model created in the previous exercises is used to answer this question without having to do further testing.
  1. In Granta MI, go to Machine Learning > Optimize. 
  2. Create a new Criteria Set.
    1. Click Create new Criteria Set.
    2. Enter a Name for your new Criteria Set.
    3. Select the model that you created during the Training exercises with the highest Model Quality.
    4. Click Save.
  3. Enter target values for the output features you want to optimize.
    Each target consists of a Goal and a Value, entered into the table on the right-hand side.
    1. Enter a target for Ultimate tensile strength.
      1. Click the Goal cell in the Ultimate tensile strength row, and select Greater than from the drop-down list.
      2. Click the Value cell, and enter 750. Don’t edit the Weight.
    2. For Surface finish, select a Goal of Less than and enter a Value of 10.
  4. Enter constraints on the input features
    . Each constraint consists of a Constraint type and Value, entered into the table on the left-hand side.
    1. Click the Constraint cell in the Border power row, and select Between.
    2. Click the Value cell, and change the maximum value from 400 to 300.
      Press ENTER or click away from the cell to apply the change.
    3. Leave the rest of the input features unconstrained (Constraint = Full Range).
  5. Click Optimize to send the trained model and Criteria Set to the engine for optimization.
You can see the status of your job on this page, the Optimize a model page, and in the MI Job Queue.