Fault Detection Technology In Liquid-hydrogen Production Process Based On PCA

From the aerospace technology developing process in America, Europe, Russia andJapan, liquid hydrogen, being the high efficient and pollution free propellant, is the idealestfuel for improving the launch vehicle performance. When launch vehicle technology reaches acertain level, liquid hydrogen and liquid oxygen must be selected for rocket propellent. In thesame time, the developing of aerospace technology will drive the liquid hydrogen applicationpromotion. Liquid hydrogen, being the ideal energy, its use range will be expanded to, forexample, aviation,`automotive field, and will`play a growing role.The liquid hydrogen production process is very complicated. If there is some fault, it willnot only affect the development cycle and launch task of aerospace model in our country, butalso it will create disaster. However, liquid hydrogen production in our country started lateand we have little experience. Especially the fault detection about liquid hydrogen productionreport have not been seen. So it is very meaningful to carry out the study of liquid hydrogenproduction process fault detection.This paper first introduces the development and current situation of fault detection andfault diagnosis technology. And for the liquid hydrogen production characteristics, a studysuitable for using PCA analysis method to carry out fault detection is put forward. Next, theprocess of a real liquid hydrogen production line is described in detail, and then a physicalmodel is made. At last, according to the physical model and our experience, the fault model ofthe liquid hydrogen production line is studied, measured variables are determined. It lays thephysical foundation for the follow up fault detection using math method.On the basis of above mentioned, PCA method is introduced in detail, principle modelapplication in the liquid hydrogen production process is analyzed and verified deeply. Theresult shows that using SPE(square prediction error) statistics can detect effectively the mainfault in liquid hydrogen production, but the fault which affect the process slightly can not bedetected. A optimized PCA fault detection method based on D-detection index is put forward,and its application to liquid hydrogen production fault detection is studied. Simulationverification results show that fault detection technology based on optimized PCA not only candecrease the error prediction phenomenon, but also can improve the detection sensitivity. Forsome faults which affect the process slightly, traditional PCA can not detect them, but theoptimized PCA can detect them effectively.The study, analysis and verification show that the principle component analysis methodin this paper can be used comprehensively to detect the fault of liquid hydrogen production process, and its algorithm can be used to establish application software and develop the realtime fault detection system of liquid hydrogen production process.