论文部分内容阅读
介绍大机组快速响应智能诊断系统。该系统采用概率神经网络作为故障分类器，收敛速度为反向传播（ＢＰ）算法的２万倍，并稳定收敛于贝叶斯优化解，避免了ＢＰ网络局部最小的弱点，可以在线快速追加故障。进一步提高系统诊断能力；同时，采用智能化信号处理技术自动提取全息话获得丰富的诊断信息．极大降低对操作人员的要求，实现对大机组常见故障快速、简捷、自动的智能化诊断并减少对专家的依赖。研制出实用的“傻瓜”式智能诊断软件，已在国内多家大型石化企业装机使用。
Introduce the large unit quick response intelligent diagnosis system. The system uses the probabilistic neural network as a fault classifier, and the convergence speed is 20,000 times that of the BP algorithm. The system converges to the Bayesian optimal solution stably and avoids the local minimum weakness in the BP network. The fault can be quickly added online . Further improve the system diagnostic capabilities; the same time, the use of intelligent signal processing technology automatically extract holographic words to obtain a wealth of diagnostic information. Greatly reduce the operator’s requirements, to achieve common faults large units quickly, simply, automatically intelligent diagnosis and reduce the dependence on the experts. Developed a practical “fool” type intelligent diagnostic software, has been installed in many domestic large petrochemical enterprises use.