在分析了生产过程两类故障的特点和常规故障诊断方法的缺陷的基础上 ,提出了生产过程早期故障检测与诊断方法。利用故障发生后过程系统中相关变量或征兆信息之间的解析冗余关系 ,将它们的动态趋势信息通过人工神经网络进行融合 ,避免了常规方法故障检测与诊断的滞后性。理论分析和仿真表明了该故障诊断方法的有效性。 Based on the analysis of the characteristics of the two types of faults in the production process and the defects of the conventional fault diagnosis methods, the early fault detection and diagnosis methods in the production process are proposed. Using the analytic redundancy relationship between the relevant variables or symptom information in the process system after the fault occurs, their dynamic trend information is fused by the artificial neural network, which avoids the lag of the fault detection and diagnosis of the conventional method. Theoretical analysis and simulation show the effectiveness of the fault diagnosis method.