Study On Fault Diagnosis System For Papermaking Wastewater Treatment Process Based On Principle Component Analysis

Process monitoring, diagnosing and further removing faults during the productionprocess, is rather meaningful to improve the production efficiency, product quality and thesafety of staff and equipments as well.In this research, two Fault Diagnosis (FD) Models (FDM) based on principal componentanalysis (PCA) for two different scales of papermaking wastewater treatment process, labscale and plant scale, are developed separately.During the development of FD model for the lab scale papermaking wastewatertreatment process,4process variables (ORP, pH, DO, LL) of wastewater are the inputs ofPCA model,100batches data under normal operating condition and another100batches datawhich includes some fault are utilized respectively in the building and verifying of PCAmodel. The study results show that when accumulative contribution rate is85%andconfidence coefficientαis95%, the number of principal component is2, the control limits ofT2and SPE statistics are27.3253and1.2421, and the FDM dovetails the physical faultcondition closely. The whole results lay excellent foundation for the following FD modelestablishment of paper mill wastewater treatment system.During the development of FD model for paper mill wastewater treatment system, whichincludes more variables and more volatile data,7process variables (TCODI, SCODI, SSI,pHI, TCODO, SSO) of wastewater are the inputs of PCA model,146batches data undernormal operating condition in2009and another820batches data in2010which includessome fault are utilized respectively in the building and verifying of PCA model. The studyresults show that when accumulative contribution rate is85%and confidence coefficientαis99%, the number of principal component is4, the control limits of T2and SPE statistics are14.0929and2.7526, and the FD model successfully dovetails23of29SBR faults occurredfrom August to December in2010. The accuracy of this model is79.31%.The potential suggestions for further research is brought forward:(1) Denoisingtreatment of raw data could be performed firstly before the building of PCA model, and otherkinds of advanced PCA methods could also be employed to improve the accuracy of FDmodel.(2) If the FD system and the online monitoring system of papermaking wastewatertreatment are connected, then the online FD system of papermaking wastewater treatmentprocess could be achieved.(3) In order to remove the faults rapidly, the classification ofidentified faults should also be studied.