Using Neural Network Models to Affect the Bottom Line for Linerboard Applications, 2003 Spring Technical Conference Proceedings
The use of neural network based “virtual sensors” for predicting paper strength properties has been reported on numerous occasions beginning as far back as 1990. 1,2,3 All of these papers discuss the potential for improved
operations and potential savings based upon the use of these models. However, we have very few cases that have been presented that discuss the important final steps being taken to address the issues associated with using these models to actually achieve cost savings potentials in real time.
This paper will discuss the final steps required for developing a model into more than just a prediction model. These are the steps that make the model a true online tool that can positively affect the operational bottom line. In addition, some actual examples will be presented that represent actual savings opportunities ranging from $300,000 to $800,000 per year.
Using Neural Network Models to Affect the Bottom Line for Linerboard Applications, 2003 Spring Techn