Abstract: To better utilize historical alarm information of industrial processes, this paper introduces a belief rule base model based process alarm time series prognosing inference approach which is able to evaluate the process safety performance in the future. The belief rule base model involved is established using historical alarm data of process variables, while a particle swarm optimization algorithm is used for the model parameter learning. The online implementation of the model can help predict trends of the process alarm states in the future. A numerical simulation and industrial process alarm data are used to demonstrate the effectiveness of the approach with satisfying prognosis results.

Key words:  , Belief rule base models, time series, particle swarm optimization algorithm, process alarm states, prediction

摘要： 为了更加有效的利用流程工业中的各种历史报警信息，对未来时刻的系统安全性进行评判，提出了一种基于置信规则库推理的过程报警时间序列预测方法，利用过程变量的历史报警数据建立置信规则库模型，采取粒子群算法进行模型参数学习。该预测模型可以对未来一段时间过程可能产生的报警状态变化趋势进行预测。通过数值实例仿真及实际工业过程报警数据对该模型进行验证，得到了较为满意的预测结果。

关键词: 置信规则库模型, 时间序列, 粒子群优化算法, 过程报警状态, 预测