Advanced energy-saving optimization strategy in thermo-mechanical pulping by machine learning approach

Authors: Behnam Talebjedi, Timo Laukkanen, Henrik Holmberg and Sanna Syri

Affiliation: Aalto University, Department of Mechanical Engineering, Energy Conversion and Systems, PO Box 14400, FI-00076 AALTO, Finland

Keywords:  artificial neural network; data analysis; forest industry; machine learning; refining energy simulation; thermo-mechanical pulping

Background

Thermo-mechanical Pulping (TMP) is one of the most energy-intensive industries where most of the electrical energy is consumed in the refining process. Developing an energy-efficient refining optimization and control strategy requires an advanced system identification model. Control models such as model predictive controller (MPC), which is common for controlling the refining process in TMP mills, require an identification model that can explain refining behavior based on refining variables. Therefore, there is an indispensable need to develop a model that can simulate the refining process accurately. Since refining is a time-varying process with an intricate and multivariate nature, refining energy simulation is challenging. Several factors, such as rotational speed, crossing angle, refining gap, bar dimensions and pulp consistency, influence the pulp and paper properties. Implementing all these factors into the fluid dynamic and thermodynamic equations for the simulation of the refining process requires too much computational time and effort. On the other hand, analyzing complex systems with high accuracy is facilitated by artificial intelligence (AI) development. Artificial intelligence is an alternative powerful data-based method for energy modelling of the refining process.

Our research results suggest that accurate energy simulation of the refining process is achievable by utilizing AI methods. Additionally, we propose an energy-saving refining optimization strategy by integrating the machine learning algorithm and heuristic optimization method.

Methods

• In the first step, refining specific energy consumption (RSEC) and pulp quality identification models are developed using Artificial Neural Networks.
• In the second step, the developed identification models are incorporated with the Genetic algorithm to minimize the total refining specific energy consumption while maintaining the same pulp quality and production.

Results and findings

Simulation results prove that a deep multilayer perceptron neural network is a powerful tool for creating refining energy and quality identification models with the model correlation coefficients of 0.97, 0.94, 0.92, and 0.67 for the first-stage RSEC, second-stage RSEC, final pulp fiber length, and freeness prediction, respectively.

Findings confirm that the average total RSEC reduction of 14 % is achievable by utilizing the proposed optimization method.