Chemical Engineering Department, ETSEQ, University Rovira i Virgili, Av. dels Països Catalans 26, 43007 Tarragona, Spain
Information Technology Laboratory, Cinvestav Tamaulipas, Scientific and Technological Park TecnoTam, 87130 Victoria, Tamaulipas, Mexico
TIRME S.A., Carretera de Sóller Km 8.2, 07120 Palma de Mallorca, Islas Baleares, Spain
This paper describes a study of operational parameters by using the multivariate data analysis and neural networks for a municipal waste incinerator located in Majorca (Spain). The basis of the study also includes the chemometric techniques: linear multivariate regression to develop a model with certain predictive capabilities; linear principal component analysis, which allow the number of variables to be reduced from 17 to 4, thus fostering visualization in a low-dimension space; and linear discriminant analysis to categorize plant data accordingto the month (probability ≈ 70%). Neural network predictive capability was good, with relative errors around 6-8%. These techniques allow all the variables to be analysed simultaneously and focus on the variables which have a significant impact. In this way, the interrelationships between sets of variables, causal relations among input/output variables, seasonal motivated deviations as well as observation variations have been identified.