AI-driven solution for in-line inspection

THE CHALLENGE

Company X is a global manufacturer and service provider in the pulp and paper industry (among other industries). The Company approached materialsIN, www.materialsIN.com, a materials informatics company, with a materials and process-related problem involving a tissue creping process--a problem that is pervasive in the entire paper industry.

The process involves a very complex methodology in which chemicals sold by Company X are sprayed onto the equipment surface. The chemical selection process, ratio of chemicals fed, and volume of chemicals fed is often a very difficult and technical process in which engineers study a machine during a machine audit and make recommendations for a coating “package” that allows the widest window of operation for the given attributes of a machine (speed, temperature, grades produced, quality desired, etc). Engineers will select the package through experience (mostly) and polymer physical and crosslinking properties in order to maintain the optimum coating “set up” at the creping blade.

THE MATERIALSIN SOLUTION

materialsIN Pro-Opt is a product that monitors, reports and provides prescriptive suggestions on processes, in this case a tissue creping process. It does this by gathering, organizing, and analyzing large amounts of data captured from the sensors on the equipment.

The end result takes the art out of creping and replaces it with hard data science--allowing the operators to have more insight into their machine performance.

Many key attributes, from feedrates to physical properties (hardness, mass, moisture, ash content, etc) to chemical usage to tissue quality, will be monitored through direct sensors and through data inference.

The primary objective of Project A are:

to provide the user with recommendations for the correct feedrates and the correct time to change a blade;
to utilize key learnings to provide further feedback to the operators for proper coating feedrates AND wet end setpoints in order to maintain tissue quality;
to identify process variables, ideally adhesive flow rate or release flow rate, that can be adjusted during the blade life in order to 1) minimize sheet quality impacts and/or 2) extend the life of a blade by avoiding going outside of designed sheet quality specs for a longer period of time.

MATERIALSIN METHODOLOGY

materialsIN leverages a proprietary data-driven methodology, which has been developed over 40 years by Dr. Krishna Rajan, a leader in the space. The company uses machine-learning to expedite knowledge/development of materials; it harnesses statistical learning methods with data and physics-based computational techniques information to solve customers’ materials-related issues.

materialsIN collects, assesses, and organizes data (the quality of which determines the effectiveness of the analysis) and employs a proprietary data selection, processing and transformation process to handle complex data problems.

The methodology, as shown below, consists of five steps: (i) data selection, (ii) data preprocessing, (iii) data transformation, (iv) data mining, and (v) interpretation of analysis results. The result is a data-analytics approach that empowers customers to select the best candidate materials for their products, as it: establishes new correlations, identifies outliers, enlarges and evaluates databases, and formulates predictions.

PARTNERSHIP STRUCTURE

Company X leverages its sales teams and marketing reach to take the private-label solution, powered by materialsIN, to its global customer base in the paper industry. Company X will partner with materialsIN on additional products as it has identified a considerable list of materials/process-related issues which confront both the Company and industry.