改进BP神经网络在流型判别中的应用

提出了一种应用于人工神经网络进行流型判据的新方法。采用了自适应梯度下降法和改进模拟退火法等措施，加快了ＢＰ网络的收敛速度，增强了其跳出局部检小值的能力，提高了神经网络模拟非线性系统的能力，并分析了可作为流型判别用的神经网络的输入参数、输出参数。为了证实该方法的可行性，应用前人的实验数据进行验证。证实该方法是一种十分有效的两相及相流流型判别方法。

The Application of an Improved BP Neural Network in the Discrimination of Various Flow Patterns

A new method for the discrimination of flow patterns is proposed in this paper, which is based on the use of an artificial neural network. With the help of a self adaptive gradient reduction method and an improved simulation annealing approach, etc the convergence rate of a BP network can be accelerated. There emerged, as a result, an enhancement in the network's ability to avoid a local minimum magnitude, which contributes to an increased capability of the network to simulate a nonlinear system. In addition, an analysis was made of the network input and output parameters, which can be used for pattern discrimination. To demonstrate the feasibility of the recommended method, the authors have employed the experimental data of scholars forerunners to check and verify their work procedures. It has been proved that the above method can be assessed as a very effective one for the discrimination of two phase and multi phase flow patterns.