Linköping University | Department of Computer and Information Science
Bachelor’s thesis, 16 ECTS | Datateknik
2020 | LIU-IDA/LITH-EX-G-20/031-SE
Measuring the change in
concen-tration of suspended particles in
water using ultrasound
Mäta förändringen i koncentrationen av suspenderade partiklar
i vatten med ultraljud
Oscar Lavén
Martin Hallgren
Supervisor : Adrian Horga Examiner : Martin Sjölund
Upphovsrätt
Detta dokument hålls tillgängligt på Internet - eller dess framtida ersättare - under 25 år från publicer-ingsdatum under förutsättning att inga extraordinära omständigheter uppstår.
Tillgång till dokumentet innebär tillstånd för var och en att läsa, ladda ner, skriva ut enstaka ko-pior för enskilt bruk och att använda det oförändrat för ickekommersiell forskning och för undervis-ning. Överföring av upphovsrätten vid en senare tidpunkt kan inte upphäva detta tillstånd. All annan användning av dokumentet kräver upphovsmannens medgivande. För att garantera äktheten, säker-heten och tillgängligsäker-heten finns lösningar av teknisk och administrativ art.
Upphovsmannens ideella rätt innefattar rätt att bli nämnd som upphovsman i den omfattning som god sed kräver vid användning av dokumentet på ovan beskrivna sätt samt skydd mot att dokumentet ändras eller presenteras i sådan form eller i sådant sammanhang som är kränkande för upphovsman-nens litterära eller konstnärliga anseende eller egenart.
För ytterligare information om Linköping University Electronic Press se förlagets hemsida http://www.ep.liu.se/.
Copyright
The publishers will keep this document online on the Internet - or its possible replacement - for a period of 25 years starting from the date of publication barring exceptional circumstances.
The online availability of the document implies permanent permission for anyone to read, to down-load, or to print out single copies for his/hers own use and to use it unchanged for non-commercial research and educational purpose. Subsequent transfers of copyright cannot revoke this permission. All other uses of the document are conditional upon the consent of the copyright owner. The publisher has taken technical and administrative measures to assure authenticity, security and accessibility.
According to intellectual property law the author has the right to be mentioned when his/her work is accessed as described above and to be protected against infringement.
For additional information about the Linköping University Electronic Press and its procedures for publication and for assurance of document integrity, please refer to its www home page: http://www.ep.liu.se/.
Abstract
Measuring the concentration of particles in water is important in many areas. Indus-tries might measure it to run more efficiently while scientists might measure it to study the pollution of a body of water. Regardless of the area, the standard for taking a measurement is done by filtering out the particles by hand. This is a very slow and expensive method, so other alternative methods have been developed. However, all the alternative methods can only estimate the concentration. Therefore, Deepoid AB aims to investigate if ultrasound can be used to measure the concentration of particles in water.
This thesis shows that a direct ultrasound signal can be used to measure changes in the concentration of particles in water. It also shows how this method is much faster than measuring the concentration of particles in water by hand.
Contents
Abstract iii
Contents v
List of Figures vii
List of Tables ix 1 Introduction 1 1.1 Motivation . . . 1 1.2 Background . . . 2 1.3 Aim . . . 2 1.4 Research questions . . . 2 1.5 Delimitations . . . 2 2 Theory 3 2.1 Standards for measuring the concentration of suspended particles in water . . 3
2.2 Alternative methods for estimating the concentration of suspended particles in water . . . 4 2.3 Ultrasound . . . 4 2.4 Piezoelectric Transducers . . . 5 2.5 Signal Processing . . . 5 2.6 Related Work . . . 8 3 Method 9 3.1 Measuring the change in concentration with ultrasound . . . 9
3.2 Measuring the concentration by hand . . . 9
3.3 Evaluation . . . 10
4 Results 13 4.1 Pre-study . . . 13
4.2 Initial Tests . . . 14
4.3 The Equipment . . . 17
4.4 The signal processing . . . 17
4.5 Evaluation . . . 17
5 Discussion 21 5.1 Results . . . 21
5.2 Limitations . . . 22
5.3 The work in a wider context . . . 23
5.4 Source criticism . . . 23
6.1 Research questions and conclusions . . . 25 6.2 Future work . . . 26
Bibliography 27
List of Figures
2.1 The different phenomenons that decreases the intensity of a sound wave when it
hits a particle. . . 4
2.2 Two frequencies that line up if the sampling rate is 100 Hz. . . 6
2.3 A signal with 25.5 Hz frequency sampled at 100 Hz for 1 second. . . 6
2.4 A signal with 25.5 Hz frequency sampled at 100 Hz for 2 seconds. . . 7
2.5 A signal with 25.5 Hz frequency sampled at 100 Hz for 1 second but with zero padding. . . 7
3.1 An abstract view of the experimental setup for the ultrasound method. . . 10
4.1 The signal strength is too strong to distinguish between signals. . . 14
4.2 The effect the temperature of the water had on the measured energy level. . . 15
4.3 Measurements taken with increasingly higher concentrations of dissolved toilet paper. . . 16
4.4 Comparing the accuracy of stirring during or between measurements with a con-stant concentration. . . 16
4.5 The equipment used during testing (left) and how it was set up (right). . . 17
4.6 The different stages the signal goes through while being processed. . . 18
4.7 Measurements taken with increasingly higher concentrations until measurement 5 and 6, where the concentration is decreased for each measurement. . . 19
List of Tables
4.1 The total time taken to measure by hand and the measured concentration. . . 20 4.2 Time taken to dry each sample by hand. . . 20 4.3 The total time taken to measure with ultrasound. . . 20
1
Introduction
This is a thesis done at Linköping University. The thesis has been done for the company Deepoid AB1. Deepoid AB develops technical equipment for underwater use.
1.1
Motivation
The earth’s surface consists of 70 % water. However, only 3 % of that water is freshwater and only 0.01 % is accessible for humans to use and much of that 0.01 % is polluted according to Azizullah Azizullah et al. [1]. Goel [2] writes that water can be considered polluted when the characteristics of the water have a negative effect on aquatic life or people who drink the water. Boyd and Tucker [3] explain that water has different characteristics such as phys-ical, chemical and biological. Physical characteristics of water are, for example, temperature, color, taste and odor. Chemical characteristics of water are acidity, alkalinity, hardness and corrosiveness. Biological characteristics of water refer to all the living organisms that reside in the water. The water quality is determined by the combination of all these characteristics and how they influence the beneficial use of the water. Bilotta and Brazier [4] state that it has been established that suspended particles are an important aspect of water quality and its effect on aquatic life has been a worldwide interest in the past 50 years. They explain that suspended particles are defined as particles that are held suspended in a water column by turbulence.
Knowing the concentration of suspended particles in water is relevant in many areas. Wastewater treatment plants and paper mills are examples of industries that need to know the concentration of suspended particles in their water. Wastewater treatment plants want to make sure that the water is not polluted and can be returned to the water cycle. Paper mills need different concentrations of raw material mixed with the water for making different types of paper.
The standard method used when measuring the concentration of suspended particles in all these different areas is filtering by hand. However, this method is very slow and expen-sive, since it requires manual labour. Therefore, alternative methods that are cheaper and faster have been developed. However, these methods can only estimate the concentration.
1. INTRODUCTION
Deepoid AB wants to investigate if ultrasound could be used to measure the concentration of suspended particles in water in all these different areas.
1.2
Background
Deepoid AB develops technical equipment for underwater use. Mostly sensors and commu-nication devices that use ultrasound. They also help other companies develop projects in the same area. They want to develop a method that measures the concentration of suspended particles in water, since they already have software and hardware for sending and receiving ultrasound signals. This method is intended to be used in industries like paper mills and wastewater treatments plants. They also plan on making it available to the public so that it can be used by citizen scientists. A citizen scientist is defined as a volunteer who does not necessarily have any expertise in the area but helps collect and process data.
1.3
Aim
Deepoid AB wants to develop a method that can measure the concentration of suspended particles in water with both large and small particles as well as high and low concentrations. Before they start developing a method, they first want to know if it is possible. Therefore, the aim of this thesis is to provide a proof of concept to Deepoid AB that ultrasound can be used to measure the changes in the concentration of suspended particles in water. We also aim to compare the time it takes to measure the concentration of particles in water with ultrasound against measuring it by hand.
1.4
Research questions
These are the questions this thesis project aims to answer:
1. Is it possible to measure if the concentration of suspended particles in a volume of water has increased or decreased, using ultrasound?
2. When measuring a change in the concentration of suspended particles in water, is the method developed in this thesis faster than measuring it by hand?
1.5
Delimitations
The choice of sensors used to transmit and receive ultrasound is not evaluated.
2
Theory
In this chapter, some standards for measuring the concentration of suspended particles in wa-ter by hand are presented and described. Alwa-ternative methods for estimating concentration are also introduced. Thereafter the basics of ultrasound and signal processing are explained. Furthermore, there is a section that briefly explains a piezoelectric transducer and lastly, a section that presents related works.
2.1
Standards for measuring the concentration of suspended particles in
water
There are several standards for measuring the concentration of suspended particles in water. The paper mills use standards like ISO 4119:1995, while wastewater treatment plants use standards like EPA-NERL: 160.2, 2540 D or D5907. Standards like these, all use the same basic steps:
1. Take a sample of the water to be measured 2. Weigh the filter
3. Pour the sample through the filter
4. Dry the filter with the sample in a drying oven 5. Weigh the filter and sample
6. Redo step 4 and 5 until the filter and sample have achieved a constant weight 7. Subtract the filter’s weight from the final weight
The different standards differ on how to take the sample from the source, what pore-size to use for the filter, what precision the scale needs to have, the minimum amount of time the filter and sample need to dry before weighing and how to define constant weight.
2. THEORY
2.2
Alternative methods for estimating the concentration of suspended
particles in water
Davies-Colley and Smith [5] explain that since measuring by hand can be time-consuming, many industries that need to measure the concentration of suspended particles have opted into using turbidity. Turbidity measures the amount of light scattered by particles. The more light that is scattered, the more particles are present.
The microwave method, as the name implies, uses microwaves to measure the concentra-tion of particles in water. This method is suitable in industries like paper mills, where they have a large range of concentrations. Nakayama [6] developed a method that could mea-sure concentration ranging from 0.6-10 %. He also mentions another method called the fan method. The fan method measures the torque of a rotating fan in the water. The more torque measured, the higher the concentration.
All of these methods are relative measurements. Which means that a base value is needed to determine the particle concentration.
2.3
Ultrasound
Sound is created when a body vibrates. When that body vibrates, sound waves are prop-agated out from the body with the same frequency as the body. Ultrasound waves behave the same as normal sound waves, except that ultrasound waves have higher frequencies. The cut-off point between normal sound and ultrasound is the point where most people are unable to hear the sound, which is approximately 20 kHz. Povey [7] explains that when ultrasound travels through a medium, the intensity of the wave is decreased with distance travelled. In other words, the energy gets spread out. The amplitude of an ultrasound wave correlates with the amount of energy that the wave carries. The higher the wave energy, the higher the amplitude. The intensity of a wave represents the average energy per area unit in a specified direction. In idealized materials, the intensity would only decrease because of the wave propagation. However, in natural materials, when an ultrasound wave collides with particles several effects will occur that decrease the intensity even more. Shuo et al. [8] explain that these phenomenons are reflection, scattering and diffraction. If the ultrasound wave has a wavelength lower than the particle diameter but fairly close, the ultrasound wave will get scattered in all directions. The portion of the scattered wave that propagates back to the sound source is referred to as the backscattering coefficient. With a wavelength much lower than the particle diameter, reflection will occur instead. When the wavelength is much bigger than the particle diameter, diffraction will occur. Diffraction is when an ultrasonic wave swerves around the particle and propagate further. All of these phenomenons will result in energy loss and is referred to as attenuation. Figure 2.1 illustrates these different phenomenons. Another phenomenon that also occurs when an ultrasound wave collides with an obstacle is that harmonic waves are created. A harmonic wave is defined as a wave that has a frequency that is a multiple of the original wave frequency.
(a) scattering (b) diffraction (c) reflection
Figure 2.1: The different phenomenons that decreases the intensity of a sound wave when it hits a particle.
2.4. Piezoelectric Transducers
2.4
Piezoelectric Transducers
Piezoelectric transducers are sensors that can convert mechanical waves into electric signals and vice versa. This means that they can convert an electric signal into a sound wave. That way the sensor can be used as a transmitter. When a sound wave hits the sensor, it can convert the sound wave into an electric signal. This way the sensor can be used as a receiver.
2.5
Signal Processing
Tan and Jiang [9] writes that the Discrete Fourier Transform (DFT) is an algorithm that trans-forms a signal taken over time into its frequency components and can be achieved by the Equation 2.1 below. X(k) = N´1 ÿ n=0 x(n)e´i2πknN , k=0, 1, ..., N ´ 1 (2.1)
Where X(k) is the spectral sample of the frequency k and x(n) is the input signal of the n:th sample. There are several versions of a function called the Fast Fourier Transform (FFT). The FFT produces identical results as the DFT but is more efficient and therefore much faster to calculate. Matlab’s version1of the FFT is defined in Equation 2.2.
X(k) = N´1 ÿ n=0 x(n)WN(n´1)(k´1) Where WN=e´i 2πk N (2.2)
One application of the FFT function is to transform a digital signal into a frequency spectrum. The frequency spectrum shows which frequencies are present in the signal and the amplitude of those frequencies. The highest frequency that can be measured is always half the sampling rate. For example, if the sampling rate is 100 Hz and the frequency transmitted is 75 Hz, the frequency spectrum will still show that the dominant frequency is 25 Hz. The reason for this is that the cycles of the frequency at 75 Hz line up with the frequency of 25 Hz every time we take a sample. A sampling of 100 Hz means 1 sample is taken every 0.01 seconds. Figure 2.2 shows that when a sample is taken at 0.01 seconds, 0.02 seconds and so forth, both frequencies are either at 1, -1 or 0. The result of this is that both frequencies appear to be oscillating at 25 Hz. All frequencies that are higher than half the sampling rate will line up with a lower frequency. Therefore, the highest frequency we can measure is half the frequency of the sampling rate.
The frequency resolution is the difference in frequency of each X(k). These are called bins and do not always correspond to the measured frequency. If a signal with an amplitude of 50 and a frequency of 25.5 Hz is sampled at a rate of 100 Hz for 1 second. The closest bins to 25.5 Hz will be 25 Hz and 26 Hz. Therefore, the highest amplitude will be on 25 Hz and 26 Hz but their amplitude will be lower than the actual signal’s amplitude of 50, as shown in Figure 2.3. The only thing that affects the frequency resolution is the sampling time. Therefore, if the sampling time is increased to 2 seconds, as in Figure 2.4, the frequency resolution is doubled and the difference between each bin will be 0.5 Hz. This allows the frequency of 25.5 Hz to fall exactly on a bin and the amplitude can be perfectly estimated. However, it is not always an option to sample for a long time. In those cases, zeroes can be added to the end of the sampled data. This is called zero-padding and does not increase the frequency resolution but it makes the interpolation of the amplitude better. In Figure 2.5, zeroes have been added to make the difference between the bins 0.5 Hz, which allows the amplitude to be correctly
2. THEORY
Figure 2.2: Two frequencies that line up if the sampling rate is 100 Hz.
Figure 2.3: A signal with 25.5 Hz frequency sampled at 100 Hz for 1 second.
2.5. Signal Processing
Figure 2.4: A signal with 25.5 Hz frequency sampled at 100 Hz for 2 seconds.
Figure 2.5: A signal with 25.5 Hz frequency sampled at 100 Hz for 1 second but with zero padding.
2. THEORY
Signals which are not periodic have spectral leakage which means that the frequencies "leaks" to adjacent frequencies. The result of spectral leakage is that the amplitude estimation is less accurate and if there are several frequencies close together it will also make it harder to distinguish between them. This error can be reduced by a longer sampling time or by apply-ing a window function. However, window functions have the drawback that they decrease the frequency resolution, which means that the amplitude estimation accuracy gets worse.
A signal often contains noise. Noise is part of the sampled signal that is not of interest but can interfere with the results. What counts as noise differs depending on what is measured. A biologist who studies whale sounds might consider radar signals from a submarine as noise, while the submarine might consider the sound from the whales as noise. Filters can be applied to remove the noise from a sampled signal. Low-pass filters block frequencies over a given cut-off frequency. High-pass filters block frequencies below a given cut-off frequency. Band-pass filters block frequencies outside two given cut-off frequencies.
2.6
Related Work
Shen and Lemmin [10] present their new method on how to profile suspended particles in water. They emphasize that their backscattering profile can calculate the attenuation with-out previous information on both particle size and size distribution. The fact that they can calculate without prior information differs from previous works. Their method involved em-ploying two ultrasonic transducers across each other and profiling both the backscattering coefficient and the diffraction coefficient. One transducer would be enough for measuring backscattering but since they also measure the diffraction two opposing transducers are nec-essary. Sung et al. [11] also employed two opposing transducers to measure the attenuation of the sound wave to calculate the concentration of suspended particles. They found that the attenuation was mainly determined by the concentration of the suspended particles and the temperature of the water.
Spelt et al. [12] developed a theory for the attenuation of sound caused by particles with small volumes. Their theory was very accurate and predicted the same result as their exper-imental measurements. They also show that the particle size distribution can be determined from the attenuation-frequency data. However, they also state that determining the particle-size distribution is very sensitive to the choice of frequency range and the physical properties of the particles.
In a field study, Siadatmousavi et al. [13] compared the accuracy between light and ultra-sound when estimating the concentration of suspended particles. They compared an Optical Backscatter Sensor (OBS) with a Pulse-coherent Acoustic Doppler Profiler (PCADP). They found that estimating the concentration with ultrasound is less accurate than light when try-ing to detect very small particles. However, increastry-ing the sampltry-ing time to a few hours mitigated the difference in accuracy.
Sahin et al. [14] used a 10-MHz acoustic Doppler velocimeter (ADV) when estimating the concentration of suspended particles. They discovered that they had an upper and lower bound for the concentration they could measure. If the concentration was too high, the backscatter became saturated. Meaning that, after a certain point, adding more suspended particles does not further increase the measurement. When the concentration was too low, there was not enough backscatter to detect the suspended particles.
1https://www.mathworks.com/help/matlab/ref/fft.html#buuutyt-6
3
Method
This chapter begins describing the method that was used to measure the changes in particle concentration in water. Secondly, how the particle concentration was measured by hand is presented. Lastly, the method to evaluate both research questions is described.
3.1
Measuring the change in concentration with ultrasound
A concentration of suspended particles in a container filled with water was estimated by measuring the amplitude of an ultrasound signal. While the signal was transmitting, the wa-ter in the container was stirred to ensure that the particles were suspended and as evenly distributed as possible. The ultrasound signal was transmitted from one side of the con-tainer and received on the opposite side as can be seen in Figure 3.1. The signal was sent several times during one measurement. To transmit and receive the ultrasound signals, two piezoelectric transducers were used. The raw data from the receiver were processed to de-termine the amplitude of the signal. To do this, the signals were identified and cut out from the raw data. The signals were then zero-padded to the next power of two and passed into an FFT function to get a frequency spectrum. The mean of the amplitude in the frequency spectrum of all signals was calculated as well as the standard deviation. Each signal which deviated more than a certain threshold from the mean were removed. The new mean was then calculated with the new set of signals. The amplitude of the current signal could then be compared to the amplitude of the previous signals with different concentrations of particles to determine the change in concentration.
3.2
Measuring the concentration by hand
The concentration of suspended particles in a container filled with water was measured by scooping up a sample, filtering out the particles and weighing the particles. The concen-tration of the suspended particles in the container was known throughout the experiment to enable a comparison to the measured concentration. In order for the concentration to be known, the volume of the water in the container was known and the particles were weighed before they were added to the container. After the particles were added to the container, the water was mixed to ensure an evenly distributed concentration of particles. A sample was
3. METHOD
Figure 3.1: An abstract view of the experimental setup for the ultrasound method.
then taken and the volume of the sample recorded. The sample was then filtered through a filter with a known weight and a pore-size small enough to catch all particles. The filter and the particles it contained were then dried in an oven until they were dry to the touch. The filter was then weighed and the weight recorded. The filter was then dried again for one-quarter of the total drying time and reweighed. If the two weighings differed more than 0.01 g, the filter was once again dried for one-quarter of the total drying time and reweighed. This process repeated until two consecutive weighings were achieved which did not differ more than 0.01 g. The concentration was then calculated by using the Equation 3.1.
g/L= m1´m2
V (3.1)
Where m1is the weight of the filter, in grams. m2is the weight of the dried filter containing the particles from the sample, in grams. V is the volume of the sample, in liters.
3.3
Evaluation
To answer the research questions, the method using ultrasound was tested and compared to the method of filtering by hand. The first research question was answered by determining if the amplitude of the ultrasound signal was related to the concentration of particles in the water. The amplitude should decrease when the concentration increases and increase when the concentration decreases. Firstly, a measurement with clear water was taken. This was done to have a value to compare the first concentration to. Secondly, several measurements with increasingly higher concentrations were taken. Thirdly, several measurements were taken after filtering out particles to lower the concentration. The concentrations during this evaluation were unknown since it was only designed to evaluate if the amplitude of the signal is related to the concentration and not their exact relationship.
The second research question was answered by comparing the time taken for both meth-ods to measure a known concentration of suspended particles. The time started when the sample was taken and ended when the measurement was known. The test was carried out in a container with 15 liters of water and each measurement increased in concentration by 0.05 g/L, starting with 0 g/L. The concentration of 0 g/L was only measured using the ul-trasound method and was done to have a value to compare the first measured concentration against. Firstly, particles were added to the water to achieve an increase in concentration by 0.05 g/L. Secondly, a measurement was taken with the method using ultrasound while being timed. Thirdly, a sample was taken out of the container while being timed. Fourthly, the
3.3. Evaluation
water volume of the container affect the results of the ultrasound method. To prevent this, the same volume of water that was removed by taking a sample was added to the container. Since water was added, the concentration was diluted. Therefore, particles weighing as much as the collected sample should contain were also added to the container. This increased the concentration to what it was before the sample was taken. This process was repeated several times. The samples taken were then dried and weighed and the time taken recorded. The mean time taken for the method of measuring by hand was then calculated and compared to the time taken to measure by using ultrasound.
4
Results
This chapter begins with the initial decisions that were made when researching. Secondly, all the results from the initial tests are shown to give an insight into why every decision was made while developing the method. Thirdly, how the received signal was processed is described. Fourthly, all the equipment that was used is presented. Lastly, a section that evaluates both research questions.
4.1
Pre-study
Many different methods were found that used ultrasound to estimate particle concentration in water and all of them measured the backscatter. However, measuring the backscatter in-volved advanced mathematics, and given the time frame, the decision was made to skip using backscatter. Instead, the decision was made to measure the direct signal. Measuring the direct signal means that the strength of the ultrasound that reaches the receiver on the other side is measured. The more particles there are in the water, the more of the ultrasound wave is reflected and diffracted before reaching the receiver.
Afterwards, how the ultrasound waves behave when colliding with particles as well as how to process the received signal was researched. Based on the research, a decision was made to compare a received signal to the previous signal. If the intensity of the received signal that has passed through a certain concentration is lower than the previous signal, it means that the concentration is higher than before. If the received signal is stronger than the previous signal, the concentration is lower than before.
After getting a good understanding of all the theory needed to develop the method. The software and hardware that the company had provided was studied. They provided diving computers that had integrated sensors that were able to send and receive ultrasound signals. The diving computers transmitted the ultrasound signal at 40 kHz and sampled the signal at 96 kHz. In addition, scripts to extract the raw data, update the source code and erase the recording on the diving computers were provided. These scripts were continuously used while testing.
The concentration is measured with the same volume of water as well as the same dis-tance and angle between the transmitter and receiver. This is done to reduce the impact of background noise on the measurements since it will be constant.
4. RESULTS
To mitigate the problem of spectral leakage and achieve a high frequency resolution, a long signal was implemented. However, the diving computers were made for communica-tion underwater and not estimating particle concentracommunica-tion. Therefore, the sensors could not produce a continuous ultrasound signal for longer than 10 seconds because of the software in the diving computers.
4.2
Initial Tests
If the diving computers were placed too close together the received ultrasound signal was too strong, as can be seen in Figure 4.1. The received signal was normalized to values between 1 and -1 when extracted. This led to the signal strength in the extracted data having the max value, which meant that distinguishing between two signals with different amplitudes was impossible. If one signal was twice as strong as another, it would still give the same amplitude value. The diving computers were therefore placed diagonally in the container to increase the distance between them. In addition, it was easy to accidentally move them during the tests. Therefore, Deepoid AB 3D-printed two stands for the diving computers. The stands were glued to a piece of wood that was placed in the container used for testing. They were not glued directly to the container to prevent damage to the container. The piece of wood was the same width as the container to prevent it from moving. Two weights of 1 kg were also placed on the piece of wood to further prevent it from moving.
Figure 4.1: The signal strength is too strong to distinguish between signals.
Only one test per day could be conducted and it could not last more than 45 minutes. This was because of three reasons. Firstly, the temperature of the water affected the measurements. Warmer water resulted in weaker signals, as seen in Figure 4.2. Therefore, the water in the container needed to be room temperature when conducting the tests. This meant that filling the container with water the day before running a test was needed to make sure that the temperature of the water was stable. Secondly, the diving computers had the limitation that they could only store data recorded for approximately 45 minutes. Lastly, even with the stands for the diving computers, it was not possible to ensure that they were placed the same in two separate tests. Meaning that, if the diving computers were to be picked up from the container and put back in, the result were not guaranteed to be the same even if the concentration did not change.
4.2. Initial Tests
Figure 4.2: The effect the temperature of the water had on the measured energy level.
Several steps were taken to ensure that the difference in the received signal was because of a different concentration of particles and not a measuring error. The first step was to send 20 signals, each lasting 10 seconds, to take one measurement. This was done so that a mean amplitude could be calculated. The second step taken to reduce the measuring error was to remove signals which deviated too much from the mean. Matlab’s function rmoutliers()1 was used to remove signals which differed more than three standard deviations from the mean. The threshold value of three was the default value used by the function. The mean was then recalculated without the removed signals.
Dissolved toilet paper was used as a substitute for suspended particles since access to fine particular matter was unavailable. The first test with a concentration of particles was done by stirring the water to evenly distribute the particles and then letting them settle before taking a measurement. The exact concentration of each measurement is unknown. However, particles were added to the water after each measurement to always increase the concentration for the next measurement. The result can be seen in Figure 4.3 and shows a promising trend but it was very inaccurate. Measurements 2, 4 and 8 has higher amplitude than their previous measurement, even though the concentration is higher. The reason was that the particles were not guaranteed to settle evenly distributed. If more particles settled in the centre of the container the signal would be weaker than if more particles settled at the edges of the container. Additionally, the toilet paper did not dissolve into very small pieces, which could also contribute to the inaccuracy. Therefore, the toilet paper was ground up with an electric mixer to create finer particles. To see if slow-moving water would more evenly distribute the particles, a test was conducted where the water was slowly stirred with two spoons when the signal was transmitted. The test was done with a small unknown concentration that stayed constant through all measurements. This was done to see the measuring error caused by the method and hardware. The results in Figure 4.4 shows that it is indeed more accurate to stir during measurements than it is to stir between the measurements and let the particles settle. The standard deviation when stirring between measurements was 1.14 mV and the difference between the maximum and minimum values were 3.37 mV. The standard deviation when stirring during measurements was 0.26 mV and the difference between the maximum and minimum values were 0.69 mV.
4. RESULTS
Figure 4.3: Measurements taken with increasingly higher concentrations of dissolved toilet paper.
Figure 4.4: Comparing the accuracy of stirring during or between measurements with a con-stant concentration.
4.3. The Equipment
4.3
The Equipment
The equipment used when testing the ultrasound method is listed below. Figure 4.5 shows each item on the list as well as how the experiment was set up.
1. Two 3D printed stands glued to a thin piece of wood to stabilize the diving computers during the experiments.
2. A plastic container with the dimensions 36.5x56.5x40 cm that the experiments were carried out in.
3. A digital scale with precision of 0.01 g to weigh the filter before and after filtering the samples.
4. 2 Weights of 1 kg that were placed on the piece of wood to prevent it from moving. 5. Two piezoelectric sensors of the model KS-A1640H10.5S integrated into two separate
diving computers. The sensors were used to send and receive the ultrasonic signals. 6. An electric mixer that was used to stir the samples to ensure small particle sizes.
Figure 4.5: The equipment used during testing (left) and how it was set up (right).
4.4
The signal processing
The signal processing was done in Matlab 2020a. The procedure was divided into five main steps. Firstly, the raw data from the receiver was extracted using the script provided by Deep-oid AB. Secondly, each 10 second signal in the raw data were cut out and saved individually. Thirdly, those signals were passed to another script that zero-padded the signals and per-formed the FFT function on them. The script then returned the sum of the amplitude value of all frequencies in the range of 1 Hz-48000 Hz. Fourthly, another script then calculates the average as well as the standard deviation for each measurement and removes signals if neces-sary. Fifthly, the average of each concentration is plotted and presented on a graph. Figure 4.6 shows how the raw data looks as well as the cut-out signal before and after the FFT function is applied.
4.5
Evaluation
Two different tests were carried out to answer the research questions. The first test confirmed that a change in the concentration of particles in water can be measured with ultrasound. The
4. RESULTS
(a) The raw data of ten signals. (b) The raw data of one signal.
(c) The frequency spectrum of one signal.
Figure 4.6: The different stages the signal goes through while being processed.
Answering the first research question
To evaluate the ultrasound method, validation that the signal strength was related to the con-centration of particles in the water was needed. The container was filled with 15 litres of water the day before. A measurement of the clear water was then taken. Particles with an unknown weight were added, stirred with an electric mixer until it was ground up into small particles and then measured again. More particles were added and the new concentration measured five more times. Particles were then filtered out and the new concentration mea-sured. Particles were filtered out from the water and the new concentration measured a total of two times. A measurement consisted of 20 signals which were transmitted with an inter-val of 10.5 seconds, each signal lasting 10 seconds. The water was constantly stirred with two spoons. The results is shown in Figure 4.7.
Answering the second research question
To evaluate if the method with ultrasound is faster than measuring by hand, a comparison was done to the standard method ISO 4119:1995. The reason this method was chosen is that it is used in paper mills. The particles measured in paper mills are mostly fibres, which toilet paper is as well. Therefore, ISO 4119:1995 should be an appropriate standard to compare with. The standard follows the steps described in Section 2.1. However, we did not have access to a drying oven, so a regular kitchen oven was used instead. Constant weight is defined as two consecutive weightings that do not differ more than 0.01 g. The standard also
4.5. Evaluation
Figure 4.7: Measurements taken with increasingly higher concentrations until measurement 5 and 6, where the concentration is decreased for each measurement.
states that each time the sample is weighed and constant weight is not achieved, it has to be dried again for at least one-fourth of the total drying time so far.
The container was filled with 15 litres of water the day before the experiment was carried out. A 15-litre bucket was also filled with water the day before. A measurement of the clear water was then taken with the ultrasound method. Toilet paper weighing 7.5 grams, which is a concentration of 0.05 g/L, was added. The toilet paper was stirred with an electric mixer until it was ground up into small particles and then measured with the ultrasound method. A 1-litre sample from the container was then taken and stored in a glass container. To not decrease the water level, which could affect further measurements, 1 litre of water was added to the container from the bucket. To make sure the concentration did not change, toilet paper weighing as much as the amount of paper that the sample taken should contain was added to the container. Five measurements were taken with both methods with each concentration increasing by 0.05 g/L. The samples in the glass containers were then filtered through a coffee filter and dried in an oven at 105˝ C. For the initial drying time, the samples were checked every 30 minutes. When the sample was dry to the touch, it was taken out and weighed. After being weighed, the samples were put back into the oven for roughly one-quarter of the total drying time until two consecutive measurements were achieved. The time it took to sample, filter, dry and weigh the samples were recorded and are shown Table 4.1. Table 4.2 shows the drying time for each individual concentration. The time it took to measure the concentration by hand increased as the concentration increased. The average time for measuring by hand was 4.5 hours compared to measuring with ultrasound which only took 8.5 minutes independent of the concentration measured. This means that our method is, on average, 31.7 times faster than measuring by hand. The change in the concentration measured by the ultrasound method can be seen in Figure 4.8.
4. RESULTS Concentration (g/L) Filtering time (m) Drying time (m)
Total time (m) Measured concentration (g/L) 0.05 45 95 140 0.04 0.10 60 145 195 0.10 0.15 60 145 195 0.11 0.20 90 190 280 0.19 0.25 90 425 515 0.24
Table 4.1: The total time taken to measure by hand and the measured concentration. Concentration (g/L) Initial drying time (m) Second dry-ing time(m) Third drying time time (m) Fourth drying time (m) 0.05 60 15 20 -0.10 90 25 30 -0.15 90 25 30 -0.20 120 30 40 -0.25 210 60 70 85
Table 4.2: Time taken to dry each sample by hand. Concentration (g/L) Measuring time (m) Extracting and Process-ing time(m) Total time (m) 0.05 3.5 5 8.5 0.10 3.5 5 8.5 0.15 3.5 5 8.5 0.20 3.5 5 8.5 0.25 3.5 5 8.5
Table 4.3: The total time taken to measure with ultrasound.
Figure 4.8: The change in concentration measured by the ultrasound method.
5
Discussion
In this chapter, the results are analyzed and criticized. Thereafter the limitations of the thesis and the work in a wider context are discussed. Lastly, there is a section where the sources are criticized.
5.1
Results
Based on the theory of ultrasound, the expectation was that an energy loss would occur when the ultrasound wave hits the particles. The energy loss could then be measured by analyzing the spectrogram of the received signal. However, we were unsure if we would be able to see the difference in the concentration of particles or if the energy loss would be too small. Another uncertainty was if the relationship between the energy loss and concentration of particles would be linear or non-linear. If it was linear it would be easier to map the energy loss to a certain concentration than if it was non-linear.
The results show that while increasing the concentration of particles the amplitude does indeed decrease. However, after a certain concentration, the amplitude tended to stay the same or even increasing. We believe this is because the relationship between the signal strength and the concentration of particles is non-linear. The higher the concentration the smaller the difference in signal strength. This makes the measurement very unreliable when our measuring errors become larger than the difference caused by the increase in concentra-tion. For example, the difference between 0.15 g/L and 0.20 g/L in Figure 4.8 is less 0.69 mV, which is the range of error in Figure 4.4 when stirring during the measurement. So, if we did not know the concentration, we would not know if the amplitude stayed the same because the concentration did not change or if it is a measuring error.
In our setup, we could not reliably measure concentrations above 0.15 g/L. This is most likely because our transmitter’s output signal is too low, so increasing the output might make it easier to measure higher concentrations. However, that could instead cause problems when measuring lower concentrations, just as shown in Figure 4.1, where the received signal is maxed out. A range of different transmitters with different output strengths might be needed to measure different ranges of concentration.
We have only tested our method with dissolved toilet paper as suspended particles. Other materials might affect the attenuation differently and therefore yield different results.
5. DISCUSSION
Getting the particles to be uniformly distributed also proved to be an important aspect. As shown in Figure 4.4, the standard deviation decreased by a factor of five when stirring the sample while measuring. Having a low standard deviation means that every measurement does not differ as much, which in turn means that the accuracy is better. We believe the cause of this improvement is the fact that the particles are more uniformly distributed when the sample is stirred.
Measuring the concentration by hand took on average 4 hours and 30 minutes which was much longer than anticipated. Therefore, we discussed with the sewage treatment division at Tekniska verken1 regarding how long it takes for them to measure the concentration of suspended particles by hand. They state that it takes roughly 5-10 minutes to filter a sample, 2 hours to dry it and a few minutes to weigh it. So, the average measurement time of 4.5 hours is not a representative time. However, the fastest of our measurements of 2 hours and 20 minutes seems to be an accurate estimate of the time it takes to measure the concentration by hand. This means that measuring by hand is not 31.7 times slower than the ultrasound method, but rather 16.4 times slower. However, the accuracy achieved when measuring by hand was much better than when measured by ultrasound. Only one measurement was off by more than 0.01 g/L when measuring by hand, as seen in Table 4.1. When measuring with ultrasound, on the other hand, you can only see that the concentration has changed but it is impossible to see by how much, as seen in Figure 4.8.
When measuring the concentration by hand, higher concentrations took longer to filter and dry than lower concentrations. This was because of two reasons. Firstly, the filter size used could not contain the full sample, so the sample had to be filtered in portions. Therefore, the later portions that were filtered through both the particles already in the filter and the filter itself. Higher concentrations have more particles, which meant that the later portions of those samples needed to be filtered through more particles and thus took longer time. Secondly, higher concentrations also took longer to dry. They contain more particles so more water needs to evaporate. Furthermore, the water might be trapped in the centre of a pile of particles and has to travel longer to be able to evaporate. The process of filtering and drying could most likely have been sped up by using several filters and thus eliminate the extra time caused by higher concentrations. Using standard lab equipment might also have sped up the process.
When measuring the concentration by hand, all moisture was evaporated from the toilet paper. However, when weighing the toilet paper before adding it to the water, it was not dry and most likely contained some moisture. This means that the concentration was probably slightly lower than we believed and might explain why some of the measurements were slightly lower than the expected results.
5.2
Limitations
The temperature’s effect on the signal strength affected the results quite a lot, which limited us to only testing once per day. Sound travels faster in warmer water so we thought the signal would be stronger in warmer water. However, the signal strength decreased as the temperature increased, as shown in Figure 4.2. Since the temperature will vary in a real-time environment, the temperature needs to be measured and taken into consideration if our method was to be implemented. Because of this, the validity of our method is not very high since we were not able to verify any results. For example, when discovering that the accuracy got much better when stirring during the measurements, only one test was done to verify. If we had time to verify with more tests it would have been more reassuring that the results were accurate.
1https://www.tekniskaverken.se/
5.3. The work in a wider context
5.3
The work in a wider context
If a product is made with citizen science in mind, people will use the product outside of industries and in rivers, lakes and oceans. Transmitting ultrasound signals can disturb and even be harmful to aquatic life. Mann et al. [15] write that ultrasound is used to keep fish from entering certain waters but it is unclear if they can hear the ultrasound frequencies or if they hear lower-frequency side-bands. However, even though most fish have a hearing range of 2-4 kHz, they show that at least the American shad can hear up to 180 kHz, which lies in the ultrasound spectrum. Furthermore, Frenkel et al. [16] showed that ultrasound also can cause damage to fishes’ skin when exposed to a 1 MHz frequency at a distance of 15 cm from the source. Depending on the intensity of the signal, only 5 seconds of exposure was enough to cause damage. The signals transmitted in our experiments are quite long but very weak. However, the effects it can have on aquatic life should be studied more closely before a potential product is released for the general public.
Another problem related to citizen science is littering. The product can never be made biodegradable, so if it is left in the water it will stay there for a long time. The impact of this problem will increase as the popularity of the product increases. The more people that use it, the more people will accidentally lose their product. Furthermore, the product will most likely be more popular if it is small, light and easy to carry around, which also means that it is easier to lose.
5.4
Source criticism
Every source used in this thesis is either a book or an article. Every article is peer-reviewed which should mean that they are reliable. We do not know if the books are peer-reviewed but one of them is published by Elsevier, which should be trustworthy. We do not recognize the publishers of the other books, therefore, they might not be reliable. However, even if they are peer-reviewed the information needs to be chosen carefully. For example, when information is passed on multiple times between people, misinterpretations can occur. All it takes is one misinterpretation of information and thereafter everyone that reads the information is misinformed.
6
Conclusion
This chapter shortly summarizes the aim and research question of the thesis as well as the conclusions to them. It also mentions some interesting topics for future work.
6.1
Research questions and conclusions
This thesis was made because Deeopid AB wanted to investigate whether a method for mea-suring the concentration of particles in water using ultrasound was feasible. The first research question of this thesis was to investigate if it was possible to measure changes in the concen-tration of suspended particles in a volume of water using ultrasound. The second research question was to find out if the method developed in this thesis is faster than measuring the concentration of particles by hand. A method was developed that used two opposing piezo-electric transducers to transmit and receive ultrasound signals. The method transmitted the ultrasound signal at 40 kHz and sampled at 96 kHz. The signal was transmitted for 10 sec-onds 20 times, to take one measurement. The signals were passed to an FFT function which transformed the signal to its frequency domain. The mean of all signals’ amplitude were cal-culated and any signal which deviated from the mean with more than a factor of three of the standard deviation were removed. The mean of all signals’ amplitude were compared with a previous measurement to measure the change in concentration from that signal. Two evalu-ations were made to answer the research questions of this thesis. The first test measured the amplitude of the signal after both increasing the concentration and decreasing it. The second test compared the time it took to measure a known concentration with the method developed in this thesis with measuring a known concentration by hand.
The conclusion of the first test was that it is possible to measure if the concentration of suspended particles in water has increased or decreased. This conclusion was drawn because the signal strength was related to the concentration since the signal strength decreased when the concentration increased and vice versa. Figure 4.7 displays the result from that test. The second test allowed us to conclude that the method developed in this thesis is roughly 16.4 times faster than measuring it by hand. However, the relationship between the signal strength and the concentration was non-linear, as seen in Figure 4.8. Which meant that the method developed in this thesis could not determine how much the concentration had changed, only that it had increased, decreased or stayed the same. Furthermore, if the concentration was too
6. CONCLUSION
high, the measuring error caused by the method and hardware was larger than the change in amplitude caused by the change in concentration.
6.2
Future work
There are several topics which would be interesting to investigate for future works. The ratio between the different phenomenons were not captured in this thesis. This thesis only cap-tured how much energy loss the phenomenons caused in total. Therefore a topic for future work could be to investigate if determining the energy loss caused individually by scattering, reflection and diffraction could improve the method developed in this thesis. By using the wave frequency and the particle diameter, the ratio of the different phenomenons when a ul-trasound wave hits a particle could be found. Thereafter the material properties would need to be studied since different materials can cause different energy losses. With the attenuation data for the material the energy loss for each phenomenon could be determined.
In this thesis backscatter was overlooked because of the complexity. However, another interesting topic would be to compare the results when measuring the concentration of par-ticles with a direct signal against measuring it with backscatter, to see which one is more accurate.
One could also increase the signal strength to investigate whether the method developed in this thesis could measure higher concentrations.
Testing could also have been done with different materials, different sizes, higher concen-trations and lower concenconcen-trations. Based on these tests, an algorithm could be developed that estimates the concentration of particles in the water based on the attenuation with type of material, particle diameter and frequency of the ultrasound wave as parameters.
Bibliography
[1] Azizullah Azizullah, Muhammad Nasir Khan Khattak, Peter Richter, and Donat-Peter Häder. “Water pollution in Pakistan and its impact on public health — A review”. In: Environment International 37 (2011), pp. 479–497.DOI: https://doi.org/10.1016/ j.envint.2010.10.007.
[2] P.K. Goel. Water Pollution: Causes, Effects and Control. New Age International, 2009.ISBN: 9788122418392.
[3] Claude E. Boyd and Craig S. Tucker. Pond Aquaculture Water Quality Management. Kluwer Academic Publishers, 1998.ISBN: 9781461554073.
[4] G.S. Bilotta and R.E. Brazier. “Understanding the influence of suspended solids on wa-ter quality and aquatic biota”. In: Wawa-ter Research 42 (2008), pp. 2849–2861.DOI: http: //dx.doi.org/10.1016/j.watres.2008.03.018.
[5] R. J. Davies-Colley and D. G. Smith. “Turbidity, Suspended Sediment, And Water Clarity: A Review”. In: Journal Of The American Water Resources Association 37 (2001), pp. 1085–1101. DOI: https : / / doi . org / 10 . 1111 / j . 1752 - 1688 . 2001 . tb03624.x.
[6] Shigeru Nakayama. “Microwave Measurements of Low Pulp Concentration in Paper-making Process”. In: The Japan Society of Applied Physics 33 (1994), pp. 3614–3616.DOI: https://doi.org/10.1143/JJAP.33.3614.
[7] Malcolm J.W Povey. “Study of ultrasonic heat meter measurement error caused by sound attenuation in different water”. In: Procedia Engineering 205 (2017), pp. 4038– 4044.DOI: https://doi.org/10.1016/j.proeng.2017.09.878.
[8] Shi Shuo, Fan pengfei, and Liu lining. “Particulate characterization by ultrasound”. In: Pharmaceutical Science & Technology Today 3 (2000), pp. 373–380.DOI: https://doi. org/10.1016/S1461-5347(00)00310-2.
[9] Li Tan and Jean Jiang. Digital Signal Processing: Fundamentals And Applications. Elsevier, 2013.ISBN: 9780124158931.
[10] C Shen and U Lemmin. “Ultrasonic measurements of suspended sediments: a concen-tration profiling system with attenuation compensation”. In: Measurement Science and Technology 7 (1996), pp. 1191–1194. DOI: https : / / doi . org / 10 . 1088 / 0957 -0233/7/9/001.
BIBLIOGRAPHY
[11] C. C. Sung, Y. J. Huang, J.S.Lai, and G. W. Hwang. “Ultrasonic measurement of sus-pended sediment concentrations: an experimental validation of the approach using kaolin suspensions and reservoir sediments under variable thermal conditions”. In: Hydrological Processes 22 (2008), pp. 3149–3154.DOI: https://doi.org/10.1002/ hyp.6899.
[12] Peter D. M. Spelt, Michael A. Norato, Ashok S. Sangani, and Lawrence L. Tavlarides. “Determination of particle size distributions from acoustic wave propagation measure-ments”. In: Physics of Fluids 11 (1999), pp. 1065–1080.DOI: https://doi.org/10. 1063/1.869977.
[13] S. Mostafa Siadatmousavi, F. Jose, Qin Chen, and H. H. Roberts. “Comparison be-tween optical and acoustical estimation of suspended sediment concentration: Field study from a muddy coast”. In: Ocean Engineering 72 (2013), pp. 11–24.DOI: https: //doi.org/10.1016/j.oceaneng.2013.06.002.
[14] Cihan Sahin, Mehmet Ozturk, and Burak Aydogan. “Acoustic doppler velocimeter backscatter for suspended sediment measurements: Effects of sediment size and at-tenuation”. In: Applied Ocean Research 94 (2020), p. 101975.DOI: https://doi.org/ 10.1016/j.apor.2019.101975.
[15] David A. Mann, Zhongmin Lu, and Arthur N. Popper. “A clupeid fish can detect ultra-sound”. In: Nature 389 (1997), p. 341.DOI: https://doi.org/10.1038/38636. [16] Victor Frenkel, Eitan Kimmel, and Yoni Iger. “Ultrasound-induced cavitation
dam-age to external epithelia of fish skin”. In: Ultrasound in Medicine & Biology 25 (1999), pp. 1295–1303.DOI: https://doi.org/10.1016/S0301-5629(99)00069-1.
