中文摘要：

为提高炼化过程的安全性，有必要对工艺过程参数故障趋势进行预测及关联验证。利用特征数据分割方法提取工艺参数的特征点与特征线段，对数据分割片段进行线性趋势拟合，通过分割片段趋势结合规律得到趋势预测结果；根据炼化工艺参数的内在关联，研究炼化过程中工艺参数的拟合关系，对故障趋势预测结果进行验证，通过关联参数验证步骤确定预测结果的可信度。研究结果表明：该方法通过对工艺参数的分割、拟合和预测，得到系统的故障趋势，与现场数据相符。

英文摘要：

For improving the safety in refining process, it is a must to predict fault trends of process parameters. Considering problems of destroying integrity of data unit and data isolation in traditional trend prediction methods, fitting relationships between process parameters were studied. A characteristic data partitioning algorithm based on internal relationship between refining technologic parameters was worked out. Combined with warning information, the method was used to analyze and predict possible fault trend. Using the method, a case study was made on a reflux tank in a certain refinery in China. It was shown that fault trend prediction result of tank conforms to the field data.