Instrumentinvest Report I-I 13/01 2013-11-29 Aerosol AB 1 Lars Ström Stack monitor testing by means of fluorescent particles Lars Ström Summary Nuclear installations have to monitor effluent air streams for radioactivity. Typically a sample of the air flow is withdrawn, and brought through a pipe to measuring instruments. The International Standard ISO 2889 sets rules for the performance of such stack monitors. This report shows how the standard’s conditions can be verified in a practical case. The demonstration took place in the now closed nuclear power station Barsebäck 1. The ISO 2889 distinguishes between singe-point sampling and multi-point sampling. Single-point sampling is strongly recommended. For a location to be suitable for single point sampling the air velocity shall be about the same over the whole sampling cross-section, and the contaminants to be measured be well mixed into the stream. If such favourable conditions do not exist, then multi-point sampling is an alternative. But then the representativeness of a sample withdrawn has to be proven. The properties of the Barsebäck sampling site and sampling installation were investigated by means of a fluorescent monodisperse test aerosol. The 10 µm particles were introduced into the air stream a distance upstream of the sampling location for the ordinary monitoring. The concentration of particles was measured in the stack by means of a newly developed optical particle counter. Particles were also collected on filters, and the particles counted with computerised image analysis. Air velocity in the sampling section was measured in 16 places. From these measurements the coefficient of variation (standard deviation/mean, COV) was calculated for both air velocity and particle concentration. The COV of air velocity turned out to be 20 %, on the limit for single-point sampling. The particle concentration COV was 69 – 80 %. This very high COV was created on purpose, by selecting the injection point for the test aerosol close to the sampling area. With a longer way for the air to travel, the test aerosol should be better mixed into the air flow and the COV lower. The transmission efficiency of the sampling line between the stack and the measurement equipment was measured. The ratio between the concentration recorded by the collecting filter and the concentration in the stack was 6 – 7 %, well below the standard’s requirement of 50 % for 10 µm particles. This work was supported by the Swedish Radiation Safety Authority, project number SSM2013-876. Document1 ▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬ Instrumentinvest Aerosol AB Tryffelstigen 12 SE-611 63 NYKÖPING Sweden Tel nat 0155-216862 int +46-155-216862 Org.nr 556559-7696 Bankgiro 5395-5639 E-mail lars.strom@instrumentinvest.se www.Instrumentinvest.se Instrumentinvest Report I-I 13/01 2013-11-29 Aerosol AB 2 Lars Ström Contents 1 2 3 4 Introduction ....................................................................................................... 3 1.1 The ISO 2889 rules ..................................................................................... 3 1.2 Control methods ......................................................................................... 3 Equipment used at the Barsebäck demonstration.............................................. 4 2.1 The Barsebäck 1 plant ................................................................................ 4 2.2 Generator for fluorescent tracer particles ................................................... 5 2.3 The turning carrier for sensors ................................................................... 6 2.4 Air velocity measurement .......................................................................... 7 2.5 Particle measurement with an optical counter ............................................ 7 2.6 Particle measurement by means of filter collection ................................... 8 2.7 Evaluation of filter samples ........................................................................ 9 Air flow ........................................................................................................... 11 3.1 Air velocity observations .......................................................................... 11 3.2 Velocity distribution and velocity coefficient of variation....................... 12 3.3 Calculation of the total air flow ................................................................ 13 3.3.1 Turning probe. ................................................................................... 13 3.3.2 Scilab velocity integration................................................................. 14 3.3.3 Flow calculated from the Barsebäck 2 formula ................................ 14 3.3.4 Facility statement of stack flow. ....................................................... 15 Flow of tracer particles.................................................................................... 15 4.1 4.1.1 Optical particle counter ..................................................................... 16 4.1.2 Filter array sampling ......................................................................... 17 4.2 Particle distribution and particle concentration coefficient of variation .. 17 4.2.1 Optical counting ................................................................................ 17 4.2.2 Filter array sampling ......................................................................... 18 4.3 5 Particle concentration observations .......................................................... 15 Calculation of total particle flow .............................................................. 19 4.3.1 Turning probe directly....................................................................... 19 4.3.2 Turning probe velocity and concentration interpolation. .................. 19 4.3.3 Filter array sampling ......................................................................... 19 Transmission efficiency of the sampling line ................................................. 20 5.1 Sampling efficiency in Barsebäck 1 ......................................................... 21 5.2 Contamination of a filter from Test 2 ....................................................... 22 6 Comments on the instrumentation .................................................................. 22 7 References ....................................................................................................... 23 8 Sammanfattning på svenska ............................................................................ 24 Instrumentinvest Report I-I 13/01 2013-11-29 Aerosol AB 3 Lars Ström 1 Introduction 1.1 The ISO 2889 rules The ISO standard “Sampling airborne radioactive materials from the stacks and ducts of nuclear facilities” from 2010, reference 1, sets some rules for the measurement of gases and particles in outgoing ventilation air. The standard refers to the case when a sample must be withdrawn for radiometric analysis. The first rule is that one-point sampling is to be preferred. Multi-nozzle samplers have higher internal losses than simple ones. From this follows the second rule that the sampled air stream should be well mixed, so the one-point sample is representative for the whole stream of air. The third rule is that the air moves straight and even at the sampling point. Finally it is required that 10 µm particles shall not undergo any serious losses on its way to the measurement instruments. The standard expresses these rules quantitatively. 1.2 Control methods The purpose of this demonstration is to show how the ISO rules can be checked. Some new instruments, especially designed to investigate compliance with the ISO rules, are tried for the first time. The demonstration is limited to the control of particles. The rules also apply for gases, but if an installation is good enough for particles, it will also do for (inert) gases. Stack monitor testing is done by generating test particles in the stack air stream. If the particles are 10 µm and monodisperse, sampling of these particles will relieve the distribution pattern of such particles. Alternatively a polydisperse test aerosol is used, which is much easier to generate. Then a size-selective optical sampler is used for analysing for 10 µm particles. Unfortunately optical particle counters suitable for use in an industrial ventilation environment cannot measure 10 µm particles because of inlet losses. At best the relative concentrations can be obtained. This suffices for calculation of the COV, but is not enough for determining the transmission efficiency of the sampling line. This transmission is often the weakest link in stack monitoring. In the past Swedish monitoring installations have been tested with a monodisperse test aerosol with a tracer substance and filter sampling. The same tracer is used for determining the transmission efficiency of the sampling line. This technique works well, but is quite expensive in the steps of particle fabrication, tracer marking and analysis. There is also the risk of contamination of samples. The dilution from generator to sample is of the size of order 10-6. The complicated preparation of the test aerosol and its injection must be well separated from sampling and analysis. A better method is desirable. Instrumentinvest Report I-I 13/01 2013-11-29 Aerosol AB 4 Lars Ström The idea of measurements based on particle counting rather than bulk analysis of a tracer in a sample has been investigated. The main advantages with the new technique was expected to be 1. 2. 3. 4. Easy analysis of particles in the air, by means of a new optical counter Easy analysis of particles collected on filters, by particle counting Risk of contamination errors in samples much reduced Easier manufacture of tracer particles To realize such a test method Instrumentinvest Aerosol AB has developed Generator for fluorescent monodisperse test aerosols, for field use Optical particle counter for fluorescent particles with low inlet losses Technique for counting fluorescent particles on filter 2 Equipment used at the Barsebäck demonstration 2.1 The Barsebäck 1 plant The nuclear power station Barsebäck 1 was shut down in 1999. The ventilation system and the stack monitor are still operating, and were used for this demonstration. Niklas Wallgren represented the Barsebäck plant, Tommy Hansson and Hannes Tingfors from ISS participated in the practical work. The operations in the plant took place 19 – 23 August 2013. Sections through stack are shown in Figure 2 and Figure 4. The stack is not circular. The permanent stack monitor has a sampler of the multi-nozzle type, with four nozzles of 50 mm diameter. The sampled flows are brought together in a transport pipe of inner diameter 100 mm, carrying the sample down to the measurement equipment, at the flow rate of 56 litres/second. There a secondary flow of 0,60 litres/second is withdrawn, with particle collection on filters of 50 mm diameter. Above the sampler there is a grated gangway or bridge over the stack, which is accessible over a ladder on the stack inner wall. From this bridge the stack measurements of this report are made. Instrumentinvest Aerosol AB Report I-I 13/01 2013-11-29 5 Lars Ström 2.2 Generator for fluorescent tracer particles Based on earlier work (reference 2) a vibrating orifice aerosol generator for field use was developed. Such a generator is deceptively simple; the central part fits easily in the hand. However, for field use, in an environment where the equipment can be contaminated with radioactivity and must stand rough handling and decontamination, the equipment became rather large, Figure 1. Barsebäck 2013 report Figure B Figure 1. Generator for monodisperse fluorescent particles of 10 µm diameter. On the front panel there are controls for the nozzle vibration frequency, amplitude and wave form. To the right there are valves and meters for the flow of liquid. To the bottom left there is room for a filter for the air pressurizing the interior of the box, to prevent inside contamination. The generation rate is typically 115 000 particles/second. The particle size at the Barsebäck tests was calculated to be 8,7 µm, based on liquid flow and disintegration frequency. The corresponding aerodynamic diameter is 9,1 µm. The diameter measured by means of a microscope was 8,9 µm. The particle diameter varied slightly during the tests, because the pressure, driving the flow of liquid, had to be restored at intervals. Instrumentinvest Aerosol AB Report I-I 13/01 2013-11-29 6 Lars Ström 2.3 The turning carrier for sensors To measure air velocity and to sample particles on the same level as the ordinary stack effluent monitor, probes was mounted in that level, hanging from the grated bridge above, Figure 2. The probes were set on the ends of a cross bar, fixed to the lower end of a vertical axis. The axis was rotated by means of a lever on the upper end of the axis. This rotated the cross bar, so that the probes were moved on a circle. There were four predefined settings for the lever, a, b, c, d. Readings taken on the same letter are simultaneous, but the particle and velocity probes are separated by the length of the cross bar, 600 mm. Figure 2. Cross and longitudinal sections of the Barsebäck 1 stack at the sampling level of the stack monitor. The vertical axis was mounted in a stand hanging on the bridge handrail. The whole setup could be moved on the rails to four different positions in the air stream, 1, 2, 3, 4. Instrumentinvest Report I-I 13/01 2013-11-29 Aerosol AB 7 Lars Ström 2.4 Air velocity measurement The velocity probe was a Prandtl tube, diameter 7 mm (also called Pitot – static tube). The differential pressure output was measured with a micro manometer Airflow PVM 100. Each record is the mean of five individual readings. The manometer was calibrated against a laboratory standard Furness MDC FC00 0 - 100 Pa and an inclined tube alcohol meter, Sartorius 655. The Airflow tended to show about 0,2 % too low. This small correction was not applied to the measurements. 2.5 Particle measurement with an optical counter The particle concentration was measured with a newly developed optical particle counter. It has a shrouded probe inlet, with the same proportions as described in reference 3. The inlet diameter is 34 mm, the flow 3 litres/second, inlet velocity 3,3 m/s. The outer flow in the shroud is filtered and used as a sheath of clean air in the optical sensing section. The geometry predicts that 2/3 of the total air flow should be filtered and form the sheath, the rest counted for particles. Unfortunately it is quite difficult to demonstrate the sampling properties of a shrouded probe, so the evaluation is based on the assumption that the free air concentration in the stack is three times the concentration measured by the optical counter. Barsebäck 2013 report Figure M Figure 3. The optical particle counter. The tracer particles are fluorescent. This discriminates against other particles in the air. The fluorescence is exited with an axial illumination, in the direction of the incoming aerosol flow. Fluorescent light is emitted in all directions, permitting a simple concentric geometry with straight-through air flow. Instrumentinvest Aerosol AB Report I-I 13/01 2013-11-29 8 Lars Ström 2.6 Particle measurement by means of filter collection Particle distribution was also measured with ordinary 50 mm glass fibre filters, mounted on a cross-shaped frame and hoisted to the level of the sampling section, Figure 4 and Figure 5. Figure 4. The cross-shaped support for filters A … G (blue) together with the permanent stack sampler manifold with its sampling points R … U (green). Barsebäck 2013 report Figure C Figure 5. Filter holders A and B on an arm of the cross-shaped support frame. The air inlets are here capped. Instrumentinvest Aerosol AB Report I-I 13/01 2013-11-29 9 Lars Ström The filters were connected to the suction side of a blower, equipped with an accurately manufactured Venturi tube on the inlet. Inlet pressure and differential pressure was indicated by manometers, controlled against an alcohol filled U tube manometer. The venturi compared well with a sharp-edged orifice. The flow was kept at 3,43 l/s. The filters were mounted in filter holders with shaped inlets, Figure 5. The flow through each filter in its holder at a given pressure drop was measured in advance. The variation in flow was small, the standard deviation less than 2 % of the mean. So the flow through each filter is assumed the same, 0,49 l/s, or 147 litres during the sampling time of 5 minutes. – Inlet diameter was 0,93 cm2, inlet velocity 5,25 m/s. The same kind of 50 mm diameter glass fibre filters were also installed in the filter holders of the ordinary stack sampler. 2.7 Evaluation of filter samples The fluorescent particles collected on the filters were photographed, and the fluorescent particles appeared as “blobs” in the images. A computer program was used to find particles, reference 4. Figure 6 shows a photo where the blobs identified by the program have been coloured white. Barsebäck 2013 report Figure E Figure 6. Particle “blobs” from particles collected on filter 387 in position G in Test 1. Figure 6 shows a small area of a filter, processed by the JMicroVision program. The image shows 14,9 mm × 9,5 mm of the filter paper. Most blobs are 11 – 15 pixels in area, or 70 to 80 µm across, much larger than the particles they represent. The whole filter holds 745 blobs in the size range 10 – 96 pixels, of which 560 are estimated to represent particles. Note that the number label contributes to the background with some large blobs. Instrumentinvest Report I-I 13/01 2013-11-29 Aerosol AB 10 Lars Ström The blobs from the image analysis are of different sizes, and can be displayed in a histogram. Figure 7 shows such a histogram for Filter 385, used in position E in Test 1. There is also plotted the histogram of the unused filter 388, spare filter in Test 1. Only blobs in the range 10 to 96 pixels were regarded. Number of blobs per class. 250 200 150 Filter 385 Cross position 1E Photo P1010987.jpg Filter 388 not used Photo P1010989.jpg 100 50 0 1 11 21 31 41 51 61 71 81 91 Class number=blob size in pixel. Barsebäck 2013 report Figure F Figure 7. Histogram over blobs found in filters 385 and 388, Note that pixel here is a unit of surface. The high background was not expected. Theoretically the lower limit of detection is one particle. In the present situation at least a hundred particles are necessary to clearly stand out from the background. Test 3 was intended to demonstrate the contribution to the background from the air in the stack. The contribution turned out to be small. The background seems to stem mainly from the evaluation, so it should be possible to bring the background down considerably. Particle generation as a stream of droplets involves the risk of droplet collision and the formation of double-sized particles, doublets. What impression a doublet would make in an image is difficult to say, but a first guess is twice as many pixels as a singlet. In histograms like the one above one can sometimes see a small increase in blob numbers around 30 pixels. Instrumentinvest Report I-I 13/01 2013-11-29 Aerosol AB 11 Lars Ström 3 Air flow The air flow in the stack should be straight and evenly distributed where the sample is drawn. If one-point sampling is to be considered, ISO2889 requires that the velocity in the central 2/3 of the cross-sectional area has a coefficient of variation (COV) of less than 20 %. COV is the ratio of the standard deviation to the mean. To find the velocity mean and distribution the standard recommends the conventional geometry of two lines of measurements at right angles across the air conduit. Such measurements cannot be taken in the Barsebäck stack, so a turning carrier suspended from a grating above the sampling location was employed, as described in section 2.3. The standard also requires that that the air is not rotating in the stack. The average resultant angle should be less than 20 degrees. 3.1 Air velocity observations Air velocity was measured as described in section 2.4. Measurements were taken in16 points distributed over the stack cross section at the sampling level of the ordinary stack sampler, in Positions 1a … 4d, Figure 2. The following table summarizes the observations. Probe position 1a 1b 1c 1d 2a 2b 2c 2d 3a 3b 3c 3d 4a 4b 4c 4d 5a 5b 5c 5d 6a 6b 6c 6d Test 1 4,73 4,40 3,96 4,36 4,12 4,32 3,61 3,83 5,29 5,48 6,16 5,72 6,14 6,90 7,21 6,27 Velocity m/s Test 2 Test 3 4,73 4,73 4,28 4,12 4,40 4,08 4,65 4,47 4,40 4,69 4,32 4,24 3,72 3,70 3,74 3,83 6,63 7,12 7,35 6,83 5,48 6,30 6,78 6,90 Test 4 4,86 4,08 4,12 4,47 4,40 4,32 3,51 3,61 5,16 5,77 Instrumentinvest Aerosol AB Report I-I 13/01 2013-11-29 12 Lars Ström In Test 2 the turning carrier was not correctly placed in the positions 4 and 5. The unintended positions have been reconstructed by means of photographs taken during the tests, and are called 5a … 6d. In Test 3 the intention was to measure the “natural” background of particles in ventilation air, and samples were taken only in eight positions. In Test 4 there was a failure in the instrumentation, and the test was interrupted. The observed velocities are in good agreement. 3.2 Velocity distribution and velocity coefficient of variation By means of a computer program, Scilab, reference 5, the velocities observed in Test 1 and Test 2 were converted to curves of equal velocity over the stack cross section. Within the area covered by the observations the program interpolated with credible results, but when extrapolating outside this area, out to the stack wall, the program produced evidently erroneous lines of equal velocity. To rein in the program a number of reasonable velocities were invented for the stack wall. The result is shown in Figure 8. Test 2 is less reliable, because of the less favourable positions of the measurement points, and the uncertainty in the reconstruction of their position. Figure 8. Curves of equal velocity generated by Scilab from data of Test 1 and Test 2. The grey area marks the central 2/3 of the whole stack area. The blue circles indicate imagined measurement points on two orthogonal lines. A measurement along orthogonal lines can be simulated by calculating the velocities at the required measurement points by means of the Scilab. The resulting COV should be reduced because of the smoothing effect of interpolation. Instrumentinvest Report I-I 13/01 2013-11-29 Aerosol AB 13 Lars Ström In the table below the COV is calculated on four data sets, all from the central 2/3 of the cross section Data set Velocity from the suspended probe Test 1 Velocity from the suspended probe Test 2 Velocity on orthogonal lines, from Scilab, Test 1 Velocity on orthogonal lines, from Scilab, Test 2 Number of data points 13 11 14 14 COV % 20 22 20 19 The coefficient of variation is near 20 %, on the limit of the acceptable for onepoint sampling according to ISO 2889. The mean and the standard deviation are calculated on a limited number of observations. Both numbers have a statistical uncertainty, and their quotient, the COV, still more so. 3.3 Calculation of the total air flow To calculate the total particle emission with the exhaust air the flow must be known. The conventional way to calculate the total flow is to allocate a fraction of the stack area ΔA to each velocity record. The whole area A is ∑Δ A= A The whole air flow Q is ∑ΔA∙v = Q The area allocation is a somewhat subjective task. Usually the measurement points for velocity are so distributed that the area fractions are of equal size. 3.3.1 Turning probe. Each measurement point is allocated the same area, A/16, although the distribution of measurement points could be more even. This gives equal weight to each point, and the flow is simply the mean velocity times the total area, 6,20 m2 Test 1 vmean = 5,16 m/s Q = 32,0 m3/s Test 2 vmean = 5,48 m/s Q = 34,0 m3/s A carefully performed measurement in the field, under favourable conditions, should in most cases come within 3 % from the true value. In Test 1, and especially in Test 2, the results must have larger error margins mainly because of too few measurement points. Instrumentinvest Report I-I 13/01 2013-11-29 Aerosol AB 14 Lars Ström 3.3.2 Scilab velocity integration The interpolation program used for creating the curves of equal velocity in Figures 6 and 7 can also be used for summing velocities. Because of the irregular shape of the stack cross section, it was divided in 12 rectangular slices. The mean velocity of 100x100 points in each slice was multiplied with the slice area, and the flows of the slices summed. Test 1 Q = 30,7 m3/s Test 2 Q = 34,3 m3/s The result might be influenced a little by the three invented velocities, necessary to keep Scilab within bounds outside the interpolation area. 3.3.3 Flow calculated from the Barsebäck 2 formula Following the work in Barsebäck 2 in 2001, reference 6, a formula was devised for the stack flow. It is based on measurements of velocity at two predetermined positions, from the bridge across the stack, above the sampling area, REFÖ1 (-600, -400) and REFÖ2 (-600, -1000). The data were taken with a hand-held thermal flow meter, Compuflow GGA 665 P. The same instrument was now used in Barsebäck 1. The stack in Barsebäck 2 has the same dimensions as the stack in Barsebäck 1. Measurements were also taken in corresponding positions at the lower grating, REFU1 and REFU2. The velocities are the mean of five individual observations at each point Monday 13:20 Tuesday 11:20 Wednesday 13:30 REFU1 8,60 6,62 6,90 REFU2 8,60 8,02 8,18 REFÖ1 7,32 8,44 6,94 REFÖ2 8,76 7,38 6,08 The Barsebäck 2 formula is Q=5,11×(vREFÖ1+vREFÖ2)/2 [m3/s] , [m/s] The velocity readings at the upper grating translates to the following flows Monday 41,1 m3/s Tuesday 41,4 m3/s Wednesday 33,3 m3/s The stack flow seems to vary considerably. The measurements at the lower grating also show great variations. It is unlikely that the stack flow varies that much. The Compuflow meter might not be suitable for flows of very high turbulence, even if as here each flow is the mean over 2 × 5 individual observations. It should also be noted that the I-I formula is devised for Barsebäck 2. In Barsebäck 1 the grating of the bridge across the stack is less permeable for the air flow because of heavy deposits on the grating. This may increase the velocity in the free space, where the readings are taken, giving a false impression of a higher flow. The temperature in the stack was stable, 22,5 – 23,5 C. Instrumentinvest Report I-I 13/01 2013-11-29 Aerosol AB 15 Lars Ström 3.3.4 Facility statement of stack flow. From the control room of the power station are reported the following flows, measurement point K305 in the stack: Min m3/h Monday 164742 Tuesday 164693 Wednesday 164698 Thursday 164728 Max m3/h 165215 165193 165215 165175 Mean m3/h 164966 164962 164964 164952 This flow is measured with a probe in the stack sampling area. The recorded flow is extremely stable. The mean flow is 45,8 m3/s, much higher than the results obtained in the manual measurements. The total flow calculation is used as a check on the measurements. The following table shows that the measurements could be better, or the stack flow more stable. Measurement 3.3.1 Prandtl 16 points Test 1 3.3.1 Prandtl 16 points Test 2 3.3.2 Scilab integration Test 1 3.3.2 Scilab integration Test 2 3.3.3 Barsebäck 2 formula 3.3.3 Barsebäck 2 formula 3.3.3 Barsebäck 2 formula 3.3.4 Facility Day Tuesday Wednesday Tuesday Wednesday Monday Tuesday Wednesday Mon- to Friday Flow m3/s 32,0 34,0 30,7 34,3 41,1 41,4 33,3 45,8 The Prandtl tube measurements are regarded as the most reliable. For calculation of tracer particle concentrations in the stack (see below), a constant flow of 33 m3/s is assumed. 4 Flow of tracer particles The purpose of this section is to describe two new particle counting techniques for obtaining the COV for the particle distribution over the sampling section in the stack. The injection point for tracer particles was chosen to create a high COV in the sampling section. The high COV observed is thus not a property of the Barsebäck plant, but depends on the selected point for tracer injection. 4.1 Particle concentration observations Particles were measured in two independent ways: with the optical particle counter on the turning mount, and by means of an array of filters on a frame hoisted to the sampling level in the stack. Booth methods were adapted to the peculiarities of the Instrumentinvest Report I-I 13/01 2013-11-29 Aerosol AB 16 Lars Ström Barsebäck stack. The particle flow was also measured by the ordinary stack monitor. 4.1.1 Optical particle counter The optical particle counter was set to sample a few minutes in each of the 16 measurement points of the turning probe. The sampling time was made longer where the particle concentration was expected to be low. The following concentrations were observed. Particle concentration, particles/litre Test 1 Test 2 Test 3 Test 4 Tracer Tracer No tracer Tracer source: source: source: Lower Lower Fan floor grating grating Sampler position 1a 1b 1c 1d 2a 2b 2c 2d 3a 3b 3c 3d 4a 4b 4c 4d 5a 5b 5c 5d 6a 6b 6c 6d 0,04 1,52 0,78 0,10 1,03 4,45 6,95 2,68 8,81 8,69 9,32 8,15 1,19 2,56 4,67 0,51 0,03 1,13 0,78 0,07 0,49 2,41 6,43 1,63 0,02 0,01 0,01 (0,08) 0,01 (0,21) 0,01 0,00 0,41 1,48 1,74 1,50 1,12 0,79 0,52 0,48 1,00 0,63 1,33 4,48 4,47 1,91 6,08 5,90 10,70 9,75 The Lower grating is about 9 m below the sampling area; the Fan floor further upstream the ventilation channel. In the No tracer column there are two records within brackets. In those cases some work was deliberately undertaken in the stack, liberating much dust from the gratings and other dirty surfaces. Instrumentinvest Report I-I 13/01 2013-11-29 Aerosol AB 17 Lars Ström 4.1.2 Filter array sampling The cross-shaped frame mentioned in 2.5 was used for obtaining filter samples from the stack monitor’s sampling level. Sampling time was 5 minutes, sample volume 147 litres/filter (Test 3 294 litres/filter). The filters were analysed for particles with the photo - image analysis method described in 2.7. The following blob counts were obtained. Position on frame Test 1 Number of blobs A 176 B 616 C 548 D 196 E 1191 F 1817 G 745 H Spare 189 Background 185 Conc part/litre -0,1 2,9 2,5 0,1 6,8 11,1 3,8 Test 2 Number of blobs 198 813 210 100 519 1574 764 113 137 Conc part/litre 0,41 4,59 0,50 -0,25 2,60 9,77 4,26 Test 3 Background measurement Number of blobs 91 117 133 207 111 188 155 108 After subtraction of the background the particle concentration was calculated. The background was taken from positions A and D, which were outside the particle flow, and the spare filter H, which was never used. Test 3 was made especially for investigating the background of particles in the air. Evidently the air contributes very little to the background in comparison with other sources. 4.2 Particle distribution and particle concentration coefficient of variation The COV:s are for particle concentrations, not particle flows. 4.2.1 Optical counting The concentrations observed by means of the turning probe in Test 1 and Test 2 were converted to curves of constant concentrations by means of the program Scilab. As for the velocity, the measured points had to be completed with some invented points at the stack walls to keep the computer program’s extrapolations within reasonable limits. The obtained level curves are shown in Figure 9 below. The differences between the distributions depend on the sampling pattern; the particle source was the same. Instrumentinvest Report I-I 13/01 2013-11-29 Aerosol AB 18 Lars Ström Figure 9. Particle concentration levels in Test 1and 2. 4.2.2 Filter array sampling The cross has only four sampling points within 2/3 of the stack area, B, C, F, G. Four data points is very little for estimating the COV. Test 1 2 Number of data points 4 4 Mean particle conc. part/l 6,72 4,78 Standard COV deviation part/l % 4,66 69 3,81 80 The table below summarizes COV calculated on different data sets, based on optical counting and filter sampling, all taken within the central 2/3 of the cross section. Table xx Particle COV calculated on different data sets Data set Concentration from the turning probe Test 1 Concentration from the turning probe Test 2 Two orthogonal lines Test 1 Two orthogonal lines Test 2 Filter array sampling Test 1 Filter array sampling Test 2 Number of data points 13 11 14 14 4 4 COV % 72 67 76 90 69 80 The coefficient of variation is very high, and the ISO rules do not permit one-point sampling in this situation. With the particle source further upstream, the COV becomes smaller. The limited set of data collected during the interrupted Test 4, when the tracer particles were injected from the Fan floor gives a COV of 49 %. Instrumentinvest Report I-I 13/01 2013-11-29 Aerosol AB 19 Lars Ström 4.3 Calculation of total particle flow To investigate the consistency of the obtained numbers, the total flow of tracer particles is calculated. This is also a demonstration of the use of data obtainable with the particle counting technique. Each velocity record v and its corresponding product of velocity and concentration v∙c is allocated a fraction ΔA of the whole stack area A ∑ΔA=A The whole particle flow P is ∑ΔA∙v∙c= P The generator for tracer particles worked at 115 000 p/s. There are no known particle losses (except a few per cent doublets) between generation on the lower grating and sampling below the upper grating. The measured particle flow should equal the generator output. 4.3.1 Turning probe directly. Particle concentration was measured in 16 points. As with the air flow, each measurement point was allocated the same area, A/16. The particle flow is calculated to be Test 1 Test 2 127 595 particles/second 133 875 particles/second The particle flow is higher than the generation rate. The fault may lie in the area allocation. 4.3.2 Turning probe velocity and concentration interpolation. Based on turning probe data the Scilab program interpolates a nearly continuous distribution of velocity and concentration, with 120 000 area fractions. Multiplying each velocity with its corresponding concentration and doing the sums, the following results were obtained Test 1 Test 2 98 597 particles/second 89 393 particles/second These particle flows are lower than generator output, 115 000 p/s. In this case the error source could be the invented data points at the stack wall, necessary to prevent the Scilab program from extrapolating to impossible values. In Test 2 the unsuitable sampling points also contribute to the error. It is astonishing that the two estimates of the particle flow, based on the same data set from the optical counter, can be so different. This deserves further study. 4.3.3 Filter array sampling For each filter on the filter array hoisted to the sampler level a concentration is calculated. From the velocity distribution shown in Figure 8 the velocity of air at each filter is obtained. Then each filter is allocated a surrounding area, for which Instrumentinvest Report I-I 13/01 2013-11-29 Aerosol AB 20 Lars Ström the filter is assumed to be representative. The size of this area is an estimate. With these numbers the total flow of particles can be obtained. Test 1 Filter position 1A 1B 1C 1D 1E 1F 1G Concentration particles/litre -0,1 2,9 2,5 0,1 6,8 11,1 3,8 Area fraction Test 2 Filter position 2A 2B 2C 2D 2E 2F 2G Concentration particles/litre 0,41 4,59 0,50 -0,25 2,60 9,77 4,26 Area fraction 0,19 0,15 0,15 0,19 0,12 0,12 0,08 0,19 0,15 0,15 0,19 0,12 0,12 0,08 Air velocity m/s 6,3 6,1 4,5 4,1 3,8 4,9 5,0 Total flow Particle flow particles/second -724 16620 10325 350 19342 40465 9442 95821 Air velocity m/s 7,0 6,5 4,9 4,1 3,5 4,4 6,8 Total flow Particle flow particles/second 3080 26088 2078 -1216 7347 35638 10578 85422 These particle flows should be the most reliable. The data points are admittedly few in this very uneven flow field, but there are no invented concentrations. Maybe the test aerosol generator output is not 115 000 particles/second; some particles might be deposited inside the generator head because of the very turbulent air flow surrounding it. A further support for these low particle flows can be had from calculating the stack air flow on the same numbers. One gets 31,1 respectively 32,9 m3/second, close to earlier estimates, section 3.3. 5 Transmission efficiency of the sampling line The ISO 2889 requires that the loss of 10 µm particles in the transmission line is less than 50 %. Smaller particles have smaller losses. Larger particles have higher losses, but are supposed to have little significance, if there is no special reason to think otherwise. Formulas have been devised for calculating the transmission losses; most prominent the Deposition computer program, reference 7. Calculations were a help when they predicted that a lower flow should give better transmission in the sampling line of Oskarshamn 3 stack monitor, reference 8. A practical test proved the prediction to be right, at least qualitatively. Generally calculations with the Deposition program have given poor results when compared with measurements on Swedish 100 mm transmission lines. Instrumentinvest Report I-I 13/01 2013-11-29 Aerosol AB 21 Lars Ström The formulas are only as good as the assumptions that must be done about the piping. If the pipes are leaky or dirty, or are not built as assumed the calculations will be misleading. Still calculation is a good help in designing a new system. 5.1 Sampling efficiency in Barsebäck 1 Tracer particles were collected on filters in the filter holders of the ordinary stack monitor. Analysis of the filters gave Test Test 1 Test 2 Test 3 Test 4 Sampled volume Conditions Litres High COV 5187 High COV 3801 Background 2205 Low COV 4330 Blobs counted 1849 See 5.2 279 1596 Concentration, background subtracted p/l 0,23 0 0,24 Fraction of stack concentration 6,6 % 6,9 % The stack concentration is calculated from the tracer generator particle output, 115 000 p/s, divided with the stack flow, 33 m3/s, page 14. The mean stack concentration becomes 3,5 particles/litre. The stack concentration presented to the multinozzle sampler can be also be estimated from the filter array hoisted to the sampler nozzle level. As can be seen in Figure 4, the filters A … G in the array form pairs surrounding the sampler inlets R … U. By forming the proper means, the apparent stack particle concentration is calculated to 3,9 particles/litre, Test 1. In lack of better information it is assumed that the flows through the four nozzles are all the same. The ratio between the concentration recorded with the ordinary monitor filter and the concentration in the stack is used as the transmission efficiency of the sampling line, including the nozzle arrangement. The transmission efficiency is low, 5,9 % 6,9 %. Using the apparent stack concentration, the transmission efficiency becomes 0,23×100/3,9=5,9 %. The stack monitor in Barsebäck 2 was checked in 1999 (reference 6), and the transmission was found to be 24% - 25%. There is a significant difference between the sampling lines in Barsebäck 1 and Barsebäck 2. Maybe there are deposits in the Barsebäck 1 pipe similar to those on the gratings. There are deposits also on the Barsebäck 2 gratings, but not so solid. Instrumentinvest Aerosol AB Report I-I 13/01 2013-11-29 22 Lars Ström 5.2 Contamination of a filter from Test 2 Filter F406 was used in cartridge number 2 for collecting particles in the regular stack sampler, under Test 2. Barsebäck 2013 report P1020005 Figure K.jpg Figure 11. Image of a part of filter F406. Figure 11 shows a portion (16 × 4 mm) of the filter. To the right the filter deposit looks normal; to the left the blobs have different shapes and sizes, indicating that they are not part of a regular sample. Collected samples can be screened for contamination with a stereo loupe, and seen that way the strange deposit looks like a fingerprint. 6 Comments on the instrumentation The air flow and the COV of the air flow were successfully determined with conventional technique, adapted to the circumstances. The observed stack flow did not agree with the flow reported from the plant control room. Tracer particle generation was only partly successful. The particle generation rate may be lower than assumed. The generator is unreliable and stopped working during Test 4. Particle concentrations and the COV of the particle concentrations were successfully measured with the optical particle counter, with some reservation for a calibration factor. Particle flows from different evaluation methods differ significantly. Internal contamination of the counter proved a problem, and some parts of the detector should be made disposable. The electronics of the counter need a makeover. Particle concentration was measured successfully with the filter array and image analysis of the fluorescent particle deposit. The sampling points were too few for the very sharp concentration gradients. Instrumentinvest Report I-I 13/01 2013-11-29 Aerosol AB 23 Lars Ström 7 References 1. ISO 2889:2010 Sampling airborne radioactive materials from the stacks and ducts of nuclear facilities 2. Lars Ström The generation of monodisperse aerosols by means of a disintegrated jet of liquid. The Review of Scientific Instruments 40 (6) 778-782. June 1969. 3. Chandra, S, McFarland, A. R. Comparison of aerosol sampling with shrouded and unshrouded probes. American Industrial Hygiene Association Journal 56(5):459-466. 1995. 4. Roduit, N. JMicroVision: Image analysis toolbox for measuring and quantifying components of high-definition images. Version 1.2.7. http://www.jmicrovision.com (accessed 2 July 2006). 5. Scilab 5.4.1. www.scilab.com 6. Lars Ström Kontroll av utrustning för skorstensmonitering av aerosoler, system 553. Instrumentinvest Aerosol AB rapport I –I 00/01. 2000-02-20. 7. McFarland, A. R., A. M. Nagaraj, H. Ramakrishna, J. L. Rea, J. Thompson. 2002. Deposition 2001. Aerosol Technology Laboratory, Department of Mechanical Engineering Texas A&M University College Station TX. http://www.mengr.tamu.edu/research/AerosolLab/index.htm 8. Lars Ström Kontroll av skorstensmoniteringens funktion vid sänkt provtagningsflöde. Oskarshamn 3, oktober 2003. Instrumentinvest Aerosol AB rapport I–I 04/02. 2004-05-18. Instrumentinvest Aerosol AB Report I-I 13/01 2013-11-29 24 Lars Ström 8 Sammanfattning på svenska Den internationella standarden ISO 2889 anger prestanda för utsläppsmonitering vid nukleära anläggningar. Standarden (mer än 100 sidor med bilagorna) betonar funktionen; det tekniska utförandet exemplifieras, men är inte del av standarden. Standarden skiljer på enpunktsprovtagning i den moniterade luftströmmen, och provtagning i flera punkter i strömmen. Enpunktsprovtagning rekommenderas starkt, men innebär också att det ställs krav på att den provtagna luftströmmen är välblandad. Har inte luftströmmen denna egenskap kan flerpunktsprovtagning tillämpas. I ett sådant fall måste man på något annat sätt övertyga sig om att provet är representativt. För vissa ämnen, bland annat aerosolpartiklar, kan överföringen från provtagningen till analysapparaten medföra betydande förluster. Standarden ställer här det ganska stränga kravet att förlusterna för 10 µm partiklar och mindre får uppgå till högst 50 %. Demonstrationen avser visa hur villkoren från ISO 2889 kan verifieras i ett praktiskt fall. Delvis användes nyutvecklad teknik, anpassad till ISO-kraven. En speciell testaerosol med partikelstorleken 10 µm infördes i luftströmmen, ett stycke uppströms provtagningen för den fasta moniteringen. Partiklarna innehöll ett fluorescent spårämne för att underlätta detektering. Koncentrationen av partiklar mättes i ventilationskanalen, där samtidigt den ordinarie moniteringen tog prov. Två olika tekniker användes, dels en nyutvecklad optisk partikelräknare, dels insamling på filter, med datoriserad bildanalys för att räkna de fluorescerande partiklarna på filtret. Även lufthastigheten mättes i ett flertal punkter. För att en provtagningsplats skall vara lämplig för enpunktsprovtagning kräver ISO 2889 att lufthastigheten är någorlunda jämn; variationskoefficienten (standardavvikelsen/ medelvärdet, COV) för mätningar fördelade över tvärsnittet skall vara mindre än 20 %. Lufthastighetsmätningar utförda med ett Prandtls rör och mikromanometer i 16 punkter gav COV på 20 %, och alltså just godtagbart för enpunktsprovtagning. Även fördelningen av partiklar över provtagningstvärsnittet skall vara jämn. För demonstrationen hade dock generatorn för testaerosol placerats så, att en mycket ojämn fördelning erhölls. COV låg i området 69 till 84 %. Partiklar från längre uppströms i ventilationssystemet kan väntas ha betydligt lägre COV. Överföringen av partiklar från provtagningen i skorstenen till insamlingsfiltret mättes också. Den befanns vara endast 6–7 %, mycket mindre än ISO-kravet 50 %.