A Tool for Process Optimization: Neural Network Software

A continuing program of process optimization is the key to high asset productivity. The increasing complexity of processes makes this more difficult to achieve, but it also provides more opportunities and rewards. Deinking process optimization has become more important recently due to increasing complexity (more process steps, reduced water usage and increased water recycle) and sudden economic changes (dramatically higher wastepaper costs and deinked pulp prices). Neural network software, when used by an experienced process engineer, can sift through the large amounts of time-stamped data available in modern computer control systems and find process variable interdependencies. Since the neural network ranks and quantifies these “cause and effect” relations, this information becomes the basis for making process adjustments to optimize operations. The neural network model can be used to test these adjustments before making changes to the actual process. This paper provides information on the neural network analysis used to optimize the deinking process at Alabama River Newsprint in Claiborne, Alabama where deinking chemical costs were reduced by 32%.