Generating fuzzy rules by learning from olive tree transpiration measurement - ...

• Author(s): Siqueira, J.M. ; Paço, T.A. ; Silvestre, J.C. ; Santos, F.L. ; Falcão, A.O. ; Pereira, L.S.
• Date: 2014
• Persistent ID: http://hdl.handle.net/10174/9542
• Origin: Repositório Científico da Universidade de Évora
• Subject(s): fuzzy algorithm; Granier sap flow; olive transpiration; data analysis
• Document Type: Part of book or chapter of book
• Language: English

Description

The present study aims at developing an intelligent system of automating data analysis and prediction embedded in a fuzzy logic algorithm (FAUSY) to capture the relationship between environmental vari- ables and sap flow measurements (Granier method). Environmental thermal gradients often interfere with Granier sap flow measurements since this method uses heat as a tracer, thus introducing a bias in transpiration flux calculation. The FAUSY algorithm is applied to solve measurement problems and pro- vides an approximate and yet effective way of finding the relationship between the environmental vari- ables and the natural temperature gradient (NTG), which is too complex or too ill-defined for precise mathematical analysis. In the process, FAUSY extracts the relationships from a set of input–output envi- ronmental observations, thus general directions for algorithm-based machine learning in fuzzy systems are outlined. Through an iterative procedure, the algorithm plays with the learning or forecasting via a simulated model. After a series of error control iterations, the outcome of the algorithm may become highly refined and be able to evolve into a more formal structure of rules, facilitating the automation of Granier sap flow data analysis. The system presented herein simulates the occurrence of NTG with rea- sonable accuracy, with an average residual error of 2.53% for sap flux rate, when compared to data pro- cessing performed in the usual way. For practical applications, this is an acceptable margin of error given that FAUSY could correct NTG errors up to an average of 76% of the normal manual correction process. In this sense, FAUSY provides a powerful and flexible way of establishing the relationships between the environment and NTG occurrences.