Interpolators Discussion

[BenW0]

Add separate test parts for estimating the performance of intermediate features as one of the parameters changes around the conditions of the main evaluation print:

• Angle with respect to the print plane (angle of elevation)
• Aspect ratio of the feature along the feature length (longer vs. shorter holes/fins/etc)
• Aspect ratio of the feature across the feature (pillars mapping to fins)
• Orientation on the build plate (rotation in the build plane) -- do this one with the angle of the part something besides horizontal/vertical

For each interpolation test, we do a lengthy process where we do a separate test part for each feature type. For example, the aspect ration along the feature length test for pillars would print pillars of different heights (from bumps through pillars 2 or 3 times higher than the model aspect ratio). For each aspect ratio there would be the normal set of 10 pillars of different sizes (based on the output of a normal calibration run), using the wizard interface to iterate only until a green/yellow transition is known for each aspect ratio.

Possibly, when we have obtained a green/yellow transition for a given aspect ratio, we can just skip printing that part in the subsequent iteration, since we're not trying to zoom in too much...

TODO:

• Design the new parts
• On the front end,
  • Allow looking up the results of a previous trial
  • Allow starting a new trial based on the results of an old one
  • Allow selection of the kind of calibration run to perform (i.e. which model to use)
• On the back end
  • Implement server-side (or client-side cached) storage of current parameters. We can't do this with urls

[BenW0]

I think it may be helpful to create a consistent way to characterize the different kinds of features for linking pieces of the main test part to different interpolations. Excluding fillets (for now, perhaps forever), all the features are defined by

• Sign - positive or negative
• Primary aspect ratio (height/diameter for pillars, for example)
• Secondary aspect ratio (width/thickness for vertical fins, for example)
• Angle of inclination (horizontal/vertical for test parts). We'll map this to 0 (0° = flat) - 1 (90° = vertical)
• FUTURE: Angle of rotation (in the x/y plane)

We can intuitively combine sign with primary aspect ratio (negative primary aspect ratio means it is a hole/negative slot). This gives us 3 coordinates that any linear feature can be mapped to. How do we map non-straight features into this system???

So our vertical pillars are currently mostly [7, 1, 1], horizontal slots are roughly [-7, 2, 0]. Each experimental trial fills in a point on the 4D space with the feature size associated with that [aspect, aspect, angle] coordinate.

But how do we interpolate after we have all this data???

[BenW0]

Today's idea for describing an arbitrary piece of geometry in a way that lets us treat the variables as more or less independent parameters:

• Sign
• Primary aspect ratio (thickness/length) along the main axis of the patch
• Secondary aspect ratio (thickness/width) along the secondary axis of the patch
• Angle of inclination of the patch with respect to the build plane (not angle with respect to the main axis)

Note that as the secondary aspect ratio approaches 1, the angle of inclination needs to be checked both across the face and along the major axis, since it is approaching a post.