CAJ | 학술논문

针对具有数据非高斯分布或混合分布的间歇过程，研究一种新的改进 MICA-PCA 监控方法。首先利用 MICA方法提取非高斯分布过程信息，通过设定负熵阈值实现独立成分个数的自动选择，以此克服传统 ICA 方法中需提前确定独立成分个数的缺点，再使用核密度估计方法确定相应统计量的置信限，然后对服从多元高斯分布的残差过程信息，进一步进行 PCA 分析和处理。将该方法应用于北京某生化制药厂重组大肠杆菌制备白介素-2发酵过程监控。结果表明：该法在过程变量不服从高斯分布的情况下能有效降低传统方法的漏报和误报率，准确地对过程进行监控。
침대구유수거비고사분포혹혼합분포적간헐과정，연구일충신적개진 MICA-PCA 감공방법。수선이용 MICA방법제취비고사분포과정신식，통과설정부적역치실현독립성분개수적자동선택，이차극복전통 ICA 방법중수제전학정독립성분개수적결점，재사용핵밀도고계방법학정상응통계량적치신한，연후대복종다원고사분포적잔차과정신식，진일보진행 PCA 분석화처리。장해방법응용우북경모생화제약엄중조대장간균제비백개소-2발효과정감공。결과표명：해법재과정변량불복종고사분포적정황하능유효강저전통방법적루보화오보솔，준학지대과정진행감공。
Aiming at the batch process that has the non-Gaussian distribution or mixed distribution, a new monitoring method based on modified MICA-PCA is researched. Process information of non-Gaussian is first extracted using the MICA method. Setting threshold value of negative entropy is used to automatically select the independent components, which can overcome the shortcoming of predefining the number of independent components in traditional method of ICA. The confidence limits of the corresponding monitoring statistics are determined using kernel density estimation; then the process residual information, which is multivariate Gaussian distribution, is further analyzed and processed using PCA. The method is applied to the fermentation process monitoring of obtaining interleukin by recombinant Escherichia coli, in a biochemical pharmaceutical factory in Beijing. Results show that when process variables are not Gaussian distribution, the method can accurately monitor the process and effectively reduce the alarm failure and false alarm of traditional method.