ニューラルネットワークを用いた音響診断によるポンプ異常の検出
Detection of abnormality of pumps by acoustic diagnosis using neural network
著者:
角皆 学 Manabu TSUNOKAI 高瀬 健太郎 Kentaro TAKASE 萱田 良 Ryo KAYATA
発刊日:
公開日:
Acoustic diagnosisBearingCondition monitoringMisalignmentNeural networkPump
概要
A method that applies the neural network to acoustic data to detect abnormalities of horizontal pumps is proposed. In the experiments, running sound of a pump with bearing damage or misalignment was acquired from various distance while another pump in the same room was running or not running as a possible noise source. By applying the neural network to the appropriately processed frequency spectrum of acquired data, abnormality can be accurately detected regardless of the distance between the sensor and the pump or presence of noise. Learned neural networks also showed high interpolation and extrapolation characteristics to magnitude of abnormality and the distance. In addition, by combining the learned neural networks for each of the bearing abnormality and the misalignment, the state of the pump was accurately indentified.
