Robert Lanouette

Process modeling with neural networks using small experimental datasets

Abstract This paper reports some work done to improve the modeling of complex processes when only small experimental data sets are available. Various solution strategies based on feed-forward and radial basis function (RBF) neural networks have been tested for three problems including two wood pulp applications. Experimental data sets obtained from D-optimal design and from a random selection throughout the experimental space were compared for their ability to lead to the proper model. In addition, the influence of activation functions, the number of levels in stacked neural networks and the composition of the training data sets have been studied. The study shows that designed training data sets are more desirable than random experimental sets, due to their higher orthogonality. The use of neural network is a powerful tool for modeling complex processes even when only a small set of data is available for training. However, special care must be exercised to insure that good predictive models are obtained.

Comparison of two multicriteria decision aid methods: Net Flow and Rough Set Methods in a high yield pulping process

Abstract This investigation presents a synthesis of two multicriteria analysis methods, Rough Set Method (RSM) and Net Flow Method (NFM), applied to the multicriteria optimisation for the manufacture of paper using jack pine as the source of fibres. This work is the result of a collaboration between different Canadian and French laboratories. The two optimisation methods, based on different approaches, are applied to the same Pareto domain of non-dominated operating conditions. The Rough Set Method (RSM) uses a set of decision rules that are based on the preferences of experts, when presented with a small set of diverse conditions extracted from the Pareto domain. These rules are then applied to the entire Pareto domain to determine the preferred zone of operation. In the Net Flow Method (NFM), the preferences of experts are defined with three threshold values and one set of weights that are used to classify the entire Pareto domain. The NFM is a hybrid of two methods between ELECTRE and PROMETHEE. To compare these two methods, they were used to optimise the identical process. Results clearly show that the two methods gave nearly identical optimal solutions and well within inherent experimental errors.

Multicriteria optimization of a high yield pulping process with rough sets

Abstract The optimization of complex processes usually involves many competing objectives; in this case there is typically no solution that yields optimal values for all of the objective criteria and the decision-maker must therefore find a reasonable compromise. In recent years, new multicriteria methods have been developed to assist the practitioner in achieving a judicious compromise among the various competing objectives. One method, the rough set method (RSM), is able to encapsulate the preferences of an expert within a simple set of logical rules that are used to rank a large number of feasible solutions according to these preferences. The RSM was used in this investigation to determine the optimal operating region of a high yield pulping process. This pulping process has seven input variables that can be manipulated to optimize four objective criteria characterizing the product: brightness, specific refining energy, extractive content, and rupture length. Results show that an optimal solution zone can be easily defined and zones of decreasing preference can be drawn. This information is very useful to the practitioner for choosing the desired operating conditions and for analyzing the robustness of the preferred operating scenarios from a process control point of view.

The breakdown mechanism of earlywood and latewood in refining

The principal objective of this research project is to study the breakdown mechanism of Jack pine (Pinus banksiana) earlywood (EW) and latewood (LW) in thermomechanical pulping by means of microscopic observations. Characteristics such as fiber splitting, shortening, delamination (internal fibrillation), external fibrillation, etc. are evaluated. Physical changes in the EW and LW fibers are qualified and quantified with the aid of light microscopy as well as scanning electron microscopy. The impacts of the observed changes on pulp and paper properties are assessed to establish possible interrelation between the fiber characteristics and paper properties.