Smart System to Automate Control of a Million Dollar Printing Press

HP’s Corvallis site develops and produces multi-million-dollar printing presses larger than a shipping container that print from large rolls of paper at up to 16 feet per second. Today, a lot of expertise is required to select the right paper, prepare the jobs for printing, select the right settings, and operate the press successfully to produce valuable printed product instead of tons of waste paper.

We want to automate the selection of paper and settings used for a job and control the operation of an HP PageWide press, something that today requires the deep expertise of well-trained hard-to-find staff.

We’ll work on several use cases. One has inputs of the PDF to print, the paper to use, and press information. The outputs control many aspects of press operation, like paper speed, dryer temperature, and paper tensions. We liken this project to figuring out how to make a press as simple to use as clicking the green button on a copier.

The project goal is to refine and prepare for productization a system that encapsulates this expertise and makes it available in multiple ways. One core deliverable is a set of fully automated and high performance PDF content analytics methods that look for specific characteristics that, when present, require the press to operate with specific settings.  Another deliverable is a flexible microservice “rules engine” and rules that encode expert knowledge as executable rules and can be updated without deploying new software.     Another deliverable is a user-friendly UI and application logic to collect inputs, present outputs, and orchestrate operation.

Objectives

At a high level, the overall objective is to evolve and refine the existing implementation developed over two previous Capstone projects and move it towards high quality and performant code that can be directly leveraged into product release.  Along the way, key objectives are to refine the technical approaches and tools, develop tests to measure and optimize capabilities and performance, and iteratively develop these capabilities in collaboration with HP domain experts who will test and provide feedback based on real-world projects and production runs.

The project can be organized in waves of modular deliverables that build upon each other and can be split among people if desired. We can scale the scope according to the number of people involved.  The milestones can evolve by mutual agreement as things progress. We envision the team will work using an agile methodology, including sprints, demos, and both feature and story backlogs, using us sponsors as the “product managers” to provide feature and priority guidance.

There are three areas of work:

  • For the PDF analysis routines, we have C++ code that needs to be enhanced and tuned for capability, performance, and a robust set of automated tests.  
  • For the microservice, we have an innovative nodejs Javascript-based rules-engine that needs refinement and to be extended to support additional rulesets and a test suite. 
  • For the application, we have a modern React-based application UI experience with responsive design and we need to expand it with more capability and fix a few existing issues.  It has a headless application that orchestrates operations and needs further refinement and testing.

Initially:

  • Break down the project into a roadmap and intermediate milestones.
  • Develop familiarity with the problem domain by characterizing concisely a broad set of expert knowledge extracted from white papers, training materials, guides, and true human experts.
  • Understand the existing codebase, for which there are code walkthrough videos and other documentation.
  • Investigate technical approaches and available open source tools. Do experiments to gain familiarity.
  • Outline a technical path forward, backed by developed and demonstrated expertise.
  • Assemble the necessary development and deployment environments, which includes a cloud-based deployment using Docker and multiple containers.

With the fundamentals in place, development will be iterative and in collaboration with HP technical and product experts.

  • Improve the capability in collaboration with experts who will test and provide feedback based on real- world projects and production runs.
  • Provide increased range of input use cases to receive the job request
  • Provide improved input data analysis, looking inside the PDF to extract important characteristics of the content be printed, or retrieving technical characteristics of specified paper(s)
  • Provide better recommendations and explanations using automated rule-based reasoning, that converts inputs into outputs in one or more steps.
  • Provide more robust testing and performance optimization, etc.
  • Output results that show the recommended settings along with a textual explanation
  • Gather evaluation feedback, that captures responses from experts, both based on expert knowledge and actual production results, and is used to improve the system.
  • In all areas, apply a thoughtful technical architecture and implementation.
     

Motivations

Today, a lot of expertise is required to select the right paper, prepare the jobs for printing, select the right settings, and operate the complex HP press successfully to produce valuable printed product instead of tons of waste paper.

We want to automate the selection of paper and settings used for a job and control the operation of an HP PageWide press, something that today requires the deep expertise of well-trained hard-to-find staff.

Qualifications

Minimum Qualifications:

Software development skills

Preferred Qualifications:

nodejs, React, html5, Java, Javascript


Details

Project Partner:

Pieter van Zee

NDA/IPA:

NDA/IP Required

Number Groups:

1

Project Status:

Accepting Applicants

Website:
http://hp.com/go/pagewidewebpress
Video:
https://www.youtube.com/watch?v=k6nCD9ZUGz0
Keywords:
AI expert system rules engine microservices UI React nodejs java javascript
Card Image Capstone