Near-field microwave sensing technology enhanced with machine learning for the non-destructive evaluation of packaged food and beverage products

This part is a complete description of the antenna design and operation. It consists of the description of the design parameters, the impedance matching efficiency, and the experimental near-field evaluation. Furthermore, we consider the functionality of detecting contaminants utilizing the electrical area (E-field) spatial protection evaluation and the time area reflectometry.Antenna design parametersFigure 3(a) Front view of the design, (b) rear view of the design, (c) entrance view after fabrication, (d) rear view after fabrication. Red labels point out the following: L1 and L2 signify the width and size of the antenna’s transmission line, whereas R1 and R2 denote the longer and shorter radii of the 4 ellipses comprising the antenna construction. Additionally, L3 and L4 correspond to the dimensions of the V-cut triangular form in the floor aircraft.Selecting the acceptable antenna sort is pivotal in designing an efficient MW sensing system. Exploring the position of antennas on this context entails contemplating components reminiscent of bandwidth, polarization, radiation sample, and achieve; nevertheless, given our deal with making use of the sensing approach in the near-field area, our concern lies not on the achieve. As said in the system description part, we’re utilizing a PCB-printed monopole antenna with a flower form, derived from32. Omni-directional antenna has a large angular protection, enabling it to emit and acquire the retrieved scattering parameters from all instructions. This functionality facilitates the detection of contaminants at completely different positions inside the jar. The simplicity and omnidirectional nature of the monopole antennas make them well-suited for close-range purposes, nevertheless, acquiring a large bandwidth antenna is a problem in phrases of radiation energy stability throughout the complete bandwidth, impedance matching, and electrical dimension constraints. Taking into consideration all these parameters, we optimize the antenna to adapt to our particular software necessities, specializing in attaining the desired stability in the radiation sample, the required bandwidth, and the impedance matching. The antenna is designed in accordance with the following requirements: the metallic flower construction of the antenna and the floor aircraft have a thickness of (0.03,textual content {mm}) and are printed on a regular substrate of FR-4 ((epsilon _r = 4.1)) with a thickness of (1.55,textual content {mm}). The antenna primarily consists of 4 ellipses as proven in Fig. 3, every with a bigger radius (R1) of (24,textual content {mm}) and a smaller radius (R2) measuring (12,textual content {mm}). The major ellipse is aligned with the transmission line at a zero-degree angle relative to the line. Two further ellipses contribute to forming the flower-shaped construction by being rotated +45(^circ) and (-45^circ) with respect to each the transmission line and the centrally positioned ellipse. The fourth ellipse is aligned horizontally and is rotated 90(^circ) with respect to the transmission line. The general dimension of the antenna is (50times 45 ,textual content {mm}^2). L1 and L2 signify the width and size of the transmission line, with values of (3,textual content {mm}) and (15,textual content {mm}), respectively. The floor aircraft has a semi-ellipsoidal form with an optimum triangular lower, as illustrated in Fig. 3b. L3 and L4 are equal to (3,textual content {mm}) and (5.2,textual content {mm}) respectively and denote the optimum dimensions of the V-cut, that are decided for the finest efficiency in phrases of the impedance matching in the frequency band of curiosity.The restrict of the near-field area is set by$$start{aligned} hspace{65mm} textual content {r} < frac{2text {D}^2}{lambda }, finish{aligned}$$ (2) the place (textual content {D} = 6.7,textual content {cm}) is the most dimension of the antenna (the diagonal), and (lambda) is the wavelength. In the chosen frequency bands, we compute the near-field radiating radius (r) utilizing Eq. (2) at (1.5,textual content {GHz}) and (10,textual content {GHz}), leading to values of (4.5,textual content {cm}) and (30,textual content {cm}) respectively. In our setup, the antenna is about (5,textual content {cm}) away from the jar. This implies that our measurements are performed in the near-field area for oil-based products in the 6 to (10,textual content {GHz}) bandwidth. While for water-based products, at decrease frequencies inside the 1.5 to (3.5,textual content {GHz}) bandwidth, the jar is positioned at the edge of the near-field area. However, at the increased frequencies on this bandwidth, it falls inside the near-field area. For a complete understanding of the efficiency of the antenna array parts built-in with our sensing system, we conduct simulations and measurements to acquire the near-field. The upcoming subsections present an intensive evaluation and evaluation of the propagation of the E-field radiated by the antenna in the area of curiosity.Impedance matching and near-field measurementsThe near-field measurement system, which can also be used for impedance matching evaluation, is proven in Fig. 4. It contains the community analyzer, the transmitting antenna employed in our research, and the receiving antenna, which is a wide-band horn antenna working ranging from (2,textual content {GHz}). For the near-field measurements and later for the time-domain evaluation, that are helpful to totally perceive and analyze the addressed downside, we conduct measurements and simulations throughout a broad bandwidth starting from 2 to (12,textual content {GHz}).Figure 4The system for near-field measurements. (a) The measurement setup shows the transmitting antenna, denoted as 1, which is the flower antenna utilized in the setup. (b) The measurement setup additionally options the receiving wideband horn antenna, labeled as 2.For the impedance matching evaluation, we measure the reflection coefficients at the antenna port. We examine three eventualities which are reported in Fig. 5. The first one shows the impedance matching of the antenna in the absence of any jar (blue curve). This plot shows good matching efficiency in the bandwidths of curiosity besides at the lowest frequencies of the decrease bandwidth, as highlighted in inexperienced and orange shaded areas in Fig. 5. However, the actual system operates in the presence of a jar, which is why we examine two further eventualities, one with a jar stuffed with oil (inexperienced curve), and the different with a jar stuffed with water (crimson curve). The three curves are virtually superimposed, exhibiting that the presence of the jar doesn't considerably have an effect on the impedance matching. The antenna is general effectively matched, besides beneath 2.5 GHz. However, mismatching doesn't imply that no energy is transmitted into the jar particularly for water-based products, whereas the penetration depth is 6 cm at 2 GHz (from Fig. 2c). So it's too early to exclude the frequencies between 1.5 and 2.5 GHz at this stage, however we are going to examine this frequency band throughout the evaluation of the experiments and the classification outcomes.Next, we measure the near-field of the antenna. For this, we make use of the mechanical stepper motor linked to the receiving horn antenna for goal scanning. The two antennas are centered with respect to one another and positioned (17,textual content {cm}) aside. The horn antenna captures the radiated waves inside a scanning space that extends (20,textual content {cm}) in the vertical y-direction and (25,textual content {cm}) in the horizontal x-direction with an incremental step of (2.5,textual content {mm}) in each instructions. Since we goal to measure the near-field of the antenna, measurements are performed with out the jar. The outcomes are displayed in Fig. 6. The plots illustrate the magnitude of the near-field (normalized to the most magnitude worth throughout all frequency factors), illustrating a discount in area energy with growing frequency, particularly past (8,textual content {GHz}). This might be confirmed by the general system measurements as described in the dataset building part.Figure 5The measured amplitude of reflection coefficients for the realized antenna in three eventualities: with an empty jar, with a jar stuffed with oil, and with a jar stuffed with water.Figure 6The measured amplitude of the antenna’s near-field (normalized to the most magnitude worth throughout all frequency factors) in the absence of any jar at: (a) 2 GHz, (b) 4 GHz, (c) 6 GHz, (d) 8 GHz, and (e) 10 GHz The rectangular form featured in the plot signifies the anticipated location of the food jar inside the measurement system.E-field spatial protectionThe E-field spatial protection may be analyzed as an indicator of how successfully an antenna array illuminates a selected level inside the radiated quantity. In less complicated phrases, it assesses how completely the electromagnetic waves emitted by the antenna array attain and work together with a selected level inside the examined house. For a greater interpretation of E-field spatial protection, we simulate the system in its actual dimensions, as illustrated in Fig. 7. We select the water case as a result of the problem that top losses suggest. The simulations are performed at a frequency equal to (2.5 ,textual content {GHz}), which is the middle frequency in the chosen bandwidth for testing water jar samples. The E-field spatial protection is outlined as follows:$$start{aligned} hspace{55mm} C_E^alpha (textbf{r},f) = frac{sum _{n = 1}^Tmid textbf{E}_n^alpha (textbf{r},f) mid }{max sum _{n = 1}^Tmid textbf{E}_n^alpha (textbf{r},f) mid }, finish{aligned}$$ (3) the place (textbf{r}) is the positioning vector, T is the quantity of antennas, and (textbf{E}_n^alpha) with (alpha = i,t) and s, is the incident, whole, and scattered E-fields, respectively. Each (textbf{E}_n^i) is obtained with the n-th antenna radiating with the jar stuffed with water solely, as a substitute, the (textbf{E}_n^t) is the E-field radiated by the n-th antenna with contaminants inside the stuffed jar. Finally, (textbf{E}_n^s = textbf{E}_n^t - textbf{E}_n^i) represents the E-field scattered by the contaminants. When the n-th antenna is radiating, all the others are matched to 50 (Omega). Hence, (T = 6) completely different simulations are carried out to get the E-field spatial protection with every time a special radiating antenna.Figure 8a exhibits the (C_E^i(textbf{r})) and it's evident that we bought the lowest E-field protection at the backside of the jar. Hence, to contemplate the worst state of affairs, the contaminants, that are 5 plastic spheres with 2 mm in diameter, are positioned in the backside of the jar. Then, Fig. 8b exhibits the (C_E^t(textbf{r})), which is the protection when the contaminants are current: the whole protection seems similar to the incident one. However, the scattered protection proven in Fig. 8c clearly exhibits the presence of the contaminants at the correct location, that are clearly distinguishable.Figure 7The design system used for E-field spatial protection evaluation.Figure 8The E-field spatial protection at 2.5 GHz inside the water jar: (a) incident, (b) whole, and (c) scattered.Time-domain evaluationThe time area reflectometry is often utilized in non-destructive evaluation, exhibiting that the reflection coefficients carry related details about the object underneath take a look at; therefore we use right here to research our simulated and measured information. The goal of analyzing the time-domain response of the reflection coefficient parameters, obtained from each simulations and measurements, is to review the reflections of the sign emanating from the radiated area. This evaluation may be performed by direct time-domain measurements or by first performing a frequency-domain measurement adopted by the Inverse Fast Fourier Transform (IFFT). To establish discontinuities inside the system, a big time-domain commentary window with a small interval is required. The reflection coefficients are measured throughout a frequency vary of (2,textual content {GHz}) to (12,textual content {GHz}) at intervals of (10,textual content {MHz}) for each the measurements and simulations. By evaluating the electrical distance in each measurements and simulations throughout numerous media, alongside with evaluating the distance line ((textual content {d}x)) obtained from the time area evaluation, we efficiently detect the key reflection factors inside our system. These factors embody the antenna’s feeding port, its section middle, the jar stuffed with water, and most vital the contaminant inside the jar.Figure 9The simulation system much like the measurements besides for the presence of the international physique.In Fig. 9 we current the simulated system. The labels (a–f) inside this determine correspond to the following factors: (a) the waveguide port, (b) the feeding port of the antenna, (c) the middle of the antenna construction, (d) the middle of the face of the water jar reverse to the antenna, (e) the place of the contaminant inside the water, and (f) the edge of the jar from the second facet. Four simulations are carried out to investigate the time area habits underneath completely different eventualities. The first simulation is used as a reference, involving a jar stuffed with water with out intrusions. In two of the remaining three simulations, a plastic dice contaminant is launched, every with a special place inside the jar. The final simulation concerned altering the contaminant materials to metallic (copper). The IFFT operation is utilized to the simulated reflection coefficient information, and the distance line, (textual content {d}x), is calculated in accordance with:$$start{aligned} hspace{60mm} textual content {d}x = frac{ccdot textual content {d}t}{2},quad textual content {d}t = frac{1}{textual content {d}f}, finish{aligned}$$ (4) the place (textual content {d}x) represents the distance line in centimeters, (c) is the velocity of gentle in free house, (textual content {d}t) is the time interval between consecutive factors in the time area, and (textual content {d}f) is the frequency decision.For extra correct calculations, you will need to outline the vary decision as the antenna’s functionality to tell apart between carefully spaced targets. Discussing vary decision is essential for addressing discontinuities in the antenna and the system. For instance, to detect these reflection factors precisely, the vary decision should be smaller than the distance between any two consecutive factors in Fig. 9. Therefore, we have to take into account the worth of the vary decision in our calculations in figuring out the positions of these discontinuity factors. The vary decision is outlined as follows:$$start{aligned} hspace{65mm} Delta {textual content {R}} = frac{c}{2cdot textual content {BW}}, finish{aligned}$$ (5) the place (Delta {textual content {R}}) is the vary decision, and (textual content {BW} = 10) GHz is the bandwidth of the performed simulations and measurements. According to Eq. (5), the vary decision is calculated to be 1.5 cm.Figure 10Table 2 Analysis of the distance (Delta X) as illustrated in Fig. 10.Figure 11The amplitude of the IFFT outcomes for the measured reflection coefficients.Figure 10a exhibits the IFFT outcomes for the simulated reflection coefficients obtained at level a in Fig. 9. The labels (1), (2), and (3) inside this subfigure correspond to the reflection factors b, c, and e indicated in Fig. 9, respectively. The positions of these peaks on the IFFT plots are decided by calculating the electrical distance by multiplying the geometrical distance (obtained from the simulated design or the measurement system) by the sq. root of the relative permittivity of the propagating medium, as proven in Table 2. This permits us to discern which half of the performed simulation or actual measurements corresponds to every peak in the IFFT plot. In Fig. 10a, we're unable to detect the presence of the contaminant inside the jar, as the plots are superimposed with the reference curve (the one with out the contaminant). To handle this problem, we subtract the recorded reflection coefficients simulated when the intrusion is current from that of the reference simulation. Then, we carry out the IFFT on the remaining values to see if we are able to establish the positions of the contaminants alongside the distance line. Figure 10b shows the outcomes of this strategy, and it's evident that the peak obtained at (x = 46.5,textual content {cm}) is a consequence of the contaminant presence inside the water. This peak is notably extra pronounced when the contaminant is metallic. The displacement of the contaminant inside the jar influences the location of the reflection level indicated by a peak at (x = 61.4,textual content {cm}), whereas the peak at (x = 77.9,textual content {cm}) arises from the reflection of the rear facet of the jar. This peak seems because of this of the presence of the contaminant inside the water, influencing the propagation path of the wave inside the medium. This deviation from the reference state of affairs, the place no contaminant is current inside the jar, causes the noticed distinction at the third peak in Fig. 10b.Comparing the information in the last two columns of Table 2 reveals an excellent coherency between the theoretically estimated reflection factors positions and these evident in the IFFT plot illustrated in Fig. 10a, b, taking into consideration the vary decision. The error noticed in the ((Delta X_{c,d})) may be attributed to the assumed precise place of the antenna’s section middle at level c, as depicted in Fig. 9.To additional examine the system’s capacity to detect adjustments in the reflection coefficients based mostly on the jar contents by measurements, we report the antenna’s reflection coefficient parameters utilizing the identical system beforehand employed in the near-field research (proven in Fig. 4). We discover three distinct eventualities: one with none jar, one other with a jar stuffed with oil, and a 3rd with the identical jar stuffed with water, however this time with out the want for scanning. In Fig. 11, the labels displayed on the curve signify the estimated reflection factors. The peaks marked with (1) and (2) correspond to the feeding port and the section middle of the antenna, respectively. The three curves overlap till reaching the peak at level (3), which signifies the reflection level originating from each the assist holding the jar and the jar itself, as proven in Fig. 4. The reflections at this level distinctly reveal the affect of the complete medium inside the jar on the reflection depth. Specifically, it's extra pronounced when the jar is stuffed with water (the crimson curve) in comparison with when it incorporates oil (the inexperienced curve) or the state of affairs the place reflections are solely as a result of the assist with out the presence of a jar (the blue curve). It is noteworthy to say the consistency of the three highlighted peaks (1), (2), and (3) in each plots, the simulation in Fig. 10a, and the measurements in Fig. 11. However, the place of these factors is shifted by the size of the coaxial cable ((Delta X_{a,b})) in Fig. 10a.
https://www.nature.com/articles/s41598-024-62287-6