What is a virtual measurement?
Virtual measurements, also known as soft sensors, are data models generated from machine learning technologies that use various process variable data points to produce inferred calculations of paper property measurements. Adding an online measurement component to traditional paper testing methods provides a data-based estimate of a pulp or paper property that cannot necessarily be measured in real-time or by a physical sensor.
Online measurements should be used in addition to QCS and laboratory measurements for improved quality control. While both traditional online and lab measurement methods provide valuable insight, you do not get online visibility of sheet strength during the papermaking process without incorporating virtual measurements.
How does a virtual measurement determine strength?
Better quality control depends on process optimization. Online virtual measurements are created from the combination of available online and/or offline process and product measurements to give visibility into process performance and improvement areas.
First, a strength model is created using historical machine data to establish an initial expectation of the accuracy of the calculated strength. Multiple data modeling stepsare followed, including synchronizing data based on time as well as cleaning, filtering, validating and more.
Once the accuracy of an initial model is confirmed, it is implemented for a site-specific online calculated strength measurement based on a mill's paper machine configuration. The virtual measurement calculation is further improved by gaining a better understanding of machine operation, systematic model refinement, and the implementation of measurement suspension events during periods of un-modeled machine operation. Ongoing performance monitoring is done to maintain and optimize the strength calculation.
Variables that impact strength
• Conditioned weight • Machine speed • Freeness • Sheet width • Rush drag • Furnish mix • Refining • Chemical additions • Other machine-specific variables
