ニューラルネットワークによる導管網漏洩位置推定

概要

論文の詳細を見る
We implemented the classification procedure using the multilayer perceptron (MLP) on the leak localization problems of fluid pipe networks. In our implementation, we first segmented the whole pipe network into a finite number of regions or pipe groups. Then we trained the MLP to produce the relationship between the measurement values and the leak region. Because an MLP for 1 out of M classification problems approximates the Bayesian posterior probability, we can estimate the leak probability for each region. Using this probability value we updated the segmentation to improve the resolution around the leak position. We also considered the allocation of measurement points, and implemented a genetic algorithm procedure to improve it. Numerical simulation results are presented to show the validity of these algorithms.
社団法人日本機械学会の論文
1996-03-25