An Improved Microwave Sensor for Qualitative Assessment of Recycled Cooking Oils

An improved planar microwave sensor comprising a resistor-embedded top microstrip line and a ground-etched complementary split-ring resonator (CSRR), designed on a 1.6-mm Taconic substrate (<inline-formula> <tex-math notation="LaTeX">$\varepsilon _{r}$ </tex-math></inline-formula> = 2.2), is presented to assess the quality of recycled cooking oils. The proposed resonant sensor exhibits an improved quality factor and a reasonably larger shift in the resonant parameters under loading conditions, thereby facilitating the detection of even slight variations in electrical and chemical properties of low-loss dielectric samples. The enhanced performance of the proposed sensor is attributed to incorporating a specific valued lumped resistor on the top side of a microstrip line, which helps in coupling the maximum electric field to the CSRR and providing an extreme improvement in the <inline-formula> <tex-math notation="LaTeX">${S} _{21}$ </tex-math></inline-formula> notch (−40 dB) near resonance. The proposed sensor is designed and simulated using the full-wave simulator, the Computer Simulation Technology (CST) Microwave Studio (MWS). The equivalent circuit model is developed using the Advanced Design System (ADS) to understand the behavior of the circuit in detail. An analytical formulation is implemented using MATLAB, which provides a definite relationship between the small value of the resistor connected on the top of the microstrip line and the <inline-formula> <tex-math notation="LaTeX">${S} _{21}$ </tex-math></inline-formula> value near resonance. Several fresh and recycled cooking oil samples are measured, and it is observed that their acid values (AVs) get degraded from 1.5 to even 9.4 indicating the toxic nature of the recycled oils and their adverse effect on human health. The AV of the oil samples in the present situation is obtained under a controlled chemical environment by developing an empirical relationship between this parameter and the normalized frequency and the <inline-formula> <tex-math notation="LaTeX">${S} _{21}$ </tex-math></inline-formula> notch. The fabricated sensor provides a substantially higher value of fractional sensitivity (~25 MHz/mgKOHgm<inline-formula> <tex-math notation="LaTeX">$^{-1}$ </tex-math></inline-formula>) and larger deviation in the <inline-formula> <tex-math notation="LaTeX">${S} _{21}$ </tex-math></inline-formula> notch level (~3.33 dB/mgKOHgm<inline-formula> <tex-math notation="LaTeX">$^{-1}$ </tex-math></inline-formula>), for a unit change in the oil rancidity.

An Improved Microwave Sensor for Qualitative Assessment of Recycled Cooking Oils Apala Banerjee , Member, IEEE, Nilesh K. Tiwari , Member, IEEE, Farheen Fatima , Graduate Student Member, IEEE, and M. Jaleel Akhtar , Fellow, IEEE Abstract-An improved planar microwave sensor comprising a resistor-embedded top microstrip line and a ground-etched complementary split-ring resonator (CSRR), designed on a 1.6-mm Taconic substrate (ε r = 2.2), is presented to assess the quality of recycled cooking oils.The proposed resonant sensor exhibits an improved quality factor and a reasonably larger shift in the resonant parameters under loading conditions, thereby facilitating the detection of even slight variations in electrical and chemical properties of low-loss dielectric samples.The enhanced performance of the proposed sensor is attributed to incorporating a specific valued lumped resistor on the top side of a microstrip line, which helps in coupling the maximum electric field to the CSRR and providing an extreme improvement in the S 21 notch (−40 dB) near resonance.The proposed sensor is designed and simulated using the full-wave simulator, the Computer Simulation Technology (CST) Microwave Studio (MWS).The equivalent circuit model is developed using the Advanced Design System (ADS) to understand the behavior of the circuit in detail.An analytical formulation is implemented using MATLAB, which provides a definite relationship between the small value of the resistor connected on the top of the microstrip line and the S 21 value near resonance.Several fresh and recycled cooking oil samples are measured, and it is observed that their acid values (AVs) get degraded from 1.5 to even 9.4 indicating the toxic nature of the recycled oils and their adverse effect on human health.The AV of the oil samples in the present situation is obtained under a controlled chemical environment by developing an empirical relationship between this parameter and the normalized frequency and the S 21 notch.The fabricated sensor provides a substantially higher value of fractional sensitivity (∼25 MHz/mgKOHgm −1 ) and larger deviation in the S 21 notch level (∼3.33 dB/mgKOHgm −1 ), for a unit change in the oil rancidity.

I. INTRODUCTION
C OOKING or edible oil is one of the most commonly used groceries in many parts of the world.It is quite common to reuse cooking oil, especially in developing countries, when a significant amount of oil is left after frying certain food items.This kind of procedure is also sometimes followed in food industries to attain maximum profit at the cost of human health.It is now ascertained from many studies that consumption of these types of reused oils causes cell degeneration, myocardial infarction, increased blood cholesterol level, disturbed metabolism, and so on [1].
It has been discovered that when the edible oil is repetitively heated for reusing, it might turn rancid, thereby increasing the acid value (AV) of oil [2].
The AV is one of the main parameters to characterize oil as it is the mass of potassium hydroxide (KOH) in milligrams that is required to neutralize one gram of fat, oil, and so on.It is the number of free acids present either in a chemical compound or a mixture of chemical compounds.An increase in the AV represents an increase in the toxic nature of the oil that might lead to certain harmful consequences on human health.The AV is a parameter that can conveniently distinguish fresh oil from multiple reused oil for household and commercial applications.
In the past, several chemical methods have been proposed to determine the AV of oils, such as the complex titration technique [3], the head-space gas chromatography [4], and the voltammetry measurements [5].However, the applicability of these conventional chemical methods is quite limited in terms of on-field real-time testing due to the involvement of controlled environmental conditions.This type of limitation posed by chemical methods can be relaxed using the noninvasive RF and microwave methods utilizing the high-sensitivity resonators [6], [7], [8], [9], [10].It is important to note here that microwave methods [11], [12], [13], [14], [15], [16], [17], [18] mainly rely on measured dielectric signatures of the test sample.However, it is observed that there is no substantial difference between the dielectric properties of the fresh oil sample and its heated variant for this kind of scenario.It thus necessitates the design of an RF sensor with enhanced sensitivity and high quality factor, which can provide a noticeable difference in measured parameters corresponding to a slight change in the dielectric signature of low-loss materials.
In the past, split ring resonator (SRR)-based sensors have been proposed to measure pipeline breaching due to high pH levels [19] with sufficient quality factor, which is later enhanced to multifold times using active sensors for detecting electrically small carbon samples [20] and noncontact liquid sensing [21], [22].These circuits, however, require external biasing arrangement to achieve the required performance and introduce much circuit complexity.To exploit the resonance phenomena in oil applications, Tiwari et al. [23] reported the study of varying properties of edible oil due to the inclusion of adulterants using passive sensors.Quantification of oils 1557-9662 © 2023 IEEE.Personal use is permitted, but republication/redistribution requires IEEE permission.
Authorized licensed use limited to the terms of the applicable license agreement with IEEE.Restrictions apply. in Ayurveda medicines is also reported in [24].In [25], a single-port differential sensor and a differential T-junction sensor [26] are terminated with specified resonating elements, including lumped resistor, to cause a significant notch in the reflection coefficient.In [27] and [28], two lumped resistors are placed on two ends of the host microstrip line to realize a high-impedance structure.However, using a surface-mounted resistor on the top side of a microstrip line in a conventional complementary split-ring resonator (CSRR) sensor to achieve high quality factor and its significance for specific applications, such as the quality assessment of edible oils, has not been highlighted or emphasized earlier in the literature.The present work gives a very comparable performance corresponding to its active counterpart using a simple traditional passive structure (Table V) with improved performance and ease of sample placement.In this article, the proposed sensor exhibits a substantially higher value of the unloaded Q factor (3680), making it quite appropriate to test low-loss dielectric samples such as cooking oils.The sensor is fabricated on a 1.6-mm-thick Taconic TLY-5 substrate (ε r = 2.2) with an S 21 notch level nearing −40 dB under unloaded conditions.The sensor exhibits a fractional sensitivity of 10% in simulation and 9.5% in measurement results.The electromagnetic (EM) model of the sensor is developed in Computer Simulation Technology (CST) Microwave Studio (MWS), which provides here flexibility to control the S 21 magnitude level just by changing the lumped resistor value.The frequency of operation is 1.83 GHz since oil characterization is very sensitive at this frequency regime.It can also be seen that a very small difference in the measured dielectric signature can be easily distinguished and observed at this frequency [29], [30], [31].After performing EM simulation, the quasi-static lumped model of the sensor is developed using Advanced Design System (ADS) in conjunction with an analytical model to understand the behavior of measured S 21 as a function of lumped resistance.Finally, various oil test samples are measured with the developed sensor to obtain the corresponding change in the frequency shift and S 21 magnitude level.Later, the standard in-house chemical approach is used to record the AVs of these oil samples, i.e., the fresh oil and its repetitively heated/cooked variant.Consequently, a numerical relationship between the shift in the resonant frequency, the S 21 level, and the AV is developed to validate the proposed high Q sensor's role in detecting toxicity in reused cooking oils.The schematic for the flow is shown in Fig. 1.

A. Typical CSRR Sensing Topology
A conventional CSRR-based sensor is designed and simulated using the numerical EM simulator CST MWS under the loaded and unloaded conditions to record the response at the designated frequency, i.e., 1.83 GHz.The schematic of the conventional CSRR and its dimensions is shown in Fig. 2, where the top view shows the microstrip line to excite the CSRR [Fig.2(b)] etched on the ground plane.The figure shows that at 1.83 GHz, the resonator exhibits a quality factor of 91.5, having an S 21 notch level of −23 dB [Fig.2(c)].
This quality factor (Q) and notch level of the conventional CSRR are not sufficient, mainly when a small quantity of the low-loss sample is used in the loaded case.To overcome the shortcomings of the conventional CSRR, a modified sensor is proposed with multiple times improvements in its quality factor and S 21 notch level at the same resonant frequency.The substrate dimension is 50 × 40 mm and the substrate used is Taconic TLY-5 lossy dielectric substrate (ε r = 2.2) with a substrate thickness of 1.6 mm.

B. Modified Feed Resistive Topology
An improved CSRR sensor design using a modified feed structure is proposed to account for the limitations above.The modified arrangement is carried out to ensure a better Q-factor and notch level.The proposed RF sensor, including the microstrip line, is designed using the full-wave EM simulator, the CST Studio suite.For starting the simulation, the corresponding 50-microstrip line is first designed on a Taconic (ε r = 2.2) substrate of thickness 1.6 mm by computing the initial geometrical parameters of the line using the standard empirical relationship [32].After designing the corresponding microstrip line, the CSRR of the appropriate dimension is etched on the ground plane to resonate at the designed frequency of 1.83 GHz.The CSRR is excited using the electric field of the top microstrip line, which is perpendicular to the CSRR plane.
To start the possible design modification, first, two limiting cases of feed topology are analyzed, as shown in Fig. 3.It can easily be seen that the S 21 profile of the structure can be regulated using certain modifications in the feed line.The change in feed configuration results in to change in coupling between the feed line and CSRR.Moreover, the overall structures encompass dielectric and conductor loss with negligible radiation loss for the chosen frequency range.Therefore, a lumped element can be inserted in series with the main signal line without loss of generality, maintaining that the resistance value selected can provide a nearly similar S 21 and S 11 magnitude profile.Notably, the main concern is limited to designing magnitude-based sensors; hence, only these two parameters are analyzed here.To corroborate the argument, an S-parameter analysis corresponding to continuous feed line with lumped resistive element is carried out, refer to Fig. 4. A close look at the plot shows a close match between the S 21 responses of conventional structure [Fig.2(c)] with the feed line with a 20-resistor [Fig.4(a)].However, there is one more degree of freedom to control the notch of the structure without disturbing any other part of the design.The variation of R is then carried out in a small discrete step size to record the behavior of the desired S 21 parameter.At the same time, S 11 parameters are also observed to ensure that the proposed design is not inducing any additional reflections, as shown in Fig. 4(c).
The E-field confinement changes with the change in resistance, and at a particular R, the maximum confinement of the E-field takes place.For a very small value of R, i.e., 0.01 [Fig.4(a)], S 21 is close to that of the continuous microstrip  line on the top.In other words, this kind of situation is similar to that of the conventional CSRR topology.As the resistance is slowly increased, a certain value is obtained where the maximum field confinement takes place, which is −40 dB for 10 [Fig.4(b)], and then starts approaching an open-circuit condition (Fig. 3) as the resistance is increased up to 100 and above [Fig.4(a)].From this plot, it can be noticed that the targeted limiting profiles can be generated by simply varying the absolute value of resistance.The input-output impedance for the sensor is 49.92 as observed from CST Studio during the simulation.
More interestingly, the modified feed design also provides various S 21 notch levels.In the present scenario, the maximum notch level obtained is near −40 dB for the corresponding 10-resistance of package RC0603JR-0710RL of PHG2A-KIT-ND, which is incorporated in the top microstrip line, as shown in Fig. 5.The entire structure is then simulated in CST-MWS, and the transmission plot for several parametric resistor values is taken to choose the best transmission notch for the proposed sensor.The comparison with the conventional sensor is tabulated in Table I.
The length of the slot is taken as 0.3 mm corresponding to the width (0.3 mm) of the SMD resistor model to accurately solder it in the gap (Fig. 6).

C. Equivalent Circuit Analysis
The detailed parametric analysis of the designed sensor with a particular lumped resistance value is carried out in ADS software using the designed lumped equivalent circuit with EM-generated S-parameters as the goal function.
The designed equivalent circuit (delta form) and its modified (star form) are shown in Fig. 7 and the schematic is shown in Fig. 8. L 1 and L 2 are the inductances of the host microstrip line, and "r " denotes the resistance incorporated in between the microstrip line.C 1 and C 2 are the symmetrical arrangements of the coupling capacitance from the microstrip line to the ground plane.The resonator is modeled with the parallel Rs, Ls, and Cs (RLC) circuit.
The ADS-generated plot for the optimized set of parameters (Table II), shown in Fig. 9, depicts that the EM and circuit simulated results are in close agreement with each other.The slight discrepancy in the circuit simulation might be attributed to the parasitic effect that is not considered for the sake of simplicity.

D. Analytical Formulation for Designed Sensor
The proposed configuration comprises a passive element connected in line with the host microstrip line, which helps regulate its S 21 characteristic without any physical modification in the rest of the design.The proposed sensing configuration can be considered a generalized feed topology with open-ended and through feed (conventional) as its two special cases (Fig. 3).For instance, considering the minimal value of the passive element, the overall structure response turns out to be quite similar to the case when the microstrip line is continuous between ports 1 and 2 with ground-etched CSRR, i.e., band notch response.To better illustrate the scenario, the proposed sensor structure can first be represented as a two-port network as shown in Fig. 10   where On the other hand, the S-parameters of the network can also be obtained using the corresponding impedance model of the two-port network, i.e., Z -parameters of a two-port network stated as where After performing specific mathematical manipulations, the above relation can be simplified to express the transmission coefficient of symmetrical and reciprocal two-port network Authorized licensed use limited to the terms of the applicable license agreement with IEEE.Restrictions apply.The characteristic impedance is considered here at 50 throughout the analysis as per the designed microstrip line.
The corresponding EM response can develop the quasi-static lumped equivalent model to generate the structure's Z parameters.A generalized quasi-static model of the designed sensor is presented in Figs. 6 and 7, which turns out to conventional equivalent circuit model of ground-etched CSRR for r = 0.The circuit is rearranged using standard delta-star conversion for the dashed region (aa ′ bb ′ ) to simplify the analysis, as shown in Fig. 8.
The Z -parameters of the simplified circuit can be written as ( The following relations give the star network R A , R B , and R C : The impedance from reactance X c and X L is given as Similarly, the impedance/admittance corresponding to the parallel capacitance and inductance of the RLC resonator may be represented as Therefore, the admittance of the resonator can be written as The relation (3) can now be rewritten using the relationships above (4)-( 9) The expression of N and D is provided as follows: where From (11), the transmission parameter S 21 becomes a function of the lumped resistance "r " if we consider all other parameters constant.
The CST-generated S-parameters corresponding to the conventional situation, assuming r = 10 , are then provided as an input to the numerical circuit simulator to get the numerical values of the parameters shown in Table II.When a lumped resistance is inserted in the microstrip line above the CSRR, this resistance provides a relatively high impedance path to the RF signal.This leads to a more amount of RF signal propagation to the bottom CSRR compared to the scenario when the top microstrip line is continuous.It is observed that there is some optimum value of the resistance for which the maximum notch in the S 21 level is obtained.
The above relation is then plotted in MATLAB by addressing "r " as a user-defined variable, and the plot is shown in Fig. 11.It can be clearly observed from the above figure that for a lumped resistor value of 10 , the maximum notch in S 21 level is obtained for the given scenario.This kind of phenomenon may be explained by noting the fact that for the optimum resistance value, the electric field at the bottom of the line and the magnetic field on the top line become equal, resulting in the maximum resonant condition.This maximum resonance condition ensures the maximum field confinement in the CSRR region, resulting in a small signal reaching the other port.When the microstrip line is continuous (no resistor), the RF power flowing through the top line gets coupled with the ground-etched CSRR at its resonant frequency, leading to the development of band-stop characteristics.Incorporating the resistive element of a small value does not significantly alter the transmission characteristic of the sensor.When the resistor value becomes significantly high, then the situation becomes similar to the gap-coupled microstrip line where the signal path from input to output of the line is provided only at the resonance frequency through the bottom CSRR leading to bandpass characteristics.However, for a particular value of small resistance connected to the top microstrip line, the electric field at the bottom of the line becomes almost equal to the magnetic field on the top line, thereby providing maximum notch in the S 21 level.
The explanation for the above improvement in the S 21 notch level and thereby in the Q-factor of the stopband resonator lies in the fact that the high impedance path is incorporated at the center of the microstrip line entire structure exhibiting a planar representation of a "slow wave structure."While plotting in MATLAB, the substrate specification could not be taken into the developed code, which neglects the losses due to the substrate.Due to this fact, the damping conditions are not included, and thus, the magnitude of the S 21 notch is much higher compared to the experimental response where such losses are involved.Looking into an E-field contour plot in simulation, From Fig. 12, it becomes clear that the guided wavelength for the incident one-half section is 12.35 mm,  and upon reaching the resistive path, most of the RF signal traverses to the lower sensing region confining maximum electric field to the CSRR splits.
In the present situation, the outer diameter of the CSRR element used in the analysis is 5.5 mm, which is less than one-twentieth of the effective wavelength near the operating frequency.Even otherwise, the resistor in the proposed structure is integrated near the CSRR region, where the main aim is to compare the performance of the sensor with and without the lumped resistor of any given value.Hence, the major emphasis here is to investigate the effect of the resistor on a relative scale by changing its values while keeping other parameters to be the same.While integrating the lumped resistor, a slot of 0.3 mm is etched on the host line where the resistor is embedded, thereby disturbing the λ /4 structure.Accordingly, the structure is not behaving like a quarter wavelength as per the continuous length of having a dimension of 25 mm but rather like a λ /8 line at the operating frequency, which is still a valid approximation.
The maximum resonance condition ensures maximum electric field confinement, resulting in a very small signal reaching the other port.The field distribution is shown for the conventional and the modified case in Fig. 13, keeping the same dB scale.To choose the sample holder and volume, a field depth analysis is also carried out along the positive z-axis, which is aligned to the sample holder, and obtaining an optimum field strength sufficient for interaction with the sample under test (Fig. 14), which is obtained as 5 mm.From Fig. 15, the plane of the sensing area is taken to be 0-mm reference, and then, gradually along the z-axis, the E-field (V/m) has been plotted up to 5 mm.These field values are obtained using the "cut-plane" and "fields on plane" in the 2-D/3-D tool using the CST MWS.The field on planes is noted for every 1 mm starting from 0 up to 5 mm and the values are obtained.The field tends to decrease along the z-axis; 5 mm has been carefully chosen for our sample placement because of the following two factors.
1) In subset Fig. 12(a) of Fig. 15, the field intensity is plotted for 5 mm and it shows that the strength is still sufficient for suitable interaction between sensor and sample.2) In subset Fig. 12(b) of Fig. 15, the field intensity is plotted for 6 mm and it shows that the strength has decreased in its intensity, and thus, it is suitable to stay below the 6-mm margin.
Keeping 5 mm as the upper limit, the sample height has been kept at 5 mm.At 3 mm, the field value is measured and becomes sufficient for sample interaction.As a result, the sample amount has been taken to be 3 mL.When the microstrip line is continuous (no passive element), power flow through the line gets coupled with the ground-etched CSRR at its resonant frequency, leading to the development of band-stop characteristics.Incorporating the passive restive element of negligibly small value does not significantly alter the transmission characteristic.However, when its value increases, the coupling phenomenon begins to change where some part of the signal follows the other path, as is the case with an open-ended microstrip line.For the amount of the particular value of resistance, both signal characteristics become nearly similar in magnitude and cancel each other due to the associated phase difference between them.It leads us to the minimum notch level in the transmission coefficient.After this, the signal flow through the other path increases in magnitude, leading toward the bandpass characteristic for a relatively higher resistance value.The obtained behavior can be visualized in Fig. 3. Next, to appreciate the effect of symmetry and hence balance position of resistor above the ground-etched CSRR (ref.position 0), a positional analysis of the transmission parameter and the quality factor is carried out.The resistance, when placed just above the resonator, is assigned a position of 0 mm, and then, the position is gradually shifted left by small increments (from Position 1 to Position 4) (Fig. 16) until the lumped device reaches the CSRR edge.It is observed that as the lumped device moves toward the CSRR edge (from Position 4 to Position 1), there is a gradual decrease in Q as well as the S 21 notch level.
Thus, it can be ascertained that the increased asymmetry in the resistor position results in the reduction in the S 21 notch level, which is quite evident from the unbalanced field distribution.

E. Sensitivity Analysis
The sensitivity of the proposed sensor is carried out by loading the sample with a 6 mm radius and 5-mm sample holder height (a sample volume of 3 mL) with varying permittivity and then calculating the shift in the resonant frequency.The sample is loaded simply on top of the sensing region directly without any complex arrangement unlike liquid sensing using a microfluidic system.Sensitivity is thus defined as the shift in resonant frequency per unit change in the dielectric constant of the loaded sample.The sample dielectric is varied from 1 to 3 whose plot is shown in Fig. 17.
The fractional sensitivity of the sensor is obtained to be 10% in EM simulation and 9.5% in measurement.
This sensitivity is calculated as , where f is the shift in resonant frequency when the Authorized licensed use limited to the terms of the applicable license agreement with IEEE.Restrictions apply.sensor undergoes a unity change in its dielectric constant and f 0 is the unloaded resonant frequency at which the resonator operates.The Q-factor is also improved even under loaded conditions up to 900 in simulation.A 3-D demonstration of the proposed sensor with the sample holder is shown in Fig. 18.The holder is filled with dielectric values corresponding to the test sample oils for carrying out the simulation.The sample holder is selected to be 5 mL and the oil sample used in Section V is 3 mL for all the cases (as from analysis of Fig. 14).
The tan δ indicates loss tangent.Fig. 19 plots to demonstrate the effect of increasing loss in the oil samples and how that effect would determine the resultant Q factor.Quite obviously, as tan δ will increase, Q will decrease, the extent to which Q will deteriorate when loss in the sample increases can be visualized.
As the KOH included loss is difficult to include in simulation, the idea is demonstrated with the Q factor and tan δ plot.It is observed from Fig. 19 that the value of tan δ is varied from 0.05 to 0.35, and as the value of the loss tangent increases, there occurs a decrease in the Q-factor and a lower magnitude level of the transmission parameter.

III. POWER DISSIPATION ANALYSIS FOR
THE SENSOR PROTOTYPE As shown in Fig. 5, the lumped resistor is basically integrated on the top side of the microstrip line just above the CSRR element to enhance the overall Q-factor of the designed sensor and to substantially increase the S 21 dip level near resonance.Now, the inclusion of this lumped resistor would also result in some resistive losses.In this section, a detailed analysis is carried out to closely study the performance of the designed sensor after including the resistor on the top side of the microstrip line.For this purpose, the CSRR sensor is first integrated with the resistor having a small value of 0.1 , and the S 21 notch level of the sensor is found to be −23 dB, as shown in Fig. 4(a).
From the trend in S 21 plots, it can be observed that the notch level of S 21 increases for the resistor value below and above 10 .It basically means that the S 21 level is not entirely due to resistive losses because, in that situation, the S 21 notch level would continue to decrease with the increasing value of resistance.Hence, the transmission dip observed in Fig. 4(b) may not be directly related to the power dissipation or ohmic losses.Moreover, any addition of significant resistive losses would usually result in a lower value of the Q-factor.It is, however, observed here that for the resistor value of 10 , the overall Q-factor is much higher compared to that of the conventional CSRR sensor where the resistor value is zero as the top microstrip line is continuous.
To analyze the effect of resistive losses further, a detailed power analysis is carried out in Fig. 20

corresponding to
Authorized licensed use limited to the terms of the applicable license agreement with IEEE.Restrictions apply.the present scenario when a lumped resistor of 10 is embedded on the top microstrip line.These simulations are carried out in CST MWS, and hence, no approximations are involved.The power analysis of all the components and the parameters are carried out and it is found that the power dissipation of the lumped element is only 7% at our designed frequency of 1.83 GHz.Finally, it is to be noted that the resistor used in the proposed concept is an SMD resistor from the RC0603JR-0710RL package, whose maximum power dissipation is described to be 0.1 W [34].
From the simulation, the power is 0.07 W, which is well below the rated power rating.As a result, we are not deploying the resistor against its dissipation and the entire structure with its operation is well below the maximum limiting values.

IV. AV MEASUREMENT PROCEDURE
AND DISCUSSION Cooking oil can accumulate a significant amount of acid when it is subjected to one of the external conditions, i.e., oxidation process, prolonged high temperature (ester thermolysis), or exposure to acids and bases (acid/base ester hydrolysis).
AV and Its Significance: AV (or neutralization number or acid number or acidity) is the mass of KOH in milligrams that is required to neutralize 1 g of fat, oil, and so on.The significance of the AV of oil lies in the fact that the higher the number of AVs, the higher the rancidity of oils and fats.
The designed sensor is fabricated in a 1.6-mm Taconic TLY-5 substrate (ε r = 2.2) where the measured and simulated results are shown in Fig. 21.The measured quality factor is around 915 from Table II obtained from the 3-dB bandwidth of the transmission notch.
For sample testing, oil samples are collected from different commercial food stalls that usually prepare the eateries by reusing edible oils more than once.The cooked oils undergo transesterification, which in turn increases their AV.
The measured results indicate that the resonant frequency undergoes a slight shift to 1.75 GHz because, during the placing of the SMD resistor, the etching of the gap alters the capacitive effect.The volume of the sample holder is 5 mL of which 3 mL of each oil sample is tested.The volume of the oil has a negligible effect in the measurement data verified from 1 to 5 mL of oil sample, and hence, the average value is kept.The sensor after each measurement is rubbed and polished with acetone with precision and then used for taking another measurement.The measurements are repeated three times for each sample and then averaged out to obtain the measurement results.The sample holder is made of plastic and the bottom of the holder is cut so that the sensing region is kept bare to the oil sample poured into the region.As a result, the sample holder introduces no change to the S 21 response.The role of the sample holder is only to contain the test sample as the test sample directly touches the sensor region.
The measurement setup to detect the AV of oil samples in terms of f and quality factor (Q) [see ( 15) and ( 16)] is shown in Fig. 22 using the Agilent vector network analyzer.For this, a plot between three samples is taken (Fig. 23) as follows.
3) Refined oil cooked multiple times is recorded.It is to be noted that all the oil samples are first cooled at room temperature before taking the measurement.
From Figs. 18 and 20, it is also clear that due to the sufficient sensitivity and high Q of the sensor, it is clearly able to detect the difference in properties for samples (I), (II), and (III).
With each frequency shift, there will also be degradation in the quality factor because an increased AV increases the loss content in the oil specimen and the purity diminishes concerning a fresh oil sample.Thus, as a result, the damping in the resonator increases.Therefore, there is a subsequent frequency shift as well as a drop in the quality factor as the oil gets reused more and more.To analyze the shift in frequency and quality factor, the measured set of data is imported to MATLAB in CSV format and an algorithm is developed to study their Q-factor from 3-dB bandwidth.The algorithm for the same is shown in Fig. 24 and it is repeated for the different sets of oil samples.
The quality factor turns out to be minimum for the oil that is being used multiple times and maximum for the fresh oil, as shown in Table III.
The measured results using the proposed topology are tabulated in Table III.After the determination of the frequency shift from the oil samples, a conventional experiment is carried out to establish a relationship between the frequency shift/Q-factor obtained from the passive microwave sensor and the absolute determination of the AV of the obtained refined oil sample.
The developed sensor due to its high-quality factor and sensitivity is able to determine the very small changes in the resonant parameters of the device when there occurs a unity change in the AV of the sample.The sensitivity thus becomes .
The proposed sensor also provides a substantially high sensitivity of 25 MHz/mgKOHgm −1 in terms of shift in resonant frequency and 3.33 dB/mgKOHgm −1 in terms of the S 21 notch.The procedure included the mixing of oil samples (I)-(III) with Isopropyl Alcohol (IPA) as a solvent to determine the solubility of oil in a water bath [Fig.26(a)] for about 20-30 min.Exactly 15 g of KOH [Fig.26(b)] is weighed, which is to be used for the titration.After the oil is completely dissolved, the titration [Fig.26(c)] is carried out with a 15% KOH (1.335 N) solution using phenolphthalein as an endpoint indicator using a magnetic bead Authorized licensed use limited to the terms of the applicable license agreement with IEEE.Restrictions apply.IV.
After the experiment, it was verified that the oil sample with multiple heating and cooking had a much higher AV (9-mg KOH/gm of oil) than the fresh oil (1.5-mg KOH/gm of oil).This explains that the amount of KOH (0.6 cc) required to neutralize 1 g of fat is much higher than the amount of KOH needed to neutralize fresh oil.It is to be noted that all The analysis explains that with the increase in AV of the oil sample, both the S 21 notch and the frequency decrease as obtained from the equations of analyses ( 15) and ( 16).The plot of AVs against resonant frequency shift and S 21 notch is shown in Fig. 28.A comparison of the proposed sensor with previously reported works is tabulated in Table V.
The sensitivity in this work is referred to as the change in S-parameter with the change in the dielectric parameter.
Authorized licensed use limited to the terms of the applicable license agreement with IEEE.Restrictions apply.

V. CONCLUSION
An alternative approach based on the microwave resonator method utilizing the modified form of CSRR topology has been proposed for the quality assessment of cooking oils.The proposed microwave sensor has been designed, fabricated, and tested to check the toxicity of reused oils by determining their AVs.For this purpose, an empirical formula has been established using the equivalent circuit model to find the relationship between the AV and the measured microwave parameters such as the resonance frequency and the S 21 level.An empirical formula has later been established between the acidic values of oil and the frequency shift and different in S 21 notch level of the proposed sensor.The proposed scheme can quite conveniently be employed to test the purity of edible oils without the involvement of complex systems and procedures.It appears that the reused oil can be utilized for various environmentally sustainable products such as biodiesel and organic hand sanitizers [33].

Fig. 1 .
Fig. 1.Schematic of the proposed concept for quality check of recycled edible oil.* As per the latest reports by the National News Channel on National Nutritive Week 2021.

Fig. 3 .
Fig. 3. Short and open configurations of the host microstrip feed line.

Fig. 5 .
Fig. 5. Top microstrip with resistor (R) of 0603 packages and CSRR etched on the ground plane with a slot dimension of 0.2 mm with dimensions w = 4.37 mm, c = 0.25 mm, rin = 5.35 mm, rout = 5.55 mm, d = 12.5 mm, and P 1 and P 2 are Port 1 and 2, respectively.

Fig. 6 .
Fig. 6.Schematic view of the slot length and resistor dimension.
in terms of its S-parameter, relating incident [a], and reflected [b] waves [b] = [S][a].For incident voltage waves and reflected voltage, the transmission coefficient is usually represented as

Fig. 7 .
Fig. 7. Schematic illustration of the electric circuit model of the structure's cross section.(a) Simplified electrical model diagram.(b) Corresponding star configuration.

Fig. 9 .
Fig. 9. S 21 plot of the CST EM simulation and ADS circuit simulation.

Fig. 12 .
Fig. 12. Concept showing the behavior of the proposed sensor upon resistance integration from a cavity point of view.

Fig. 14 .
Fig. 14.Field strength analysis for sample holder and sample volume having a maximum field strength of −6.6 dB max at 5 mm with respect to the z-axis.

Fig. 15 .
Fig. 15.Field strength analysis to choose optimum dimension for sample and sample holder.(a) Cut-plane field at 5 mm.(b) 6 mm along the z-direction.

Fig. 16 .
Fig. 16.Positional analysis to determine the optimum position for the feed resistor placement where W g = 0.3 mm.

Fig. 18 .
Fig. 18.Three-dimensional view of the sample placed over the resistor incorporated sensor.

Fig. 19 .
Fig. 19.(a) Analysis of transmission magnitude and quality factor versus loss tangent.(b) Response of the transmission magnitude with varying loss tangent for ε r = 2.

Fig. 20 .
Fig. 20.Power analysis for the proposed sensor at the resonant frequency.

Fig. 21 .
Fig. 21.Measured and simulated results of the proposed sensor topology.

Fig. 22 .
Fig. 22. Fabricated sensor: (a) top view and bottom view and (b) measurement setup of the proposed sensor system.
Fig. 25 then shows the measurement analysis of three different oil samples, mustard oil, refined oil, and sunflower oil with close dielectric values.The plot trend indicates that as the oil is cooked multiple times, the S 21 notch hovers toward the upper frequency regime between −10 and −15 dB.The standard deviation for all four cases is 0, 0.05, 0.03, 0.03, and 0.025.

Fig. 25 .
Fig. 25.Measurement results of three different oil samples show variation for four different cooked conditions.

Fig. 26 .
Fig. 26.Experimental verification to obtain the acid value of the refined oil samples selected (a) solubility check of oil sample using IPA solvent in a water bath, (b) 15 g of KOH (1.335 N) taken for titration, (c) titration procedure of the KOH against dissolved oil samples using phenolphthalein as an endpoint indicator, and (d) automated stirring using magnetic bead stirrer.

Fig. 27 .
Fig.27.Flowchart of the chemical method to determine the acid value and the corresponding formula used (inset) table for frequency shift and acid value determination of three samples.

Fig. 28 .× 1
Fig. 28.Analysis for the measured response of acid values with normalized frequency shift and quality factor where aa ′ = fresh sample oil, bb ′ = sample cooked once, and cc ′ = sample cooked multiple times.

TABLE I COMPARISON
BETWEEN CONVENTIONAL CSRR AND IMPROVED SENSOR

TABLE II LUMPED
ELEMENT VALUES OF THE PROPOSED EQUIVALENT CIRCUIT

TABLE III MEASUREMENT
RESULTS OF THE LOADED AND UNLOADED SENSOR STRUCTURE

TABLE V COMPARISON
OF THE PROPOSED SENSOR WITH ACTIVE AND PASSIVE COUNTERPARTS