Predicting the Failure of a Thermal Print-Head Resistor

  • Phillip Lakin

Student thesis: Doctoral Thesis

Abstract

Thermal transfer printers print variable images onto packaging such as the
Best-Before-End date on a crisp packet. These images degrade as the resistors
within the Thermal Print-Head begin to fail. Detecting a failed thermal resistor is
a simple process but predicting its failure has never been attempted before. This
research developed the first method of capturing the Thermal Print-Head signals
and generated a feature set that is used in a learning algorithm to successful predict the failure of these thermal resistors.
The electrical data from 504 failed resistors across six Thermal Print-Heads
is captured. A novel approach to using a Forward Looking Infra-Red camera,
to capture the extremely fast temperature profiles of a thermal resistor, helped to back up the real-time temperature calculations. The actual failure point of a thermal resistor, and the effect this has on a printed image, has never been captured until now and the results are also presented here. The features generated have been used in three regression algorithms. Support Vector Regression, Multiple Linear Regression and k-Nearest Neighbour algorithms are compared. K-Nearest Neighbour achieved the best results by predicting the failure point to an impressive 200m or less. For many production facilities with a high throughput, this prediction will be accurate to within two hours. Another impressive output is the algorithms ability to ensure the majority of the predictions are early, which will give a timely warning.
Knowing when any resistor will fail will give the printer’s operator the ability
to change the Thermal Print-Head at the most cost-effective point. For each consumable change, it is now conceivable that the printer can display any degraded effects of the printed image during the life of that consumable. This algorithm will give the manufacturer the tool they need to prevent any more unreadable images reaching their products.
Date of Award24 May 2019
Original languageEnglish
Awarding Institution
  • University of Portsmouth
SupervisorEdward Smart (Supervisor)

Cite this

'