Using genetic algorithms and Neural Networks to predict and optimize coated board brightness

The final brightness of coated paper is dependent upon a large number of variables, including coating machine settings and the coating formulation. In addition all these variables are interdependent upon each other. The large number of variables and strong interdependencies among them makes the task of modeling and optimizing the coating process using traditional mathematical techniques difficult and cumbersome. This work used Neural Networks to predict the final brightness of the coated sheet. 100% of the Neural Network predictions were within 5% error. The Neural Network model was then integrated with another artificial intelligence tool, Genetic Algorithms, for maximizing the top coat brightness at minimum cost. Genetic Algorithms were able to successfully optimize the model built by the Neural Network, coming up with optimal solution which can reduce cost of formulation by almost 64%. The solution was technically feasible and conformed to our understanding of the paper coating process.