Jul-2017

Online chlorine analyser for crude oil and petroleum process streams

Residual chlorine content in desalted crude oil streams can be a serious issue due to its highly corrosive nature. Online monitoring of chlorine has an important economic aspect for the oil industry.

Z W Chen and Stuart Shakshober
XOS

Article Summary

An online total chlorine analyser based on Monochromatic Wavelength Dispersive X-ray Fluorescence (MWDXRF) has been developed for this purpose. This online chlorine analyser is comprised of a low power X-ray tube, an X-ray optical chamber, and a dynamic window module sample flow cell. An intense monochromatic beam coupled with a highly efficient doubly curved collection optic provides superior detection capability. A detection limit of 0.18 ppm and a quantification limit of 0.6 ppm has been achieved. Data collected from multiple systems have shown great linear response, robustness and precision. Crude oil and desalted crude oil streams were measured for extensive periods of time to demonstrate the stability and precision achieved.
 
Introduction
Chlorides are often present in crude oils and their concentration can vary greatly, depending on the origin of the crude. The chloride concentration can also spike from contamination. These chlorides, if not removed, can react to form highly corrosive hydrochloric acid in areas such as the crude tower overheads, leading to serious operational issues. Other elements, like sodium, can also poison catalysts further downstream. Most of the chlorides in crude oil are inorganic and a high percentage can be removed by a front-end desalting process. On the other hand, instances of high level organic chlorides in crude oil are rare but can be more problematic if present. The normal desalting process is not effective for removal of organic chlorides and may not remove all inorganic or non-extractable chlorides, allowing excess chlorides to pass downstream, undetected. There are a few recent incidents where unexpected high levels of chlorine in a downstream process resulted in significant economic loss and safety concerns. The need for a reliable online chlorine analyser for refining process control is more important than ever for oil refining under the circumstances faced today.

There is no effective online method for measuring Cl in crude oil in a process stream. MWDXRF X-ray analysis was developed in the past decade and has become a powerful online process tool for total sulphur analysis in product streams1,2. However, chlorine online analysis in crude oil presents more challenges due to low level quantification requirements and sampling complications. A MWDXRF online chlorine analyser has been developed and launched to meet these needs.

Instrumentation
Doubly curved crystals (DCC) provide the basis for MWDXRF analysis and the basic geometry of a focusing DCC was demonstrated by Chen and Wittry3.4. The schematic diagram of a total chlorine analyser is shown in Figure 1. The analyser consists of an analyser engine, a dynamic window module (DWM) flow cell, a single channel analyser, a PLC, power supply and display. The engine consists of an air-cooled low power X-ray tube “T”, a point-focusing DCC optic “E”, a second collection DCC C and a proportional counter “P”. An evacuated chamber is designed to contain the optical paths. A diaphragm pump is used to maintain a 5 torr vacuum level inside the chamber. The X-ray tube is Cr target source. The excitation DCC reflects only Cr Kα line and forms an intense monochromatic beam in a small spot on a sample stream “S” outside the chamber. The excitation beam excites Cl atoms in the sample and Cl Kα fluorescent X-rays are collected by the second DCC. The collection DCC has a very narrow energy bandwidth of 10eV, which eliminates interferences from S, Ar and other elements. The sample stream flows through the DWM which has a polymer type window allowing the X-ray beam to pass in and out. The polymer window is rolled on a spool “W” so that the window can be dynamically refreshed in a specified period. The DWM can also be heated to an elevated temperature to improve flow of heavy crude oil or feed stocks. The DWM design is the key to eliminate long term background drift of the system.

Figure 2 shows a schematic flow diagram of a MWDXRF online chlorine analyser. In “Measurement” mode, a sample stream such as crude oil passes to the DWM through valve 7 and continuously flows to a sample return point. When it is time to advance the window film to refresh the window, the sample stream is blocked and a pressurized source of clean instrument air or N2 is used to purge the DWM cell and advance the film. The measurement will resume when the film advance is complete, taking approximately 90 seconds. In “Calibration” mode, a standard sample with known concentration will enter the DWM cell through valve 8. The standard sample is first filled in a small vessel which is pressurized by the instrument air. Dedicated vessels for each valve of calibration standard can be switched with quick connect fittings to obtain a linear calibration curve.

Experiments
Multiple online chlorine analysers were used in data collection. The analyser engine in each system was assembled on X-ray optical benches. An X-ray optical table with automated, multi-axis stages provided the means for critical alignment of an X-ray source, excitation DCC, collection DCC and detector. The analyser engine was evacuated down to a level around 2-5 torr by a two-stage diaphragm pump. The X-ray source in each engine was operated at 50 kV and 1.0-1.2 mA depending on the optical efficiency of the engine.

 A sample circulation system was set up by using a fuel pump for generating continuous flow for testing. The sample can be circulated through the DWM cell continuously for multiple days. The window film advancement time interval was set to four hours. Signal integration time is set to be 300s.

All analysers were calibrated using xylene-based chlorine standards, obtained from VHG Labs. The concentrations of the standards were 0, 10, 25, and 500 ppm. Each calibration standard was filled in a separate 300 ml vessel and introduced to the DWM cell. During the calibration measurement, the standard sample was static and pressurized. Raw crude oil and desalted crude oil samples were obtained from different refinery facilities. These samples were set up in the circulation system for all the data collected. Chlorine X-ray photons were continuously detected by the proportional counter and the counts were integrated over a 300s period. The counts were converted to concentration by weight based on the calibration curve. The concentration values are displaced every 300s by the PLC with no averaging of data.

A lab benchtop MWDXRF Accu-flow Cl analyser was used to cross-check one of the raw crude samples. Due to the non-homogenous nature of crude oil, quantifying Cl content is not practical with a static sample cup. A syringe coupled with a disposable sample cup was incorporated in this Accu-flow Cl benchtop analyser. In this design, a motor is used to drive the syringe to create a flow stream through the sample cup throughout the measurement, over a span of three minutes. 

Results
Calibration and linearity
For low-level Cl concentration in a matrix, the relationship between the counting rate and concentration is linear. The linear calibration curves of three different online systems are shown in Figure 3 as a net counting rate vs Cl concentration in xylene. The net counting rates were obtained by subtraction of the background rates. The background rate of each system was obtained by measuring a blank sample for 300s. The background rate of each system is indicated on the graph. A linear calibration curve through the origin was generated by fitting the calibration data points. The fitted slope, or sensitivity of the system, and the linear correlation coefficient are shown in the graph. Although Cl content in petroleum streams is typically less than 100 ppm, the 500 ppm calibration point is beneficial to reduce the counting statistics error, which improves the reproducibility of the calibration. The linear correlation coefficient is very close to unity as shown in the graph, demonstrating excellent linearity of all systems. The variation of slope of the calibration curve is due to the variation of analysis engine sensitivity.