【摘要】 对生产过程中的报警事件进行预测能够预测危险工况,指导操作人员实施相应的措施,从而避免危险事故的发生。论文提出了一种基于贝叶斯网络模型(Bayesian network)的报警事件预测方法,首先通过历史数据提取报警事件序列,分别建立单变量和多变量报警事件的贝叶斯网络,采用期望最大化(EM)算法和贪婪搜索算法相结合来确定贝叶斯网络的参数与结构,通过概率推理对报警事件进行预测。实例仿真表明,该方法可以有效地挖掘历史数据信息,实现准确的报警事件预测。更多还原

【Abstract】 The alarm event prediction in production processes can predict dangerous working conditions and guide the operator to implement appropriate measures to avoid accidents.In this paper,a Bayesian network-based method of predicting alarm events is proposed.In this method,the sequence of events was extracted from the historical data.Then the establishment of Bayesian networks for single variable and multivariable alarm events was carried out.The expectation maximization(EM) algorithm and the greedy search algorithm were combined to determine the parameters and structures of Bayesian network.Lastly the alarm events were predicted by means of probabilistic reasoning.The simulation results show that the Bayesian network-based method can mine the historical data effectively and be able to predict the alarm events accurately.更多还原