Simulation du problème de colmatage des drains en conditions contrôlées pour la production de canneberges

Authors: Dabo, Souleymane
Advisor: Gallichand, JacquesGumière, Silvio José
Abstract: Quebec is one of the largest cranberry producer in the world. The substantial yields in cranberry production in Quebec is due in part to optimized management of irrigation and drainage conditions. However, it has been estimated that 75% of the cranberry fields in Central Quebec have drainage problems. Drainage system design problems result in 25% reduction in cranberry yields drain clogging issues are responsible for a 39% reduction. The present project proposes to predict the degree of clogging based on water table heights at the mid-spacing and above the drain in addition to the water table drawdown time. To do this, predictive models using supervised learning and multiple linear regressions are used. The drainage experiments were conducted in the laboratory using a drainage simulator to analogically reproduce the degree of drain clogging. The experimental data collected at the end of each experiment served, among other things, to generate the input variables of the predictive models. The results show that an increase in the degree of clogging leads: (i) to an increase in water table drawdown time and the drain entrance resistance, (ii) a decrease in the equivalent hydraulic conductivity of the drainage system and (iii) the accumulation of water in the vicinity of the drain. The adjusted supervised learning predictive model explains 98.9% of the variation of the clogging degree, while the regression model explains only 58.9%. With the introduction of the equivalent hydraulic conductivity of the drainage system, the quality of the supervised learning model goes from 98.9 to 99.8% and that of multiple linear regressions from 58.9 to 83.8%.
Document Type: Mémoire de maîtrise
Issue Date: 2019
Open Access Date: 29 November 2019
Grantor: Université Laval
Collection:Thèses et mémoires

Files in this item:
Description SizeFormat 
35740.pdf2.82 MBAdobe PDFThumbnail
All documents in CorpusUL are protected by Copyright Act of Canada.