Multiple model optimization is a useful technique which allows predicting the optimal factor conditions that simultaneously satisfy several models’ targets. More details can be found here.
In Synthace, models are optimised by searching for the maxima or the minima of the modelled response. Alternatively, you can choose acceptable response ranges for the models to constrain the optimization within these bounds. If desired, you can choose to optimize multiple models simultaneously.
The Explore Model tab and the tools it provides in order to explore one model at a time are described here. In this article, the focus is on the Explore Multiple Models tab where you will find tools to define acceptable response ranges and to optimize multiple models simultaneously.
How to optimize a model within an acceptable range
Once you create and save the model(s) of interest, navigate to the Explore Multiple Models tab.
Under Models & Response Ranges, select the checkbox corresponding to the model of interest.
By default, the slider values under the model are set to the middle third of the response range. Adjust the acceptable response values by dragging the lower and higher slider handles. The contour plot will update accordingly to show the selected response range as a coloured area. You can also change the vertical and horizontal plot axes to two distinct factors of interest using the drop-down menus below the plot.
Choose either Minimize or Maximize from the drop-down next to the selected model.
Click the Optimize button under Multiple Response Optimization.
If the optimization is successful, the following items will update:
The model name and the desirability percentage will appear under the Optimize button. More details about the way the desirability is computed can be found here.
The slider values of the factors, under Factor Values, will change to the levels at which the response optima are achieved.
The green lines on the plot will correspond to the optimal factor levels found in the optimisation.
How to optimize multiple models
Under Models & Response Ranges, select the checkboxes corresponding to the models of interest.
By default, the slider values under each model are set to the middle third of the response range. Adjust the acceptable response values by dragging the lower and higher slider handles. For the entire response range, drag the slider handles to the opposite ends.
The contour plot will update accordingly to show the selected response ranges for each chosen model as a distinct coloured area. Note, however, that a coloured area of a model will only be displayed if the factors selected for vertical or horizontal axes are included in the model terms.
Choose either Minimize or Maximize from the drop-down next to the selected models.
Click the Optimize button under Multiple Response Optimization.
If the optimization is successful, the following items will update:
For each model included in the optimization, the model name and its corresponding desirability percentage will appear under the Optimize button. You can find out more about the way the desirability is computed here.
The slider values of the factors, under Factor Values, will change to the levels at which the response optima are achieved.
The green lines on the contour plot will correspond to the optimal factor levels found in the optimization.
Troubleshooting
If the optimization based on the selected response ranges doesn’t succeed, a warning message will be displayed below the Optimize button. Additionally, the selected response bounds are displayed for each model.
One possible cause for an unsuccessful optimisation is having response ranges that are not offering a common solution space for the factors. In this case, try extending the response ranges by adjusting the model slider values and repeating the optimization.
Unsuccessful optimization example
Example on multiple model optimization within acceptable ranges showing successful optimization as well as with non-overlapping range showing a failed optimization.
Well done for making it to the end of this tutorial.
To learn how you can use your model to decide on what your next experiments might be, click here.
To learn about how you can quickly browse all the models you have created and get some additional metrics on their quality, click here.