Dynamic signal from fuel drop spreading over a rough glass

6.1 Prototype of an optical sensor

6.1.3 Dynamic signal from fuel drop spreading over a rough glass

Another novel method that enables fuel adulteration detection is the measurement of time dependent backscattered signal (TDBS), which has not been published. The prototype of a handheld sensor is incorporated with a wireless signal detection unit, namely wireless connection to the personal computer (PC) which enables monitoring of the time dependent signal through a PC. In Fig. 6.1 are shown the measured TDBS from the prototype sensor with kerosene on top of the rough surface. It is obvious from (Fig. 6.1) that, for kerosene there is a gradual rise with more apparent hysteresis, to a maximum and subsequent decrease in the TDBS signal for the rough surface.

On the other hand, Fig. 6.2 shows the TDBS from the prototype sensor, with au-thentic and adulterated diesel oils on top of the rough surface. It is obvious from (Fig.

6.2) that, pure diesel oil on the rough surface show slow but continuous increase in TDBS. However, adulterated diesel oils exhibit interesting behavior in terms of the nature of the TDBS signal and positioning by initially showing a raise and subse-quently decrease in signal strength as compared to that of the pure diesel oil.

The difference in the behavior of fuels on the rough glass surface enable screening of fake diesel oils. This demonstrates the second novel approach for fuel adulteration detection by the same prototype. Moreover, the behavior of 10% adulterated sample is similar both in contact angle as well as in the detected signal (Fig. 6.2). In the con-tact angle measurement refer (section 6.2), the value for authentic diesel oil is closer

to that of 15% adulteration, same case holds for the case of time dependent signal.

Moreover, in the time dependent signal the 10% adulteration sample has a plateau type behavior across most part of the measured time duration, showing a strange behavior which cannot be detected by refractive index readings. This shows the pe-culiarity of this measurement method as compared to the normal measurement mode of the device.

Figure 6.1. Time dependent backscattering signal (S) from the prototype sensor with kerosene on top of the rough surface.

Figure 6.2. Time dependent backscattering signal (S) from the prototype sensor with authentic and adulterated diesel oils on top of the rough surface.

37 6.1.4 Summary

The prototype of an optical sensor was realised, this prototype detects a combination of both scattered as well as reflected light from the interface between the glass and fuel film. Firstly, the measurement mode which provide one averaged signal value was utilised, with this mode it was possible to differentiate varying mixtures of adul-terated diesel oils with high confidence. Secondly, the other measurement mode of the prototype was also investigated. With the second mode it was possible to meas-ure the time dependent backscattered signal for different diesel oil mixtmeas-ures, signals from various samples assume different shapes, providing another novel alternative for adulteration detection.

On the other hand, unlike the abbe refractometer readings which can only be dis-tinguished in second and third decimal places, the readings from the handheld de-vice can easily be interpreted by local inspectors because of differences in second digit for different adulterated fuel samples. The drawback of the handheld device is lack of mechanical stability especially in field conditions, this should be taken care of for the device to perform accurately in field conditions. Moreover, even though the initial prototype sensor is expensive, it is possible to design low cost devices, thanks to the availability of cheap CCD sensors and lasers.

6.2 EXTINCTION COEFFICIENT

Extinction coefficient was used by various researchers as analytical technique for characterizing materials especially consumer oils (Vanak Z. P., 2010; Amereih S., 2014). In this study, we were partly motivated by the pioneer work of Dombrovsky et al (Dombrovsky L. A., 2003) which studied the spectral properties of diesel fuel droplets across a broader spectral range. The positive results that were obtained in this section led us to investigate other optical properties, such as wavelength depend-ent refractive index, excess permittivity, and imaginary excess permittivity.

Here we study the wavelength dependent refractive index for one summer and one winter diesel oils, namely samples A, B and D (Paper III), similar procedures were followed for sample C (Paper II). The wavelength dependent extinction coeffi-cient was calculated by using Eq. (3.11) and the transmittance data in Fig. 4.4. To demonstrate these results, we utilized the data for sample A, similar trend is exhib-ited for sample B. From Fig. 6.3 we can make interesting observations regarding the different signals from 1190 nm to 1220 nm. Firstly, there do exist fingerprints with higher peaks at 1195 nm and lower peaks at 1215 nm. Secondly, the behaviour of kerosene is contrary to that of diesel oil, therefore it is possible to use the magnitude of the height ratios to differentiate pure diesel oils from kerosene.

The height property is insignificant for separating the different adulterated sam-ples mainly because, the height on the left side at 1195 nm is constantly higher when compared to that at 1215 nm. Therefore, the attempt to apply the ratio of heights for identification purposes fail in this case. Moreover, the curves are not ordered in the

order of increase or decrease of adulteration percentage, therefore lacking potential to separate and discriminate the different samples.

Figure 6.3. The zoomed extinction coefficient curves for kerosene, authentic and adulterated sample A in the vicinity of 1190 nm to 1220 nm (Paper III).

Next, we explore another interesting alternative in Fig. 6.4 where the curves for same samples as in (Fig. 6.3) are depicted, but now in the spectral range from 1390 nm to 1427 nm. The nature of the curves is contrary to those in (Fig. 6.3), namely there is no polarity for both the left and the right peaks. The curves are arranged in accord-ance to adulteration percentages, namely the lower curve is for the lowest adultera-tion percentage (5%) while the higher curve is for the highest adulteraadultera-tion percentage (10%). Thus, the relationship between the extinction coefficient curves and adultera-tion percentage is of a linear fashion, which enables separaadultera-tion and discriminaadultera-tion.

It is possible to resolve and separate well the curves for mixtures of sample A and sample D, while the curves for mixtures of sample B and sample D almost overlap.

For the curves of sample B (refer Paper III). Both curves for adulterated samples are higher compared to authentic diesel oil and kerosene, from the curves in (Fig. 6.3(b)) already one can observe the excess extinction coefficient, which implies excess refrac-tive index and consequently excess permittivity. This makes the spectral region of 1390 nm to 1420 nm very important for separating the samples with different adul-teration percentages. Because of these results we were able to explore other optical properties, starting first with wavelength dependent refractive index, and then the concept of imaginary excess permittivity. These were necessary to assess the differ-ences between samples adulterated by different volume fractions of kerosene, in the

39 spectral region around 1400 nm, therefore making use of the recent theory proposed in (Iglesias T. P. & Reis J. C. R, 2016).

Figure 6.4. The zoomed extinction coefficient curves for kerosene, authentic and adulterated sample A in the vicinity of 1390 nm to 1427 nm (Paper III).

6.3 WAVELENGTH DEPENDENT REFRACTIVE INDEX

Next, as it was for the case of (section 6.2) the refractive index as a function of wave-length was obtained by using Eqs. (3.12) and (3.13) which were estimated by SSKK integration in the finite spectral range from 431 nm to 1600 nm. In Figure 6.5(a) are shown the wavelength dependent refractive index curves for sample A. Strong vari-ations are observed in the refractive index values especially for the regions exhibiting stronger absorptions, namely around 1200 nm and 1400 nm. The curve for 15% is highly distinct while those for 10% and 15% are almost overlapping. The strange be-havior of sample A and especially that of 10% is an indicator of possible chemical interactions, these were suspected also in (section 6.1). The behavior of sample B in Fig. 6.5(b) is slightly different in the sense that both curves are clearly distinct from one another. The wavelength dependent refractive index data were later utilized to calculate the imaginary excess permittivity by applying Eq. (3.17). For more detailed description of the Imaginary excess permittivity calculation refer (Paper III).

Next in Fig. 6.5(c) are shown the wavelength dependent refractive index curves.

The curve for authentic sample C is the lowest while, the curves for 5% adulteration is the highest, followed by 10% and 15% consecutively. Moreover, we observe in (Fig.

6.5(c)) that, for all three samples namely, 5%, 10% and 15% adulteration, the curves are clearly distinct from each other. Therefore, SSKK provides better alternative for separating samples adulterated by different volumes of kerosene (Paper II).

In this section the wavelength dependent refractive index was explored to study adulterated diesel oil samples. Based on the results (Figs. 6.5(a), (b) and (c)), the ples are distinct. However, there is no a specific order followed by adulterated sam-ples. For example (in the case of sample B the curve for authentic diesel oil appears at the middle followed by 10% above it and 5% even further above while the curve for 15% is the lowest). Therefore, in the next section we explore the excess optical properties which reveal the presence of kerosene in diesel oils. Moreover, the prop-erties enable separation and discrimination of adulterated samples, in accordance to the volume of kerosene (Paper II and III).

41 a

Figure 6.5. The calculated refractive index curves for authentic and adulterated samples (a) Sample A (b) Sample B (Paper III) and (c) Sample C (Paper II).

6.4 EXCESS OPTICAL PROPERTIES

In this section we briefly report on the results of two most important findings of this thesis, namely excess permittivity and imaginary excess permittivity.

6.4.1 Excess permittivity

The refractive index values for samples A, B and C (both authentic and adulterated) are presented in Table 1 (chapter 4). These values were utilized for calculating the excess permittivity using Eqs. (3.15) and (3.16). The results for excess permittivity are presented in Table 4.

Table 4: Excess permittivity of adulterated fuel samples A-C.

Sample Volume fraction

Of Kerosene

Excess permittivity (10^-3)

A 5% -0.30

10% 2.63

15% 2.06

B 5% 4.25

10% 4.13

15% 2.83

C 5% 3.14

10% 3.17

15% 3.20

From (Table 4) the magnitude of excess permittivity (at 589 nm) for all the samples is small, this implies the existence of less chemical interactions. There is negative sign in the value of excess permittivity for sample A which is very different from all the other samples. Moreover, there is a unique behaviour regarding the value for sample A, namely 10% sample has highest value of excess permittivity in comparison to the values for 5% and 15%. (Paper III) This seemingly strange behaviour agrees with the behaviour of the time dependent back scattered signal measured by the new meas-urement mode of the prototype in (Paper I). The dynamic signal behaviour depicted in (section 6.1) shows a rather constant signal for 10% contrary to the plateau type signal of 5% and 15%. For the case of samples B and C only positive values for excess permittivity are observed. The values for sample B are interesting, namely 5% sample has the highest value while 15% has the lowest value, moreover, the values for 5%

43 and 10% are almost same whereas the value for 15% is very low. On the other hand, sample C has constant excess permittivity value, this might partly be due to the close-ness of its refractive index value with that of kerosene. Therefore, it is difficult to differentiate adulterated samples originating from sample C based on the single point (589 nm) excess permittivity value.

The wavelength dependent excess permittivity is important for exploring the op-tical behavior of different authentic and adulterated samples across a broader Vis-NIR spectral range. For this purpose, sample C is considered because according to Table 4 it is the most difficult case which gives almost constant value for different adulteration percentages. The wavelength dependent extinction coefficient was cal-culated using Eqs. (3.9) and (3.11). The results were utilized for calculating the wave-length dependent refractive index using Eqs. (3.12) and (3.13), transmittance values from Figs. 4.1 and 4.4, and refractive index values from Table 1. The results of Fig 6.5(c) were utilized for calculating the wavelength dependent excess permittivity by making use of Eqs. (3.15) and (3.16), and the results were applied to determine per-mittivity as a function of wavelength using Eq. (3.14). Fig. 6.6 shows the results of excess permittivity for two samples, namely 5% and 15%, there are clear nice over-lapping fingerprints at ca. 1200 nm and 1400 nm which are indicators of adulteration in the measured samples. Therefore, it is possible to discern the presence of kerosene in diesel oils based on excess permittivity curves. This is a novel finding thanks to this thesis.

Figure 6.6. Excess permittivity for mixtures of samples C and kerosene (D) (Paper II).

6.4.2 Imaginary Excess permittivity

The other interesting part of this thesis is the imaginary excess permittivity as a func-tion of wavelength, for different adulterated samples. These were calculated using

Eq. (3.17), and by utilizing the wavelength dependent refractive index data depicted in Fig 6.5 (a) and (b). In Fig. 6.7(a) the curves for imaginary excess permittivity of adulterated versions of sample A are shown. The order of the curves corresponds to the percentage of adulteration, namely the excess permittivity curve for 5% is the lowest while that for 15% is the highest. It is possible to clearly and with confidence separate and discriminate the samples based on these curves.

The results for sample B are presented in Fig. 6.7(b) where similar trend regarding the location and separation is observed. However, unlike the case of sample A where the curves for 5% and 10% are somewhat close to each other (Fig. 6.7(a)), here the curves for 5% and 10% are more clearly distinct from one another. The interesting results around 1400 nm are promising and open doors for the development of meas-urement methods, which operate in the small spectral range. This will greatly serve analysis time and will enable timely measurements to detect adulteration.

The current state of the art technologies applied for fuel adulteration studies, is based on spectroscopic measurements coupled with multivariate analysis (Bassbasi M., 2013; Marcio J. C. P., 2011; Fazal M., 2017). These methods rely on a broader spec-tral range and require commercial multivariate analytical software packages such as Unscrambler X, and PLS toolbox for Matlab, for data analysis. These are usually ex-pensive and requires long time (spectroscopic measurements and analysis), and spe-cial training, therefore usually suited for laboratory applications. One of the quests of this thesis is to come up with methods that will enable the development of field condition measurement devices. Therefore, both novel methods, namely excess per-mittivity and imaginary excess perper-mittivity are potential candidates for such a pur-pose. We infer that, these results can lead to more promising twofold solutions.

Firstly, the spectral region around 1200 nm and 1400 nm have been identified as the most important regions for identifying the presence of kerosene in diesel oils. There-fore, it is possible to utilize this narrow band and reduce the processing time for com-mercial software packages. Secondly, thanks to the novel results at 1400 nm, more doors are open for devices that exploit the concept of imaginary excess permittivity at a very narrow wavelength around 1400 nm.

45 a

Figure 6.7. The zoomed imaginary excess permittivity for samples adulterated by kerosene in the vicinity of 1390 nm to 1427 nm (a) Sample A (b) Sample B (Paper III).

6.4.3 Summary

The real and imaginary excess permittivity of diesel oils adulterated by kerosene were studied. This was achieved by combining data measured by spectrophotometer and Abbe refractometer, and thereafter utilizing the SSKK analysis to calculate

wave-46

length dependent refractive index, excess permittivity and imaginary excess permit-tivity. The excess permittivity reveals nice spectral fingerprints at ca. 1200 nm and 1400 nm, which indicates the presence of kerosene in diesel oils. The fingerprints are unavailable if refractive index or transmittance data are considered independently.

The imaginary excess permittivity method provides nicely ordered curves around 1400 nm, which clearly separate diesel oils adulterated by different volume fractions of kerosene.

This is the first time in applied optics that the model-independent SSKK analysis is utilized and proposed for excess permittivity and imaginary excess permittivity studies of any binary or multi-mixture liquids. The proposed methods in the present form are laboratory based. Nevertheless, due to availability of portable NIR spectro-photometers and small digital handheld refractometers, it is possible to design a measurement unit which is equipped with a laptop for field measurements by fuel regulatory authorities. The disadvantage of the system is requirement of well trained personnel, with background in data analysis packages such as Matlab, to carry out the measurements and analysis.

6.5 HANDHELD REFRACTOMETER METHOD

It is possible to use some of the already existing devices to solve current measurement problems. Herein, this is demonstrated, namely a novel measurement method incor-porated with the handheld Abbe refractometer was applied to solve fuel adultera-tion. We briefly consider a training set in laboratory conditions, and later the novel method is confirmed by field measurements.

6.5.1 Training set

Here we present the results of a training set in the laboratory depicting the directly observed relationship between the ideal permittivity curve, and the adulterated sample as the volume of authentic diesel oil is increased. In Fig. 6.8 are shown the curves for sample A where the dashed line represents the permittivity of ideal mixture. This was calculated by the modified ideal law Eq. (3.20) and utilizing data of Table 1. The differences between the measured permittivity and ideal permittivity is directly observed from the graph, because the points and the curve are at separate locations. However, as the volume of the adulterant is increased, the measured permittivity approaches the ideal permittivity curve. The case of sample B is even much complex because the refractive index between diesel oil (sample B) and kerosene (sample D) are very close, unlike that of samples A and D. As was mentioned earlier, the proposed method works much better when the refractive index difference is large enough.

47 Figure 6.8. Ideal permittivity for mixtures of sample A, and the measured permittivity values for adulterated samples obtained by table model Abbe refractometer (Paper IV).

6.5.2 Field measurements

Next, we present the application of a novel measurement method for field adultera-tion screening in Tanzania, these are depicted in Fig. 6.9. These were calculated using Eq. (3.20) and data from Table 2. The nature of the curves is different from those of laboratory measurements in (Fig. 6.8), this is caused by interchange in roles where by kerosene has higher permittivity value as compared to authentic diesel oil. The true permittivity points are at separate locations thus deviates from the ideal permit-tivity curves especially those for 10% and 15%. This already is an indicator for excess permittivity. From Fig. 6.9 it holds similarly to (Fig. 6.8) that, as the volume of au-thentic diesel oil is increased into the suspected sample, the values for true permit-tivity keeps approaching those of ideal permitpermit-tivity. Moreover, the trend is similar for different curves at different temperatures. Therefore, this measurement method is feasible and can be exploited to scan for adulteration (Paper IV).

Figure 6.9. Ideal permittivity for Tanzanian sample, and the true permittivity values for adulterated samples obtained by handheld Abbe refractometer data (Paper IV).

6.5.3 Summary

In this work we have demonstrated for the first time the application of handheld Abbe refractometer which traditionally measures sugar concentration, for possible application to detect fuel adulteration. To test the suspected sample firstly, pure die-sel oil is added and there after the concepts of ideal and true permittivity are ex-ploited, to study the existing relationship between the true permittivity of the meas-ured sample and the ideal permittivity. Obviously, the Tanzanian sample performs

In document Development of optical measurement techniques and data analysis methods for screening of adulterated diesel oils (sivua 51-0)