College of Engineering Unit:
HP’s Corvallis site currently utilizes multi-million dollar printing presses to produce printing jobs for large rolls of paper at up to 16 feet (4.88 m) per second. In order to begin a printing job, deep expertise from well-trained operators is required to manually select the right paper and settings such as speed of the printing, tension of the paper, drying temperature, amount of ink, etc. Mistakes in these settings can be costly and result in large amounts of waste, therefore it is vital that these decisions are made carefully. This project aims to automate the selection of paper and settings through analysis of PDF's received for printing jobs, and in the future, serve as a training model for a machine learning application to enhance on.
Currently, our application is prototyped in 3 components: the user interface, the backend rules engine, and the PDF analysis engine. Our team aimed to improve the accuracy of the PDF analysis engine this year by enhancing the application's capability to detect certain printing risks in the inputted PDF files. We assess the likelihood for three printing errors within the PDF’s: wrinkle/curl, flaking, and streaking.
-
Wrinkling/curling is a printing defect characterized by wavy and curling printed paper. It is caused by pages of large amounts of overall ink coverage, especially with widely differing coverages between opposing sides and borders. Our PDF analysis engine currently identifies this risk by searching for different hues of color and weighting their coverage according to their respective ink densities. The engine then assigns a value to the overall ink coverage and assesses the overall ink coverage and difference in coverages between the front and back pages and border to non-border areas to predict the probability of wrinkling/curling occuring on the paper.
-
Streaking is characterized by long streaks on the paper caused by large overlapping areas of ink coverage. In these areas, the paper wrinkles from the ink to the point that it hits the print head, causing the streaking marks. Our PDF analysis engine detects streaking by identifying significant areas of ink overlap between the front and back pages and areas of high saturation of red, green, and blue inks. The engine then creates a mask for the RGB regions of a duplex page to detect and extract contours, contiguous regions of those colors on a duplex page. These overlapping contours are compared against predefined parameters to provide a risk assessment based on the number of risk pixels associated in the entire image.
-
Flaking is a common defect characterized by patches of missing ink from the printed paper. It is caused high-density ink layers (especially with red, green, and blue inks), resulting in the ink to failing to adhere properly to the substrate. Our PDF analysis engine addresses this by analyzing color intensity in the HSV (Hue, Saturation, Value) color space within the PDF. This analysis estimates the ink layer thickness from the color intensity and predicts the likelihood of post-printing flaking risk.
Project Website(s):
Industry Sponsor(s):
| Attachment | Size |
|---|---|
 Expo Poster | 1.21 MB |