Neural network based double-objective optimization and application to pulp washing process improvement

For a countercurrent paper pulp washing system, the requirements of craft to residual soda in the final washed pulp and Baume degree in the first stage filtrate tank are usually inconsistent. To compromise this pair of contradiction, a neural network (NN) based two-objective optimization algorithm is proposed. Two NN models of the residual soda in washed pulp and the Baume degree in the first stage filtrate tank are obtained by a two-step identification method. An external penalty function method based on the double-objective optimization algorithm on the hot clean water input and final washed pulp output is employed. In terms of the DCS development platform of Xinhua XDPS 4000, China, a novel pulp washing process DCS is implemented and put into operation in paper mills in China with notable profit.