|학술대회||2017년 봄 (04/26 ~ 04/28, ICC 제주)|
|권호||23권 1호, p.241|
|제목||Fault detection and diagnosis of nonlinear process using various types of principal component analysis|
|초록||For process faults in chemical plant bring about tremendous economic loss, it is essential to detect faults early and take action immediately. Most chemical process, however, have highly nonlinear characteristics and there are hundreds of thousands of sensors in real plant which send out signal simultaneously. These adverse situations make operators difficult to monitor process state. To overcome these limitations, various types of monitoring techniques such as PCA (Principal Component Analysis) and kernel PCA has been developed since a few decades ago.
In this study, a few fault cases of ‘Tennessee Eastman(TE) process’ which is usually used as benchmark case and ‘HDA process’ which aim to produce benzene from toluene feed were used to compare the performance of conventional PCA algorithm with kernel PCA that was developed taking into account the nonlinearity of process. In addition, for quick troubleshooting simple diagnostics have been performed and checked its accuracy. TE process is simulated using Matlab Simulink while HDA process using Aspen Hysys v8.4.
|저자||이호동, 하대근, 김창수, 한종훈|