Hyperspectral Imaging in the Paper/Pulp Industry: A Review
Before the advent of paper, permanent records were exceedingly difficult to produce. Stone carvings, clay tablets, and parchment (animal skins) were the norm. These methods required significant investments of time and resources to produce (and in some cases store) and were therefore difficult to scale. Even our understanding of life before the dawn of paper has been impacted by these inadequacies. Indeed written records, and our understanding of humanity and its history, exploded after 100 A.D. when the Chinese invented what is now known as paper. The increase in retention and propagation of knowledge increased exponentially with the addition of the printing press in 1440 which unleashed the full potential of paper as a resource.
Today paper remains an integral resource in our daily lives. Many people don’t even realize their reliance on paper or the process which is used to make it. Let’s take a look at the major steps in that process now.
1) Start with wood chips
2) Digest w/ heat & acid to make pulp
3) Wash & bleach pulp
4) Refining/sizing fibers
5) Coloring fibers
6) Diluting fibers
7) Wire/forming fabric
11) Quality control
That last step, quality control (#11) is the process I’d like to focus on. This is the step that informs all the previous steps, and ultimately determines the quality of the product that is produced. So what does the quality control of paper entail?
The quality control of paper included both physical and chemical components. The three major components are:
That might seem simple enough, but in fact, these three parameters are highly interdependent and arise from a multitude of intricacies present within the steps of the paper-making process listed above. This is clearly illustrated in the diagram from Shen et al below.
This issue of interdependency and complex origins requires decoupling algorithms. However, three problems with this method prove hard to overcome
1) Modeling uncertainty
2) Process complexity
It’s not that algorithms aren’t the answer, they are. But you need the correct algorithms, and you need them in the correct locations. So let’s look at what WOULD work in paper production.
An artificial intelligence-aided quality control algorithm could detangle the weight/moisture/ash problem with adequate inputs. That’s because AI-aided systems are capable of analyzing complex relationships and identifying trends that are hidden behind low or overlapping signals. If, say, inputs for thick stock flow, steam pressure, and filler flow were fed into these algorithms alongside weight, moisture, and ash readings, the relationships between these factors and outcomes could be more reliably predicted.
By replacing paper weight with a compositional analysis compatible with an optical-based method it would allow for the use of AI-assisted hyperspectral imaging which would provide additional benefits still.
1) Real-time analysis
2) Analysis at various points of the process
3) 2D-spatial distribution data
Overall, this approach would provide a better understanding of what’s happening throughout the papermaking process and allow for better control of the final product. Ultimately this will result in higher quality and reduced cost of production. For these reasons, it would be pertinent for paper production facilities to invest in the implementation of AI-assisted hyperspectral imaging for the quality control of the paper.