中文摘要：

针对基于k近邻的故障检测方法（Fault Detection method using the k-Nearest Neighbor rule，FD—kNN）的在线实时监测需预估当前时刻之后的采样数据，检测性能会受到预估精度影响的问题，对FD-kNN进行扩展以适用于批次过程的实时监测。该方法根据每个采样时刻的历史数据进行建模，并根据这些模型实时监测批次过程。该方法不需要预估数据，避免由于预估误差大而带来的误报和漏报问题，同时较好地继承k近邻法则（k-Nearest Neighbor rule，kNN）在处理非线性、多模态和非高斯等问题上具有的优势。青霉素发酵过程的仿真试验验证该方法可行。

英文摘要：

The sampling data after the present moment should be predicted while the Fault Detection method using the k-Nearest Neighbor rule （FD-kNN） is applied in the online real-time monitoring, and the detection performance could be affected by the pre-estimation accuracy. The FD-kNN method is expanded to be applied in the real-time monitoring for batch processes. Some models are built by the method according to the historical data at each sampling time, by which the batch processes are monitored in real-time. It does not need to predict data in the method, and so the false alarms and missing detections can be avoided due to the large prediction error. Moreover, the method inherits the advantages of k-Nearest Neighbor rule （kNN） in dealing with the nonlinear, muhimode and non-Gaussian problems. The simulation test on penicillin fermentation process shows the feasibility of the method.