软测量技术在打浆过程中的应用

在制浆生产过程中,打浆度是一个很重要的控制参数;针对打浆过程中打浆度难于实时在线测量的问题,建立了一种基于改进BP算法神经网络的软测量模型;首先对实际生产中的原始数据,经过误差剔除及滤波处理后得到一套训练数据和校验数据样本,然后采用改进BP算法神经网络进行训练,加快了网络收敛速度。得到了打浆度的非参数模型;实践表明,该打浆度的神经网络模型能对打浆度进行较精确的预测,并为后续进行过程控制和优化控制、提高打浆质量提供了良好的基础。

Application of Soft Sensing Technology in Pulp Making Process

Beating degree is an important control parameter in the pulp making.The improved BP neural-network algorithm is applied in the real-time measurement of beating degree.The relative BP neural-network soft sensing model of beating degree is established.After errors rejection and filtering of the raw data from practical processing,a set of training and checking data is derived.Then improved BP neu- ral-network algorithm which accelerates convergence speed is applied to obtain the non-parameter beating model.Effective results indicate that this beating degree model could provide the accurate forecasting,offer the fair basis for following process control and optimization control in beating,and improve the beating quality.