Creation of paper property prediction models using cartesian genetic programming

In this study, a method to predict the properties of finished paper based on information provided from pulp suspension was examined. In the experiment performed, Cartesian genetic programming (COP), multiple linear regression (MLR) and neural network (NN) methods were employed to predict paper properties. A total of 9 types of pulp suspension were employed as samples, and 47 properties of fibre in suspension state were evaluated. Handsheets made from each type of pulp suspension were then evaluated in terms of four optical properties. Using the pulp fibre and optical property data thus obtained, the prediction performance of each method was evaluated. The results showed that the use of models created using Cartesian genetic programming enables more accurate estimation of paper properties from pulp fibre properties than other methods.