基于神经元网络的钻杆热处理温度信号校正

Neural-Network-Based Temperature Signals Tuning for Pipe Heating Process

神经网络是非线性系统建模的重要方法。反向传播 (BP)算法常常用于神经网络的权值训练中 ,但是BP算法收敛慢。为此 ,将非线性最小二乘法用于前馈神经网络的权值学习。采用这一建模方法对石油钻杆在热处理过程中的温度测量偏差进行校正。研究结果表明 ,该方法具有很快的收敛速度和很好的拟合精度 ,适用于工业过程中测量信号的在线校正。

Neural Networks is an important method in nonlinear system modeling. Back propagation (BP) algorithm is often used for the weights training of neural network, but the convergence speed of BP algorithm is slow. In this paper, nonlinear least squares method is used for the weights training of feedforward neural networks. Then the modeling method is applied on the tuning of temperature measuring error during the processes of oil pipes are heated. The research result shows that the new algorithm has fast convergence speed and good precision. The method is suitable for the on line tuning of a measured signal on industrial plant.