Fault Detection And Diagnosis For Crude Oil Desalting Process Based On Near-Infrared Spectroscopy

Crude oil desalting process is a crucial part in petroleum refining process.Due to the high integration of the petroleum refining process,the failure of the desalting process may result in overall plant shut-down and potentially major accidents.Thus,it is significant to detect and troubleshoot timely.To address the above problems,near-infrared(NIR)fault detection and diagnosis technology based on molecular spectroscopy is proposed in this paper.Firstly,the on-line spectral data of crude oil desalting process were obtained by using the NIR spectrometer which can obtain the process information from the molecular vibration level.Then,the crude oil desalting process was monitored from the perspective of microscopic molecular spectroscopy by adopting data-driven approach.Different from the traditional method based on macroscopic process variables,the NIR fault detection and diagnosis technology based on molecular spectrum can detect faults more quickly.The main research contents and contributions of this paper are as follows:1 Fault detection for crude oil desalting process based on NIR spectrum.The NIR spectrum analyzer was installed in the desalting crude oil output pipeline to obtain the information of desalting process.Based on the NIR spectral data,the principle component analysis(PCA)algorithm was used to monitor the desalting process.The Hotelling T~2 and squared prediction error(SPE)statistics were selected as the judgment indicators.The results show that the NIR-based fault detection method is about 140 minutes faster than the macroscopic process variables based method which fully reflects the advantage of NIR spectroscopy in fault detection.2 NIR-based fault detection for crude oil desalting process based on SPCA.Due to the high dimension of the spectra data and the extremely sensitive of molecular spectral data to the environment,NIR technology has a sensitive problem when combined with PCA algorithm to achieve fault detection for crude oil desalting process.To solve the above problem and improve fault detection accuracy,sparse principal component(SPCA)algorithm which can reduce the impact of very small changes at the molecular level on PC by compressing some load of principle component(PC)to zero was used.3 NIR-based fault diagnosis for crude oil desalting process based on LASSO.After the fault is detected,a fault diagnosis method based on NIR spectrum is proposed in this paper to further determine the cause of the fault.By adopting fault reconstruction and the least absolute shrinkage and selection operator(LASSO)algorithm,the NIR spectral regions affected by the fault from most to least were located.By analyzing the located NIR spectral and combining the physical and chemical properties of the crude oil desalting process,the cause of the fault is speculated which provides a reference for fault diagnosis.