J. L. Valade

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.

Effect of lignin content on thermal degradation of wood pulp

Abstract A series of pulps containing between 3.6 and 23% of lignin was prepared by a careful delignification of a high-yield bisulfite pulp. The pulps were subjected to isothermal pyrolysis in a Perkin-Elmer TGS-1 thermobalance. The measurements were carried out at 8 different temperatures from 325 to 360°C under nitrogen atmosphere. The results obtained indicate that the effect of lignin on degradation depends strongly on temperature. Below 330°C, the rate of degradation varied only little with lignin. This variation becomes more important at temperatures above 330°C in that the rate of degradation increases with decreasing lignin content. The apparent activation energy of degradation ranges from 41.4 kcal mol−1 at 23% of lignin to 67.0 kcal mol−1 at 3.7% of lignin.

A composite representation of pulp quality

Abstract Principal component analysis (PCA) performed on data from pilot scale chemimechanical (CMP) pulping trials indicates that over 90% of the variation in 11 handsheet properties can be expressed in terms of only 4 composite variables. The relaionships between the original properties and the new composite properties are displayed graphically using a biplot of the PCA scores. These techniques are combined to construct a chart capable of monitoring changes in a large number of quality responses simultaneously. Model accuracy can be monitored on the same chart resulting in a tool which will self-diagnose as well as offer a powerful overview of the process.

Effect of molecular weight of polystyrene on heat capacity and thermal transitions

Abstract Differential scanning calorimetry was used to study the effect of the molecular weight of well-defined polystyrene samples, the effect of heating and cooling rates, sample weight on the specific heat under and above the glass transition temperature Tg. It has been found, that the Cp of polystyrene was indirectly proportional to molecular weight increased up to M n = 100,000. The Tg was increasing with molecular weight up to M = 300,000 and then levelled off which was in agreement with data obtained for activation energies and rates of polystyrene degradations. Another weak transition, “third order” T1.1 has been observed after the samples were quenches from 250°C to the temperature of ice water. The effect of M on T1.1 was similar to the one on Tg. After levelling off ( M = 300,000) the T1.1 was 178°C and Tg = 108°C.