摘要
在数据层和决策层综合信息,采用多人工神经网络系统诊断故障。通过对全互连、前馈、BP人工神经网的学习训练,可识别不同类型的故障;将诊断任务分解为多个子任务,对每个子任务训练相应的神经网,最后将多个神经网的结果综合起来,以提高系统性能。
In this paper, a method for engine fault diagnosis is proposed, which synthesizes information at both raw data level and decision making level, and utilize multiple neural networks to improve generalization performance in classification. Different fully connected, feed forward artificial neural networks (ANN) were trained to recognize different fault types using backpropagation learning algorithm. System performance can be improved by decomposing diagnosis tasks into sub-tasks, trainning different ANNs for different sub-tasks, and then synthesizing the results of multiple trained ANNs.
出处
《模式识别与人工智能》
EI
CSCD
北大核心
2000年第3期338-341,共4页
Pattern Recognition and Artificial Intelligence
基金
EPSRC基金
国家自然科学基金
博士点基金
863计划
地质行业科学技术发展基金
北京师范大学青年科学基金
