METHOD FOR MEASURING REACTOR BED LEVEL FROM ACTIVE ACOUSTIC MEASUREMENT AND ANALYSIS

BACKGROUND OF THE INVENTION

[0001] The present invention is a method to determine bed levels in a reactor. In particular, the present invention uses a non-intrusive method to monitor the bed level in the reactor in a refinery. More particularly, the method is used to monitor the fluid lever in a delayed coker drum.

[0002] It is important to know the bed level in a process reactor, particularly a delayed coker, in a refinery. Delayed-coking is an important process which is used to convert vacuum resids to lighter distillates and produce coke as byproduct. Delayed-cokers are operated in semi-batch mode, with at least two coke drums. While one is being filled or coked, the other is being de-coked. During the coking cycle, the drum is gradually filled with liquid feed at elevated temperature (typically around 900 degree F) from a bottom nozzle, vapor product leaves the drum at the top, and coke forms inside the drum as the result of a complex solidification process. During the decoking cycle, the drum is first cooled by steam or water. After cooling and draining, a high-pressure rotary water jet is used to cut coke loose, which is removed from the bottom of drum for transportation. Delayed coker are recognized as one of primary capacity bottlenecks at many refineries.

[0003] Refineries that have difficulty in knowing the exact level in the drum run two risks. One is that they overfill the drum. In this case the coke inside the drum starts to stick to the top outlet and to any structure on the top of the drum. When this occurs the drum needs to be shut down and manually cleaned. This leads to significant throughput loss and maintenance cost. The second risk is that in order to prevent the condition described above a large portion of the drum is left unfilled and unused. This unused volume represents a loss of throughput

for the unit. For example, a five foot excess outage in a 100 foot tall drum represents a 5% throughput loss.

[0004] Due to the significant financial impact of this problem a few different methods to predict or measure the level have been developed. The least expensive one is to calculate the level from the feed conditions. This suffers from the difficulty in knowing the final density of the coke and is not reliable. Methods using ranging radar or sonar have to be installed inside the reactor and suffer severe fouling problems. Other methods using radioactive sources have also been developed to measure the bed level. These are more expensive, have radioactive hazards, and generally suffer from reliability issues (probably due to workers avoiding maintenance contact with radioactive devices). Other drawbacks of the radioactive devices are that they can only provide discrete level measurements, instead of continuous measurements provided by the current invention.

[0005] The dynamic response of a vessel filled with a liquid/solid to the excitation of a sound source varies with the level and properties of material in the vessel. These variations in the dynamic response result from the changes in the natural frequencies, mode shapes and attenuations when the fill level varies. In the present invention, these differences are recognized and successfully correlated to different fill levels in a vessel. By measuring the dynamic response of the vessel to a known excitation source, the fill level could be estimated. The main advantage of this approach is that it is non-intrusive, low cost and simple to install and use.

SUMMARY OF THE INVENTION

[0006] The present invention is a non-intrusive method to determine the level of material in a vessel such as a process reactor, or a liquid storage tank. In a preferred embodiment, the vessel is a delayed coker drum in a refinery. The

method uses an acoustic/vibration source to generate vibration or elastics waves traveling along the vessel wall. The characteristics of the wall vibration depend on the fill level of the vessel. One or more acoustic/vibration receivers are strategically placed on the exterior surface of the wall, preferably along the vertical axis of the vessel to measure the vibration response of the wall to the source. The measured vibration signals are analyzed to determine the amount of the change in the characteristics of the wall vibration due to the fill level. A correlation model describing the change in characteristics of the wall vibration is used to estimate the fill level. This process will be repeated to provide continuous or periodic estimate of the fill level.

[0007] For excitation source in ultrasonic frequency range (>20,000 Hz), the generated waves are typically guided waves. For certain modes, the guided waves traveling along the vessel wall leak energy into the material inside the vessel and are attenuated accordingly. The amount of attenuation depends on the fill level. Based on this principle, an ultrasonic transmitter, with a preselected center frequency, is used to generate guided-waves traveling along the wall. One or more ultrasonic receivers are attached to the exterior surface of the wall, preferably along the vertical axis of the vessel, to measure the attenuation. The attenuation of the guided waves is used for estimating the fill level.

BRIEF DESCRIPTION OF THE DRAWINGS

[0008] Figure 1 shows a schematic drawing of the vessel level measurement system based on ultrasonic guided wave approach.

[0009] Figure 2 shows the effect of fluid (water) on wave speeds, from numerical simulations.

[0010] Figure 3 shows the effect of fluid (water) on wave attenuation, from numerical simulations

[0011] Figure 4 shows the fluid level measurement resolution as a function of frequency for different modes, from numerical simulations.

[0012] Figure 5 shows the fluid level measurement resolution versus the attenuation for different transducer sensitivities, from numerical simulations.

[0013] Figure 6 shows a schematic drawing of the experiment setup for Example 2.

[0014] Figure 7 shows a comparison of the experiment results of the fill level measurement and the predictions of the model for Example 2.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0015] Ultrasonic guided wave attenuation can also be used to monitor the the fill level of a vessel. Guided waves, as they are propagating along the vessel wall, will leak energy into the fluid within the reactor. The amount of energy leaked depends on the wave modes, frequency and the fluid properties and level. By controlling the wave mode and frequency for a known fluid, the level of the fluid can be determined through monitoring the attenuation of the guided wave propagation.

[0016] Figure 1 shows a typical level measurement system based on guided wave approach. The system consists of ultrasonic transmitter (1), one or several ultrasonic receivers (2), wave matching wedges (3) for transmitter and receivers, a signal generator and amplifier unit (4), and a signal conditioning and acquisition unit (5), and data analysis and display unit (6). The location of the transmitter is preferably near the top of the vessel above the maximum fill level. The ultrasonic receivers (2) are strategically placed on the vessel wall on the same vertical line with the transmitter that is parallel to the vertical axis of the vessel along which the fluid surface rises. Both transmitter and receivers are selected to withstand high wall temperature (typically 900 degree F for a delayed coker). The wave matching wedge (3) is design to couple the transducers with

wall for efficient generation of certain wave modes, and could also be used to reduce the temperature from the wall to the transducers so that low cost transducers may be used. For multiple receivers, the number of the receivers and space between the adjacent receivers are determined by the requirement of the measurement resolution and range. A signal generator and amplifier unit (4) is used to drive the transmitter for guided wave generation. The signal conditioning and acquisition unit (5) is used to acquire the ultrasonic signals from the receivers, and a data analysis and display unit (6) is used to analyze the signals for estimating the fill level and display/report the fill level to the operation personnel, and preferably computerized. The parameters of the level measurement system are discussed in the next section.

[0017] Typically, a guided wave consists of many modes; each propagates with its own characteristics (speed and attenuation). Since more wave modes will result in more complicated signals, it is advantageous, for the attenuation measurement, to generate the least number of modes. To determine optimal modes and frequencies for the level measurement, a model is developed to simulate the guided wave propagation in the steel wall of a vessel containing a fluid, and the dispersion curves for different modes are computed with numerical simulations. Because the thickness of the wall of a typical reactor vessel is much smaller than other dimensions, the guided wave propagation in the vessel wall can be reasonably approximated by that in an infinite flat plate of the same thickness, subject to fluid loading on one side. For this model, the wall thickness is taken to be 12.7mm (0.5") which is typical for a delayed coker drum. The results with water as fluid medium are shown in Fig. 2. Each line represents the dispersion curve of a particular wave mode. The solid lines are for the part of the wall above the fluid level (Dry), and the dished lines are for the portion immersed in water on one side (1 Side Wet). It is seen that the effect

of fluid loading (water) on wave speeds is extremely small so that it does not change the dispersion characteristics.

[0018] Based on the results shown in Fig. 2, the AO and SO modes are good candidates for the attenuation measurement because: (a) they are in the low frequency range, therefore, can propagate longer distance, which minimizes the number of transducers needed; and (b) they are easier to be generated and detected; (c) at the low frequency range (<150 kHz for 12.7 mm (0.5") thick steel wall), only the AO and SO modes are generated. This makes the signal processing much easier. It should be noted that AO mode is the first asymmetric mode, and SO is the first symmetric mode respectively for guided wave propagation in a plate.

[0019] A guided wave propagates in a dry plate without attenuation except for geometric spreading and wall-material inherent attenuation, both of which are independent of fluid level and generally very small. When one side of the plate is immersed in a fluid, the portion of propagating guided waves leak into the fluid and is attenuated. However, different modes attenuate differently. Shown in Fig. 3 is the attenuation of different modes as a function of the product of frequency and thickness. It is seen that the effect of fluid loading (water) on attenuation is quite significant for the AO mode at low frequencies. For example, at 100 kHz, the AO mode decays -10 times per meter. On the other hand, the SO mode attenuates much less. Therefore; AO mode is the best mode for measuring the fill level with attenuation.

[0020] Another consideration for selecting the AO mode is measurement resolution. Let us assume that measurements are conducted at two fluid levels, h ₁ and h ₂ . It then follows from Eq. IS below that

where B ₁ and B ₂ are the wave amplitudes received by Transducer B when the fluid levels are h ₁ and h ₂ , respectively, and α _f and α are the attenuation coefficients for the vessel wall with and without the fluid respectively . Since the attenuation depends on the wave modes and frequency, the measurement resolution Ah is also a function of the mode and frequency. Shown in Fig. 4 is the resolution normalized by the wall thickness a as a function of frequency for both AO and SO modes. It is seen that at lower frequencies (less than 150 kHz for a = 12.7 mm), AO gives much better resolution.

[0021] Figure 5 shows the fluid level sensing resolution versus the attenuation for different transducer sensitivities. It clearly shows that higher transducer sensitivity and higher attenuation improves the sensing resolution.

[0022] Though the analysis above shows that lower frequencies give more attenuation and better measurement resolution, too much attenuation also reduces the measurement range. In other words, for a given attenuation coefficient and a fixed distance between a pair of transmitter and receiver, there is a maximum liquid level beyond which the wave signal becomes so weak that it is no longer measurable. The selection of the frequency also depends on the efficiency of the wave generation of the transmitter. For practical consideration, the value of the product of the frequency and wall thickness in the range of 1.0 to 1.5 (MHz-mm) may be used to select the frequency of the guided waves for level measurement. For a wall thickness of 12.7mm, an operational frequency around 100 KHz provide adequate measurements.

[0023] The emitted waves from the transmitter could be either continuous with constant amplitude and frequency, or intermittent pulses of given shape and pre-selected central frequency. When pulses are used, the pulse is generated with a pre-determined interval as required for updating the level measurement.

Example 1

[0024] An experimental setup to monitor the fluid level in a vessel is shown in Figure 6. Ultrasonic transducers are placed on the outside wall of the vessel, as illustrated in Fig. 6. During the filling process, Transducer A sends out a continuous wave of constant amplitude with certain frequency. The wave travels down the vessel wall as a guided wave and is received by Transducer B. The guided wave's amplitude decreases due to attenuation. The attenuation is much greater along the portion (d ₂ ) of the wall immersed in the fluid than along the portion (d ₁ ) of the wall above the fluid, because of leaking of energy into the fluid. Since the attenuation of wave amplitude per unit propagation length along the wall with and without fluid is known (either computed from the model or experimentally determined), the amplitude of the guided waves received at Transducer B enables us to estimate the fluid level in the vessel.

[0025] To illustrate the invention, let us assume, as shown in Fig. 6, that the distance between Transducers A and B is H. Let the distance between Transducer B and the top of the fluid be h. The distance between Transducer A and the fluid surface is thus given by H - h . Further, let the attenuation coefficients be a and a _f for the wall without and with fluid, respectively. If

Transducer A sends out a wave with amplitude A, then by the time the wave reaches the fluid surface, its amplitude becomes A As the wave further propagates to Transducer B, its amplitude further reduces to . Let B be the wave amplitude received by Transducer B. Then, the value of amplitude B is given by

[0026] In the above equation,, all parameters (A, α and a _f )are known or can be measured prior to filling the vessel. During the filling process, once the

wave amplitude B is measured by Transducer B, the fluid level h can be solved from Eq. (3S) _,

[0027] In the experiment, a matching wedge is be used for Transducer A to generate the AO mode below 150 kHz frequency range. A similar matching wedge is used for Transducer B to receive the AO mode. Based on the wave amplitude received by Transducer B, the actual fluid level can be estimated using the method discussed above.

[0028] The fluid level in the vessel was measured using the methodology described above. Attenuation of the guided waves was measured at various fluid levels. The measured attenuation is then compared with the predictions based on the model described above. The comparison is shown Fig. 7, where the open circles represent the ultrasonically measured data and the solid line represents model predictions for each given fluid level. It is seen that the model prediction agrees very well with the experimental data.

APPENDIX

Guided Waves in a Free-Standing Plate

[0029] Consider a plate of thickness 2h placed in vacuum where the x- axis is along the plate and the z axis is perpendicular to the plate . Time- harmonic wave motion in the plate is governed by the reduced wave equation,

where u is the displacement vector, ω is the angular frequency, and c _L and c _τ are the longitudinal and shear velocities in the solid, respectively.

[0030] By a direct substitution, one can show that the general solution to representing a time-harmonic wave propagating in the jc-direction with wavenumber k can be written as

Where

represent the symmetric (with respect to the plane z = 0) part of the wave motion, while

represent the anti-symmetric part of the wave motion. In these equations,

and the constants A ₁ , A ₂ , B ₁ , and B ₂ are to be determined by the traction-free boundary conditions on both sides of the plate.

[0031] To this end, we use the Hooke's law to calculate the stresses,

Where

correspond to the symmetric displacement field, and

correspond to the anti-symmetric displacement field. In the above, μ is the shear modulus of the plate.

[0032] Substituting these into the traction-free boundary conditions,

yields four homogeneous linear equations for the four constants A ₁ , A ₂ , B ₁ , and

where

[0033] For non-trivial solutions, we must have either

or

[0034] These two equations are called the Rayleigh-Lamb dispersion equations. Solutions to these equations provide the relationships between the frequency ω and the wavenumber k. Such ω - k relationship is called the dispersion relationship. Making use of the relationship ω = kc , where c is the phase velocity of the propagating guided wave, the dispersion relationship can also be expressed as velocity versus frequency curves as shown in Fig. 6. Since there are multiple solutions to (20) -(21), many dispersion curves exist for a given plate. Each curve is called a mode. The curves corresponding to (20) are called symmetric modes, because the displacement vectors (eigenvectors of(15) correspond to u _x ^s and wf which are symmetric with respect to the z = 0 plane. Analogously, the curves corresponding to (21 ) are called anti-symmetric modes, because the displacement vectors correspond to u _x ^A and u _z ^λ which are antisymmetric to the z = 0 plane. The first (lowest in frequency) anti-symmetric mode is called the AO mode, while the first symmetric mode is called the SO mode.

[0035] We note that (20) - (21) have only real solutions for real frequencies. Therefore, the guided waves propagate along the plate without attenuation.

Guided Waves in a Fluid-Loaded Plate

[0036] Consider the same free-standing plate of thickness 2h placed on a fluid on one side. The guided wave propagating in the plate can still be represented by the general expression(2). Assume that the fluid has mass density p _f and phase velocity c _f . Then, the wave field in the fluid can be written as

where C is a constant to be determined from the interfacial condition between the fluid and the plate, and

[0037] The corresponding stresses in the fluid are

[0038] On the solid-fluid interface, we assume that the displacement component in the z-direction is continuous, i.e.,

[0039] Further, the traction must be continuous across the solid-fluid interface, i.e.,

[0040] On the top surface of the plate, it is still traction-free, i.e.,

[0041] Substitute (2) and (23) into(26), (8) and (25) into (27) and (25) into (28) yields

[0042] For non-trivial solution, the determinate must vanish, which leads to the dispersion equation where and are defined by (20) and (21), respectively, and

[0043] In general, for a given real frequency ω , Eq. (30) has complex roots for k. These complex roots represent wave propagation with attenuation. The real part of k: represents the propagating part (wavenumber) whereas the imaginary part of k represents the attenuation. The attenuation is due to the leaking of energy from the plate into the fluid. The roots of the dispersion equation are presented by two sets of dispersion curves — the propagation dispersion curves featuring frequency versus the real part of k, such as those shown in Fig.8, and the attenuation dispersion curves featuring frequency versus the imaginary part of k, such as those shown in Fig.9. Note that when p _f = 0 , the dispersion equation (30) reduces to(20).