This code base was archived on March 14, 2023. This represents a mid-stream snapshot of the development cycle of a machine learning approach to predicting the line widths resulting from an EHD printing process. Development of the image analysis and dataset construction routines continues in onakanob/ehd-dataset. Machine learning models and evaluation continues in onakanob/ehd-ml.

ehd_exsitu

Image analysis for ex-situ characterization of ehd-printed patterns

Workflow:

1. Use align_pattern.py to set the offset and angle for a mosaic image so that the EHD toolpath pattern lines up with the printed pattern
2. Use the GUI in place_patches.py to place an image patch over each isolated print pattern
3. Run parse_patches.py to run image analysis on each patch, extracting metrics.

Requirements:

Section is incomplete

• sklearn

ehd_dataset

The EHD_Loader object holds multiple training datasets in the loader.datasets array, each a dataframe containing waveforms and measurements from a single experiment. When returning a dataset, the "xtype" and "ytype" arguments control how the X and Y variables (input and supervised output, respectively) will be formatted. The following options are available:

xtype

• vector
• wave
• last_wave
• last_vector
• normed_squares
• v_normed_squares

ytype

• area
• print_length
• max_width
• mean_width
• obj_count
• jetted
• jetted_selectors

ehd_model

Regressors:

• MLE
• cold_RF
• cold_MLP
• only_pretrained_RF
• only_pretrained_MLP

Classifiers:

• MLE_class
• cold_RF_class
• cold_MLP_class
• only_pretrained_RF_class
• only_pretrained_MLP_class