Overview

• When checking whether a ceramic tile is in tone or not, a specialist visually contrasts the tone of the model tile with that of the pieces coming out of the furnace

• The same printing settings can vary the final tone, as this tone is influenced by a multitude of variables that cannot be controlled (composition of the clay,humidity, convection currents in the furnace, etc.)

The Challenge

• Our customer looked for a solution that could automatically identify whether apiece coming out of the furnace was in tone or not

• Then, if the tile was not in tone, the colourimetric inputs were to be automatically readjusted, in order not to lose the production quality

  
  

The Approach

• Tone quality model: algorithm that uses colour histograms to compare the piece coming out of the furnace with the model, and check with the error margins allowed in the control set

• Print control model: Deep Learning model that changes the printing inputs when, due to some unidentified change, there is a set of tiles out of tone

The Results

• Model capable of identifying out of tone tiles with higher accuracy than expert (>0.95MBacc)