Full Conjugate GradientAlgorithm in the

Neural Network Soft Measurement

System of Dynamic Flow

Wang Yiquan

（Institute of Engineering of Machinery, Yanshan University, Qinhuangdao 066004）

Tang Yong

(Institute of Science and Engineering of Information , Yanshan University, Qinhuangdao 066004)

Jiang Wanlu

（Institute of Engineering of Machinery, Yanshan University, Qinhuangdao 066004）

Wang Hongyan

（Institute of Science and Engineering of Information, Yanshan University, Qinhuangdao 066004）

Abstract: To solve the measurement of dynamic flow in hydraulic pressure servo systems by applying neural network is an important and difficult work in the field. A new BP algorithm, full conjugate gradient algorithm, is discussed in detail. Its betterment was presented, too. Its capabilities are validated by experimentations finally. The work has extensive referenced value for BP neural network.

Key words: Dynamic flow  Neural network  Three layer perception  Full conjugate gradient algorithm  Learning factor

CLC No: TP183

国家自然科学基金资助项目(60374042). Received 20040910, received in revised form 20050208