Smart Manufacturing Controls Drastically Reduce the Need for Toxic Chemicals in Pulp and Paper Washing
Project Objective
To develop a machine learning (ML) soft sensor to enable process controls for energy efficient brownstock washing.
Benefits to Nation
• The technologies developed in this project will provide the pulp and paper industry with a unique opportunity to both reduce cost and improve environmental sustainability. • The new control strategy developed in this project can be applied to other washing processes and will be reusable for modeling and control across a wide range of energy-intensive manufacturing processes.
Benefits to Industry
• The technologies developed are reusable for modeling and control across a wide range of energy-intensive manufacturing processes in pulp and paper. Potential savings of $180k per washer annually from defoamer o usage reduction. Translating that savings to the >1,000 washers in operation nationally results in savings of $180 million annually for the pulp and paper industry. Similarly, savings of $35k per washer annually from water and energy o usage reduction translates to $35 million in energy and water savings for the industry. • Train engineering students on Smart Manufacturing in the areas of advanced machine learning sensing, data analytics, and model-based control of manufacturing processes.
Technical Approach
A statistics pattern analysis (SPA) framework is used to build a statistics pattern analysis machine learning (SPA-ML) based soft sensor. Soft sensor algorithms are implemented to measure entrained air (foam) content.The data acquired via the soft sensor is then used to enable multi objective model predictive control for brown stock washing.
Accomplishments
- Developed and deployed SM solutions for pulp and paper manufacturing that are reusable for modelling and control across a wide range of manufacturing processes.
- Reduced defoamer, water, and energy consumption in the pulp washing process.
- Passive trials demonstrated 60% defoamer usage reduction by implementing machine learning based process controls. The controls enable a reduction in defoamer usage from 0.17 kg learning defoamer/ton pulp to 0.066 kg defoamer/ton pulp. This reduction in and defoamer usage translates to a cost savings of $37.50/hour per washer.
- Active trials demonstrated a 72% reduction in defoamer usage, which translates to a savings of $43.50/hour per washer.
Deliverables
- Delivered Machine Learning enabled soft sensor algorithms to monitor entrained air content.
- Delivered multi-objective model predictive control solutions for brownstock washing
- Tested and validated the effect of reducing defoamer flow while monitoring entrained air content and drum speed at Rayonier.
Reusable Outcomes
- Machine learning based soft sensor for entrained air content for pulp and paper manufacturing. The sensor is reusable for modelling and control across a range of manufacturing processes.
- Soft sensor mathematical model and method for detecting entrained air content.