CC BY
44
УДК 681.7
EugenAnurjew1, Juergen J. Brandner1, EdgarHansjosten1, UlrichSchygulla1, ThomasStief, KlausSchubert1
1Forschungszentrum Karlsruhe, Institute for Micro Process Engineering (IMVT), Hermann-von-Helmholtz-Platz 1, D-76344 Eggenstein-Leopoldshafen, Germany
2 DECHEMA e.V., Karl-Winnacker-Institut, Theodor-Heuss-Allee 25, D-60486 Frankfurt am Main, Germany
DEVELOPMENT OF A SENSOR FOR GAS FLOW RESIDENCE TIME DISTRIBUTION MEASUREMENT THROUGH MICROSTRUCTURE DEVICES
Abstract
In the recent years a variety of microstructure devices has been developed for numerous tasks in process engineering. Micro reactors, micro mixers, micro heat exchangers and others are more or less common now in science and are gaining more and more importance in special parts of the industry. These components provide specific advantages for heat transfer and mass transfer, mainly due to the increase of the surface-to-volume ratio. The residence time reached with microstructure devices is normally in the range of milliseconds to seconds, depending on the mass flow applied and the design of the microstructures, respectively. For a complete characterization of the devices the residence time distribution is mandatory. Moreover, to allow a more precise pre-calculation, modelling or simulation of the thermal behaviour and the mass flow distribution it is necessary to know the residence time distribution as good as possible. Nevertheless, it is not easy to obtain sufficient residence time distribution information for gas flows through microstructures, which is an unsatisfactory situation.
Therefore, a sensor system was developed and tested. The design of the sensor system as well as the experimental setup and the measurement method will be described briefly in this publication. Results of first measurements will be shown. These experiments have been performed with microstructure devices made by micro-stereo lithography. A relatively large dispersion within the microstructures was found, which could be a result of a flow maldistribution through the microstructure channels.
Keywords: Microstructure device, micro machining, residence time, sensor, gas flow
Summary
New opportunities for process routes in chemical and pharmaceutical industry have been opened by the ongoing miniaturization. Microstructure devices and their great enhancement in heat and mass transfer allow a control of process parameters which was never before reached. On the other hand, the thermal and fluidic behaviour of those tiny devices has to be very well known, which leads to some basic device parameters to be well defined.
One of these parameters is the residence time distribution, especially for gas flows. Conventional measurement methods may be applied to a certain point only, since the residence time in microstructure devices is in the order of seconds or below. Thus, small integrated sensors seems to be necessary, providing measurement delay times in the range of milliseconds.
A new residence time distribution sensor based on the measurement of thermal conductivity changes was developed and tested with a special microstructure device. The device and the sensor were manufactured from a polymer by micro stereo lithography. The measurement was taken with a simple Wheatstone arrangement to monitor the change of electrical resistance of a Pt-Ir-wire placed on the sensor frame. With this system, it was possible for the first time to measure the residence time distribution of a microstructure reactor device online. A Bodenstein number was calculated from the results, showing that the dispersion of the system is high, and therefore the residence time distribution is rather broad. A high mixing and large dispersion is to see. This could be due to some flow maldistribution which leads to higher effective dispersion.
Introduction
Microstructure devices and technology has gained interest in many technological areas within the last 20 years. This is due to the specific properties of microstructure devices. Miniaturization leads to a reduction of the volume by a power of three, while the surface area is reduced by a power of two only. Thus, the surface-to-volume ratio is increased tremendously by reducing the size of devices, leading to
9 -5
values of 30.000 m /m and more, as it is reported by Brandner et al. (2006). Aside of this, the small microstructure dimensions in the range of some ten to some hundred micrometers are in the same range than the diffusion length of gases under common process parameters. This means, diffusion, mixing and, therefore, chemical reaction takes place in extremely short times. On the other hand, due to the very short distances and the high surface-to-volume ratio, heat generated by reaction processes or needed to run those can be transferred in a very easy and efficient way. Micro heat exchangers, micro mixers and micro reactors well described by different authors in literature are good examples for this behaviour. Details can be found by, e.g., Hessel et al. (2005) or Kockmann et al (2006).
To run chemical or pharmaceutical processes in microstructure devices it is necessary to control the process parameters as good as possible. As mentioned before, heat transfer and diffusion respectively mixing is not a problem that much. But due to the parallelization of microstructures into multi-channel devices, it is not quite clear whether an equal flow distribution of the fluid, especially gases, is reached or not. To clear this, it is of advantage to measure the residence time distribution (RTD) of a microstructure device. Microstructure systems are assumed to have a well defined narrow residence time distribution as a result of the defined and small dimensions of the channels. Thus the progress of the individual steps in complex reactions with consecutive and parallel steps could possibly be efficiently controlled and the conversion and selectivity could be improved.
In addition, knowledge of the RTD is necessary for the simulation of the reactor behaviour by means of dispersion, cell or other substitute models. In these cases the fluid dynamics of the reactor is integrated into the model by model parameters (effective dispersion coefficient, cell number) derived from the RTD, as it is described by Baerns et al (1999). Knowledge of the RTD is therefore important for the mathematical modelling and for performance studies of the reactor.
However, to date no sensor system has been available to measure the RTD of the catalytically active parts of microstructured reactors for gas flow with sufficient accuracy. Such a sensor system has to fulfil two main requirements: firstly, the sensor system must be small enough to be integrated into the reactor system and secondly, it must be fast enough (time constants in the region of milliseconds).
The integration is necessary because the RTD effects of the microstructured system are small compared with the RTD effects of capillaries to external sensors. In the case of conducts to external sensors it is not possible to measure the RTD effects of the reactor sufficiently precisely (see Wolf, Richter).
The mean residence time of molecules in the microstructured reactor is in the range of seconds. The sensor has to be at least one order of magnitude faster, thus the time constant of the sensors has to be in the range of milliseconds.
Principle of RTD measurement
For the experimental determination of the RTD, a detectable property of the gas flow into the reactor is changed corresponding to a defined function of time; ideally this would be directly at the reactor entrance. The function of time of this property is detected at the outlet of the reactor. This function is the basis for the determination of the RTD or the residence time sum function: the fluid elements entering the reactor are marked and their fate in the fluid system is monitored (see Baerns et al (1999)).
Electrical conductivity, light absorption and radioactivity are generally used as detectable properties in liquid-phase systems but are not usable in gas-phase systems (except radioactivity). For gas-phase systems, preferably the thermal conductivity is used.
The fluid elements that enter the flow system are frequently marked by an impulse or a step function. As response to an ideal impulse the RTD can be measured at the reactor outlet; as a response to an ideal step function the residence time sum function can be measured at the outlet. However, in real experiments it has to be noted that real test functions can only have approximate step or impulse function shapes. In addition RTD effects already occur between the place where the functions are generated and the entrance to the fluid system. In this case the course of the real test function at the reactor entrance must be measured and taken into account when determining the RTD of the fluid system.
Nevertheless, a sensor measuring the change in thermal conductivity hast to be thermally inert as well as electrically isolated. This is due to the fact that the sensor used is nothing else but a thermal conductivity detector (TCD). In general, this is a combination of electrically heated wires placed into the entrance and the outlet of the fluidic system. Thus, those wires must not come into contact with metal microstructure devices to avoid an electrical breakdown. Devices made of insulating material like ceramic or polymers are needed.
Manufacturing of Microstructure Devices
Most multi-channel microstructure devices are made of metals like stainless steel, hastelloy or other alloys. We will not go into details about manufacturing processes, those can be found in, e.g., Kockmann et al (2006).
For a first microstructure to be characterized for the RTD, it is, aside of the needs for electrical insulation described before, advantageous to use a manufacturing method which allows fast changes of the shape and the design of the microstructure device as well as of the sensor system. Thus, a rapid prototyping manufacturing method was applied to generate not only the device but also the sensor system. Here, we use the process of micro stereo lithography in polymers.
Micro stereo lithography is an additive process to generate three dimensional polymer microstructures. In figure 1 (left), the process is schematically shown.
To generate a polymer microstructure device, a three dimensional CAD model is used. The data are transferred to a computer controlling the position of a laser focus spot right on or above a model building platform. The model is then built within a liquid monomer substrate on the building platform which can be moved vertically, following the lines of the three dimensional CAD model. The platform is lowered for the distance of a layer thickness. Liquid monomer is flooding the building platform, and the focused laser exposure for the next layer starts. With the exposure, the plastic is polymerized, and the next layer is generated. Details can be found in Kockmann et
al (2006), Gebhardt (1996) or Ikuta et al (1994). With the stereo lithography process, a wall thickness of about 100 ^m is feasible. Smaller designs are possible, the question here is the applicability of those to (industrial) processes. In figure 1 (right), a CAD model of a crossflow microstructure heat exchanger and the device realized with stereo lithography is shown.
Using this technology, manufacturing many different types of micro structure devices is possible. The devices can be used either as simple models to gain information about the geometrical and technological data of certain design rules, or to run in specific applications like heat transfer or mixing. In figure 2, a countercurrent flow microstructure heat exchanger made by micro stereo lithography is shown. Figure 3 shows an example for an electrically heated device used as demonstrator for a colour-change reaction driven by temperature.
RTD measurement of gas flow through microstructured systems
Sensor
As mentioned before, the requirements of the sensor system can be fulfilled with a TCD. A TCD is in principle an electrically heated wire. This wire is located directly into the test fluid - in this case directly in front of and behind the channels of the microstructure device.
The heat transfer between the wire and the surrounding medium depends on the thermal conductivity of the medium. When it is changed the transferred heat and consequently the temperature and the electrical resistance of the wire will change. Therefore the measurement of the thermal conductivity can be attributed to the determination of the electrical resistance of the wire.
The idea now was to generate a multi-channel polymer microstructure device and a hot wire sensor frame placed as close as possible to the inlet and the outlet of the flow system to measure the change of the wire resistance. In figure 4 and 5, the specific design of the sensor frame is shown as 3D CAD drawing. Figure 6 shows a first sample of the hot wire residence time sensor frame.
The sensor consists of a meandric arrangement of the wire on a plastic frame.
-5
The frame dimensions are 14*14*2mm . The wire is mounted onto the frame by hand. The combined sensor can be seen in figure 6. After attaching the wire, the frame is installed in the microstructure reactor with the wire aligned directly in front of and behind the micro channels.
Fig.1.Micro stereo lithography process. The left scheme shows how a polymer device can be manufactured by using micro stereo lithography. On the right top, a three dimensional CAD model is shown, used as pre-process data for the generation of a polymer crossflow microstructure heat exchanger device shown (right bottom). The lines from the CAD model have been followed by the laser focus spot to polymerise the plastic monomer and therefore generate the walls of the device
Fig. 2: polymer counterflow microstructure heat exchanger made by micro stereo lithographyDevices like this one are relatively easy to manufacture and useful either for fast experimental analysis of special design variations or for visualization of device design
Fig. 3: Electrically heated polymer microstructure heat exchanger made by micro stereo lithography. Here, a cobaltchloride-isopropanol-water solution is ducted through the device and heated. By increasing the temperature above 48°C, the colour of the solution is reversibly changed
Fig. 4: 3D - sketch of the multi channel microstructure device and the hot wire sensor frame at the inlet (bottom) and the outlet
Fig. 5: Detail of the hot wire arrangement on the sensor frame
Fig. 6: Sample frame of the hot wire residence time sensor. The very thin wire stretched from one side to the other is well to see
The wire temperature chosen for the measurements was relatively low (approx. 40 K above the inlet temperature of the gases and also the temperature at which the sensor was assembled). Under these circumstances it can be assumed that the gas parameters only change to a small degree and these can be neglected. On the other hand, the wires themselves are also affected by the higher temperature - they will expand slightly. But with an thermal expansion coefficient of 8,8-10-6 K-1 the expansion is in the area of 3,5^m and can be neglected.
Besides the material parameters of the wire, the time constant of the sensor depends on the diameter of the wire. Experiments showed that a platinum-iridium wire of 12 ^m in diameter has a time constant in nitrogen of approximately 10 ms. In these preparing measurements, the electrical current through the wire was changed corresponding to a rectangular function; consequently the released heat and also the temperature changed and the change in the electrical resistance of the wire was measured using a Wheatstone bridge. The time constant can be taken from figure 7, which also shows the time constant in helium of approx. 4 ms. The determined time constant is short enough to measure the RTD.
Fig. 7: Measurement of the time constant of the sensor system in nitrogen (left) and helium (right)
The micro channel reactor for the first, orienting experimental work provides 9 layers with 14 micro channels each. The dimensions of the micro channel crosssection are 400x800^m providing a length of 10mm. The connectors for the fluids are attached to the sensor frame. The gas is led to the reactor by a diffusor system providing an opening angle of 5°, resulting in a length of 39mm. The complete system shown in figure 8 is glued together.
Fig. 8: Components of the RTD measurement system. From the right to the left: Diffusor system for the gas inlet, hot wire sensor frame (not wired yet), micro channel reactor
The electrical resistance of the sensor wire is measured by means of a Wheatstone bridge and amplified with an adjustable amplifier (offset and amplification). The sensor signals are recorded using a computer which also controls the experimental course and set-up (offset and amplification of the amplifier, setpoint of the MFCs and the magnetic valves (see next section)).
With this set-up the integral RTD of the whole reactor is measured, not merely the RTD for one channel. Under these circumstances inequalities between the channels will influence the measurements.
Experimental Set-up
The use of the TCD as sensor requires the change of the thermal conductivity of the flowing medium. Therefore a marker (tracer) was used which has a significantly different thermal conductivity than the carrier used - helium as the tracer and nitrogen
as the carrier ^ = 0 154 V-* • ' = 0 026 w/.*
The step function was used as the test function. The RTD can be calculated from the step response by differentiating. The step (concentration step of helium in nitrogen) was realized by a combination of mass flow controllers (MFC) and magnetic valves. The experimental set-up was realized at the Karl-Winnacker-Institute of the DECHEMA. A principle sketch of the experimental set-up is shown in Fig. 9.
of a polymer multi micro channel reactor
The mass flow controllers were used to create constant flows for the tracer (MFC B2) and the carrier (MFC A1 and A2). Like all controllers, mass flow controllers need some time to establish new constant flow values. As this time is far too long to carry out a step function successfully a magnetic valve was used additionally. When the valve was switched between the constant carrier flow through MFC A2 and the tracer flow through MFC B2 in a sufficiently short time (< 10 ms, see Burkert), a step function is realized. With this construction only steps between 0 and 100% tracer concentration can be obtained. To obtain smaller steps a bypass for the carrier is achieved by MFC A1.
First Results
Figure 10 shows the course of the sensor signal as a function of time at the inlet and the outlet of the microstructure. The conditions were: throughput 0.5 L/min, mean residence time (t) 99 ms, concentration step from 0 to 10 vol-% He.
If the measurements shown in figure 10 (smoothed by a moving average filter) are normalized (standard step von 0 to 1) and differentiated the residence time distributions at the inlet and the outlet, shown in figure 11, are obtained. It is evident that the RTD measured at the inlet (TCD I) is broadened by the microstructure (measured by TCD II). The squared standard deviation of these curves was calculated. The squared standard deviation of the RTDs of the microstructure can then be calculated by subtracting both values, like it is described by Aris (1959). Table 1 summarizes the values.
Using a dispersion model, the Bodenstein number Bo gives the ratio of convective transport to effective dispersion. In equation (1), the definition of the Bodenstein number is given. In principle, higher values of Bo mean smaller dispersion, lower values mean higher dispersion.
The Bodenstein number can also be calculated from the squared standard deviation of the RTD of the microstructure (see equation (2)).
V • L
Bo = R
D (1)
Equation (2) can be used to calculate the Bodenstein number from the squared standard deviation of the RTD of the microstructure .
1
a
8
a
(2)
The Bodenstein number calculated for the above-mentioned conditions is 9.5.
In figure 10, in the time region of approx. 0.5 to 0.6 s a negative amplitude of the measurements on the inlet can be seen, which was also visible in following experiments under modified conditions. The reason for this phenomenon is unknown.
ra
c
ra
w
o
w
c
a)
w
0,4 -i-
0,2--
% 0,0-
-0,2--
t = 99.0 ms (V_ges=0.5 L/min) Step: 0 -> 10 % (vol) He in N2 o ..............
-0,4
" • TCD II (outlet) ipg&JA JJ. s±
1 \Jf 1 IJTO 1
f
■ 1 1 1 O 1 1 1 O 1 1 1 O 1 " To ■ 1 1 IO 1 Ccnrrr?Tr?YT?TT77Tt77TtT>v>
H
0,0 0,2 0,4 0,6 0,8 1,0
time after step [s]
1,2
1,4
1,6
Fig. 10: Step response measured with the described micro structure sensor. Measurement parameters were: throughput: 0.5 L/min; x: 99 ms; step: 0—>10 Vol-% He in N2
Table 1: Squared standard deviation and Bodenstein number
squared standard deviation 2
throughput mean residence time step u inlet outlet reactor Bodenstein number
0,5 L/min 99,0 ms 0 ^ 10 Vol-% 0,931 1,232 0,301 9,5
o 0,3-
.q
's_
w
T3
<D
a)
o
c
a)
T3
w
a)
0,2-
0,1 -
0,0
h
0,396
J—I—1
0,594
8 10 12 dimensionless time [-]
+
+
+
14
+
0,792 0,990 1,188
time after step [s]
1,386
16
H
1,584
Fig. 11: Residence time distribution measured at the inlet and the outlet of the microstructure device arrangement. Measurement parameters were: throughput: 0.5 L/min; x: 99 ms; step: 0—»10 Vol-% He inN2
Summary
A sensor system was developed for the measurement of the residence time distribution (RTD) for gas flow through microstructured devices. This sensor system is based on thermal conductivity detectors and was tested on micro reactors constructed by FZ Karlsruhe by micro stereo lithography. The method used provides an integral RTD in the reactor, this means the RTD is not measured for one special channel but that one for the whole reactor including inhomogeneous distribution of flow between different reactor channels. In the result it gives the value which is relevant for the chemical reaction, assuming that all channels are catalytically active.
The sensors were integrated into the microstructured device in front of and behind the channels. For the measurement of the RTDs helium concentration steps in nitrogen were used. A combination of MFCs and magnetic valves was used to create these steps.
It could be shown that the experimental set-up and the sensor system are suitable for determining the RTDs. Different throughputs and concentration step heights were applied and the behaviour of the microstructured device was measured. As a result of these experiments small Bodenstein numbers were obtained. This means a high dispersion in the fluid system. To obtain more precise data further detailed research is necessary. The new sensor system provides an opportunity for this.
One possible explanation for the Bodenstein numbers found could be the uneven flow distribution to the different channels of the microstructure. This leads to a higher effective dispersion in the microstructure as a whole and hence to lower Bodenstein
numbers. Experimental determination of the flow distribution for single channels with hot wire anemometry, like it was performed by Pfeifer et al (2007), demonstrates that this uneven flow really exists.
References
1. Aris, R., Chem. Eng. Sci. 9 (1959), 266
2. Baerns, M., Hofmann, H., Renken, A., Chemische Reaktionstechnik. Verlag
Georg Thieme, Stuttgart, New York 1999
3. Bier, W., Keller, W., Linder, G., Seidel, D., Schubert, K., Martin, H., Gas to gas heat transfer in micro heat exchangers. Chem. Eng. Proc. 32 (1993), 33
4. Brandner, J.J., Anurjew, A., Bohn, L., Hansjosten, E., Henning, T.,
Schygulla, U., Wenka, A., Schubert, K., Concepts and realization of microstructure heat exchangers for enhanced heat transfer, Experimenthal Thermal and Fluid Science 30 (2006), 801-809
5. Bürkert: Data Sheet: Magnetic Valve Type 6124. Ingelfingen
6. Gebhardt, A., Rapid Prototyping, Hansa, München, 1996
7. Hessel, V., Hardt,S., Löwe, H., Chemical Micro Process Engineering Vol.
1+2, Wiley-VCH, 2005
8. Ikuta, K., Hiwatari, K., Ogata, T., “Three dimensional integrated fluid systems (MIFS) fabricated by stereo lithography”, Proc. IEEE Int. Workshop on Micro Electro Mechanical Systems, 1-6, MEMS’94, Osio, Japan, 1994
9. Kockmann, N. (Ed.), J.J. Brandner, “Microfabrication in Metals and Polymers”, Advanced Micro- and Nanosystems Vol.5, 267-320, Wiley-VCH, 2006
10. Pfeifer, P.; Schubert, K.: Hot Wire Anemometry for Experimental Determination of Flow Distribution in Multilayer Microreactors, Chem. Eng. J., (2007)
11. Stief, T., Geider, H., Langer, O.U., Brandner, J.J., Schygulla, U.,
Development of a fast sensor for the measurement of the residence time distribution of gas flow through microstructured reactors, Chem. Eng. Journal, (2007):
12. Wolf, D., Experimente und modellgestützte Simulationen zur reaktionstechnischen Charakterisierung von Mikroreaktoren für heterogen katalysierte Reaktionen in der Gasphase. in Richter, Th. (Ed.): Grundlegende Untersuchungen zur Entwicklung, Realisierbarkeit und Charakterisierung von Mikroreaktoren für industrierelevante Reaktionen in der Gasphase (MIKREAK). Abschlußbericht des BMBF-Forschungsprojektes, Förderkennzeichen 16SV671 bis 16SV677
© Eugen Anurjew, Juergen J. Brandner, Edgar Hansjosten, Ulrich Schygulla, Thomas Stief, Klaus Schubert, 2008
