When speed matters - Hydrocarbon Engineering

When speed matters

Loek van Eijck, Yokogawa, The Netherlands, questions whether rapid analysis of gases and liquids can be better achieved through use of a gas chromatograph or near infrared analyser.

Conventionally, the liquid and gas components such as those broken down by naphtha crackers have been measured by a process gas chromatograph (PGC), with the subsequent measurement values then being used for control purposes. However, using a PGC can take 5–10 mins for a complete analysis and many industries require a much faster cycle time.

In contrast, a near infrared (NIR) spectrometer can measure the components of both liquids and gas at high speed. The technology can also be applied to petrochemical processes, such as naphtha cracking furnaces in ethylene plants and other chemical processes. In these cases, high speed measurement and control have been achieved with a cycle time of approximately one minute.

Measuring in the NIR spectrum

The measurement of species using infrared absorption in either the mid infrared (MIR) or NIR regions is not a new concept. Yet only within the last three decades has the use of the highly stable Fourier transform infrared spectrometer been possible. This is because of the need to perform intensive Fourier transform calculations at speed: until the 1980s, computers simply were not fast or small enough to achieve this.

The MIR range provides the best spectral data and information, making it widely used in laboratory environments. Infrared (the region next to visible light in the range of 900–2500 nm) is a popular choice for process applications due to its practicality.

Although NIR light has lower absorbance than MIR, the path lengths employed are typically between 1–20 mm,whereas MIR employs path lengths in the sub millimetre region.

Having two optical windows so close together means that the windows can be 'filmed up' much quicker, or even blocked if care is not taken. Furthermore, the type of optical materials required for MIR are expensive and tend to be hydroscopic (water absorbing), meaning that they can be destroyed by water.

As C-H, N-H and O-H bonds show strong absorption bands in the NIR range, almost all organic substances can be measured or characterised using NIR absorption spectra. Even ions in aqueous solutions with no absorption band directly in the NIR region can be indirectly quantified. This is because they influence the spectrum of the water by having an elevating or declining and transformational shift effect, which is subtly different for each ionic species.

Quantitative analysis

(a) Ethylene process hot section (cracking furnace + waste heat recovery + gas refining system)

(b) Ethylene process separation section (front-end demethanizer process)

(c) Ethylene process separation section (front-end depropanizer process)

Figure 1. Block diagram of ethylene production process.

The Beer-Lambert equation governs the principle behind quantifying species using absorption spectra. Instruments range from simple (non-dispersive infrared) devices that employ discrete wavelengths for the identification of a single component (e.g. the measurement of CO₂ in flue gas) to full range scanning and Fourier transform devices that deliver a full information rich spectrum.

The general principle of the Beer-Lambert equation is as follows:

A = -log (I_out/I_in) (1)

A = α C • L (2)

Where 'A' is absorbance, 'I_in' is intensity of irradiating light, 'I_out' is intensity of penetrating light, 'C' is concentration and 'L' is path length.

When the path length is constant, the following simple relation is obtained:

C = B • A (3)

Where 'B' is a constant.

However, the concentration quantified in this expression is limited to only one solute for the solution, a relatively simple case. For multi component measurement, the absorption peaks of components interfere with each other, meaning the concentration cannot be calculated from a single absorbance. To measure multiple components, it is necessary to use an extended calibration model equation using multiple absorbances, as shown by the equation:

C = Σⁿ_i=1b_i * A_i (4)

To create this extended calibration model equation, NIR spectra of multiple samples whose concentrations (C) are known are measured, and the constants (b_i) are derived by solving Equation 4 using multiple combinations of the concentration and spectrum (A_i). Once the constant (b_i) is determined, the concentration (C) of an unknown sample can be calculated using its spectrum.

The calibration model can be created for quantities or qualities other than the concentration, as long as the quantities or qualities to be measured (such as density, viscosity, octane number of gasoline, acid number of ester reaction, etc.) are correlated with changes in the molecular structure or amount of ingredients. This means that the NIR analyser is capable of various measurements by using different calibration models.

Figure 2. NIR spectra of naphtha.

Ethylene plant application

Figure 1 shows a block diagram of the ethylene production process using naphtha (the final product of the petroleum refinery process) as the raw material.

The crude oil is first separated into several kinds of intermediate materials via the atmospheric distillation unit (distillation tower), depending on boiling temperature. Upon distillation, final products of the process include LPG, asphalt and everything in between. To satisfy high demand for gasoline, the heavy components of oil are reformed and resolved to increase the light intermediate materials, thus increasing the gasoline fraction. Naphtha, like gasoline, is produced from relatively light components, and is supplied to ethylene plants as a raw material.

In ethylene plants, the carbon bonds of the naphtha are broken (cracked) in the cracking furnace (cracker) and the gas produced by the cracker is distilled, reformed, separated and refined into single components. The most important components are ethylene and propylene, which are used as raw materials for many industrial products. As indicated by the red circle in the figure, a NIR analyser can quickly measure concentrations at any point.

NIR analysers are often installed in the analyser house, in the vicinity of the cracker. The analyser is fed with samples from the inlet and outlet of the cracker. Preprocessing operations such as filtering, defoaming and temperature adjustment are performed before they arrive at the flow cell for measurement, thus minimising disturbance of the measured value. The NIR analyser is operated over a measurement cycle of approximately one minute, which, as indicated earlier, is far shorter than the analysis cycle of a PGC (5–10 mins). In addition, the NIR analyser does not need evaporated samples, which offers maintenance benefits when the instrument is used at the inlet of the cracker where many components with high boiling point exist.

Figure 3. Result of benzene calibration model at inlet of naphtha cracker.

A flow cell is used for measuring samples at the inlet of the naphtha cracker. The flow cell has a path length of 10 mm and is equipped with a constant temperature water pipe, which can maintain the liquid temperature of the sample constant and is used together with an insulation jacket.

Spectrum measurement of a gas requires a cell with a longer path length (approximately 100–500 mm) than those for a liquid.

Liquid analysis

The naphtha supplied to the naphtha cracker consists of carbon hydride (carbon number 1–8) combined with components such as methane, which are gas phase at normal temperatures and pressures, dissolved in it.

Figure 2 shows the NIR absorption spectrum of the naphtha. The spectrum was measured by preparing 175 samples, including those collected from the naphtha cracker during operation and those produced by adding some reagents. The peak corresponding to the molecular structure of the constituent component can then be found, showing that the absorbances of the peaks differ according to the sample.

Figure 3 shows the benzene calibration model created by using the 175 samples. The calibration model with a good correlation was obtained where the correlation coefficient was 0.999 and the prediction error was 0.025 wt%. The prediction error is the standard deviation of the difference between the NIR analyser measurement result and the laboratory analysis result.

In addition to benzene, calibration models have been created for the following 24 items with reasonable results:

• Density.
• Distillation point (initial, 50%, 70%, end).
• Paraffin (total, C3 and C4, C5, C6, C7, C8).
• Isoparaffin (total, C3 and C4, C5, C6, C7, C8).
• Naphthene (total, C5, C6).
• Aromatics (total, C6, C7, C8).

Figure 4. NIR absorption spectra of methane and ethane.

Gas analysis

Measurement at the outlet of the naphtha cracker is important in order to obtain the component ratio and the resolved residue of heavy components. Furthermore, this allows for optimum control of the cracking furnace operation. The combination of this measurement and the analysis result at the inlet side realises further optimum control of the operation.

The spectrum of a gas is easily affected by temperature and pressure, making it crucial to accurately control these parameters. However, the procedure for creating calibration models using the spectrum of the sample and the reference values (laboratory analysis values) is identical to that for a liquid.

Figure 4 shows spectra of methane and ethane as examples of NIR absorption spectra of gas. It shows sharper peaks than the spectra in Figure 2, and is made up of many small peaks due to the molecular spin of methane. The NIR analyser can observe these peaks without being influenced by noise because of its low noise characteristic. In addition, the calibration models can be created using the same procedure as for a liquid.

(a) Calibration model

(b) Time-series trend comparison

Figure 5. Calibration model at outlet of naphtha cracker and time series trend.

To create the calibration model at the outlet of the naphtha cracker, 33 samples were prepared by the same procedure as at the inlet. Though a small amount of acetylene, butane, benzene (or similar) could be measured, the verification was conducted by focusing on the most important seven components: methane, ethane, ethylene, propane, propylene, i butane and n butane.

Figure 5a shows the created calibration model of propylene. With a correlation coefficient of 0.996 and a prediction error of 0.22 wt%, it was possible to obtain good results. Figure 5b shows the measurement results of the NIR analyser and the PGC at the outlet. During the measurement, the values of the two instruments are close enough in time zones from 0:00 – 10:00 and 23:00 – 24:00. Though the value for the NIR analyser is slightly higher than that of the PGC from 10:00 – 23:00, both trends more or less match, causing no practical problem. Although the verification period is short and the examination is insufficient, this result suggests that high speed control of the cracker is feasible in future.

Chemical process application

The respective components distilled and refined in an ethylene plant are used for manufacturing chemical products in various fields. An NIR analyser was applied to a butyl rubber production process using isobutylene, isoprene and methylene chloride. The analyser was installed in the process to continuously collect the spectra of the sample gas, employing a cell with a path length of 500 mm. The data from a PGC at the same location was used as reference, with more than 1000 samples collected over several weeks. Figure 6a shows an isobutylene calibration model with a correlation coefficient of 0.991 and a prediction error of 0.19 wt%, while Figure 6b compares the time trends of the NIR analyser and chromatograph indications. As the data update cycle of the NIR is shorter than that of the PGC, it can detect a concentration change earlier. In the future, it will be possible to control the isobutylene concentration in a more stable manner by using the NIR analyser.

Hybrid dual analyser concept

NIR analysers have long been used for liquid analysis in various industrial processes, but in recent years they have begun to be applied to gas analysis. They are also attracting attention as a means of realising rapid and simultaneous measurement of multiple components in both liquid and gas samples. In the field of petrochemical processes, including ethylene plants and related chemical processes, various efforts have been made to improve stability of control by utilising the features of the NIR analyser. Further product development will contribute to enhancing the efficiency of customers' plants.

(a) Calibration model

(b) Time-series trend comparison

Figure 6. Isobutylene calibration model and time series trend.

Naphtha steam crackers are at the heart of the single most important process in the petrochemical industry. They are the starting point for all chemicals, and as a result they process very raw hydrocarbon material, typically spot market purchases or left over naphtha from a local refinery. In both cases, from a chemical composition and property perspective, the feed can be highly variable.

Fourier transform NIR technology has many strengths, including very high speed, repeatable and multiple stream measurements, which make it ideal for process control. However, the measuring method is indirect or inferred. A complex chemometric calibration model is required to predict properties from absorption spectra. Highly changing conditions and feed mean that model maintenance can be quite intensive, and can rely heavily on laboratory analysis of collected samples. This inherently raises the possibility of introducing errors into an already complex process, indicating the need for a high degree of owner responsibility.

A possible solution to this challenge is a hybrid concept, which involves installing a Fourier transform NIR analyser together with a PGC, with both measuring the same components. This offers two potential advantages: hardware redundancy, and an automatic method of performing model updates by having a closed loop for updates and avoiding errors.

Analysis of the payback for initial investment in an additional analyser can be seen in the hundreds of man hours saved for laboratory and process personal in the manual development of these calibration models. The hybrid concept is revolutionary in the sense that the weak points of both the PGC and the Fourier transform NIR are fully compensated, providing potential customers with a reliable, fast, stable and flexible solution with minimum cost of ownership.