- Email: [email protected]
Flow-through sensor based on piezoelectric MEMS resonator for the in-line monitoring of wine fermentation
Sensors and Actuators B 254 (2018) 291–298
Contents lists available at ScienceDirect
Sensors and Actuators B: Chemical journal homepage: www.elsevier.com/locate/snb
Flow-through sensor based on piezoelectric MEMS resonator for the in-line monitoring of wine fermentation J. Toledo a,∗ , V. Ruiz-Díez a , G. Pfusterschmied b , U. Schmid b , J.L. Sánchez-Rojas a a b
Microsystems, Actuators and Sensors Group, Universidad de Castilla-La Mancha, Ciudad Real, Spain Institute of Sensor and Actuator Systems, TU Wien, Vienna, Austria
a r t i c l e
i n f o
Article history: Received 18 January 2017 Received in revised form 13 July 2017 Accepted 14 July 2017 Available online 17 July 2017 Keywords: Microresonators Piezoelectric Oscillator circuit Density In-line Grape must fermentation
a b s t r a c t The traditional procedure followed by winemakers for monitoring grape must fermentation is not automated, has not enough accuracy or has only been tested in discrete must samples. In order to contribute to the automation and improvement of the wine fermentation process, we have designed an AlNbased piezoelectric microresonator, serving as a density sensor, resonantly excited in the 4th-order roof tile-shaped vibration mode. Furthermore, conditioning circuits were designed to convert the one-port impedance of the resonator into a resonant two-port transfer function. This allowed us to design a Phase Locked Loop-based oscillator circuit, implemented with a commercial lock-in amplifier with an oscillation frequency determined by the resonance mode. We measured the fermentation kinetics by simultaneously tracking the resonance frequency and the quality factor of the microresonator. The device was first calibrated with an artificial model solution of grape must and then applied for the in-line monitoring of real grape must fermentation. Our results demonstrate the high potential of MEMS resonators to detect the decrease in sugar and the increase in ethanol concentrations during the grape must fermentation with a resolution of 1 mg/ml and 20 Pa s as upper limits for the density and viscosity, respectively. © 2017 Elsevier B.V. All rights reserved.
1. Introduction The fermentation of grape must involves the interaction between yeasts, bacteria, fungi and viruses. A correct bio-chemical process is a necessary condition but not sufficient by itself to determine the final quality of the wine, whose assessment relies on a comprehensive analysis of its chemical components, as the basic flavour of wine depends on at least 20 compounds [1]. Therefore, winemakers must carefully supervise the wine fermentation process to ensure a wine of the expected quality. One of the key parameters monitored during wine fermentation is the fermentation kinetics. This provides essential information about the steady transformation of grape must into wine due to the decrease of glucose and fructose that leads to the formation of ethanol, glycerol and carbon dioxide along with biomass, as a result of yeast metabolism [2]. This process is traditionally monitored by enologists, who manually extract and analyse discrete samples at least twice a day using instruments such as the aerometer, spectrophotometer or colorimeter. They essentially determine the density and its rate of change since these parameters provide informa-
∗ Corresponding author. E-mail address: [email protected] (J. Toledo). http://dx.doi.org/10.1016/j.snb.2017.07.096 0925-4005/© 2017 Elsevier B.V. All rights reserved.
tion about the evolution of the fermentation as a result of yeast metabolism. In an industrial fermentation process, temperature fluctuations would influence to some extent the fermentation process and thus the density. Nevertheless, it is the variations that help enologists determine the current fermentation status, and whether or not corrective measures are required as it progresses. Besides, a variety of alternative procedures have been used for monitoring the grape must fermentation such as density determination based either on differential pressure measurements [3] or on the use of flexural oscillators [4], monitoring of CO2 released during the process due to the gradual loss of mass [3,5], determination of yeast cell population evolution by means of impedance techniques [6] and turbidity measurements [7], ultrasound measurements conducted to determine the propagation velocity in grape musts [8,9], refractive techniques based on fibre optics [10], optoelectronic device based on measurements of the refractive index [11,12] and an off-line monitoring based on a piezoelectric resonator [13]. Overall, these different approaches have not enough resolution or have only been tested in discrete must samples. In this context, the simultaneous determination of the density and viscosity of liquids, through measurement of the resonant frequency (fr ) and quality factor (Q-factor) of a mechanical resonator, has already been reported [14–16]. This approach presents several advantages with
292
J. Toledo et al. / Sensors and Actuators B 254 (2018) 291–298
Fig. 1. (a) Top-view micrograph of the resonator. (b) Modal shape measured (15-mode) with a laser Doppler vibrometer.
respect to traditional methods, such as real-time analysis, on-line configuration and low liquid volumes. Piezoelectric resonators with in-plane vibration modes [17], flexural and torsional modes [18] have already been tested. However, it is well-established that a different family, such as the roof tile-shape modes, present better quality factors at moderate frequencies [19,20]. The present work evaluates the performance of this vibration mode, within micromachined self-actuated and self-sensing aluminium nitride (AlN)-based cantilever sensor for the in-line monitoring of grape must fermentation in flow-through configuration. 2. Material and methods 2.1. Resonator design and characterization The resonator was designed for the 4th order roof-tile shaped vibration mode. It resulted in a cantilever resonator with optimized electrode layout featuring five nodal lines in one direction, and 1 nodal line in the perpendicular direction. Considering Leissa’s nomenclature, the vibration mode is named as 15-mode [21]. The resonator was designed with a length of L = 2524 m, width of W = 1274 m and thickness T of 20 m. It was fabricated from a SOI wafer with a 20 m-thick device layer covered with a 650 nm-thin AlN piezoelectric film [22]. The top metallization has four striped electrodes that allow a selective excitation of the vibration modes and act as a filter for higher modes [18]. This device has already been tested in a previous work obtaining a quality factor of 116 in 2-Propanol [23]. The considered modal shape and the fabricated resonator are shown in Fig. 1 [13].
Fig. 2. Open loop response for the 15-mode measured in a liquid test (2-Propanol) with and without the dummy compensation.
2.2. PLL-based oscillator The simultaneous determination of density and viscosity of liquids, through measurement of the resonant frequency and the quality factor of a mechanical resonator, is challenging due to the low quality factors and parasitic effects present in liquid media [24]. For this reason, our approach is based on roof tile-shaped resonators in two-port configuration and a parasitic compensating device. In the two-port scheme, one of the electrodes was used for actuation (+) and the other for sensing (−) (see Fig. 1). However, our results show that a capacitive crosstalk (Cft ) across the actuation and sensing ports has a significant contribution to the output current. To minimize this parasitic component, a compensating device was introduced. This dummy device reproduces the structure of the resonator, but without the backside release, thus preventing any vibration. For this reason, we designed an interface circuit, used in a previous work [16], to subtract the dummy response from that of the resonator. Since the materials and dimensions in the
Fig. 3. Schematics of the PLL-based oscillator and the interface circuit.
dummy and resonator devices are the same, they are expected to show identical electrical behaviour with respect to their parasitic effects. This results in a clear resonance, with low baseline (see Fig. 2) and remarkable phase transition [13]. However, a phase shift (about −12◦ ) is introduced by the instrumentation amplifier (Fig. 2) at the maximum gain, representing the differences in electrical performance between the dummy, the resonator device and Cft . In order to compensate this phase shift, the resonator and interface circuit were finally included in an oscillator based on a Phase Locked Loop (PLL) instrument (Fig. 3), instead of an oscillator circuit based on discrete components [16]. The phase compensation
J. Toledo et al. / Sensors and Actuators B 254 (2018) 291–298 Table 1 Composition of model solutions representing a normal fermentation process (N1 :N9 ).
293
Table 2 Composition of model solutions representing a sluggish fermentation process (S1 :S9 ).
Solution
Fructose [g/l]
Glucose [g/L]
Glycerol [g/L]
Ethanol [% v/v]
Solution
Fructose [g/l]
Glucose [g/L]
Glycerol [g/L]
Ethanol [% v/v]
N1 N2 N3 N4 N5 N6 N7 N8 N9
110 90 70 60 40 20 8 5 2
100 80 30 20 10 2 2 2 1
0 0 5 5 6 7 7 7 9
0 1 6 8 9 12 13 13 14
S1 S2 S3 S4 S5 S6 S7 S8 S9
110 90 80 70 66 64 62 58 57
100 80 60 30 27 26 24 22 21
0 0 3 5 5 5 6 6 7
0 1 3 6 7 7 7 7 8
allowed to make the oscillation frequency approximately equal to the natural frequency of the resonator. The PLL instrument used (Zurich HF2LI) provides software-based voltage controlled oscillator (VCO), proportional-integral controller (PI) and phase detector (PD) blocks that form a control system able to track the resonant frequency when the resonator is affected by the surrounding medium. The configuration parameters used for each of the blocks were the proportional gain, Kp = 2.63 Hz/deg, time constant of the PI controller, tc = 16.84 ms, and the bandwidth of the system, BW = 100 Hz.
Table 3 Values of the coefficients obtained in the calibration process for the normal model solution.
Normal model solution
Kic [s]
C1 [m]
C2 []
C3 [m2 ]
C4 [m]
3.58 × 10−10
0.12
−20.99
2.11
−103.22
3. Results and discussion 3.1. Sensor calibration With the goal of estimating the density () and viscosity (), during the fermentation process, two variables were measured from our interface circuit: the oscillation frequency (fosc ) and the gain of the interface circuit module at operation (Gosc ), given by the ratio Vout /Vin . Furthermore, for the application of the resonator as a density-viscosity sensor, a calibration procedure was developed with a model solution of artificial grape must [20] and finally applied to the monitoring of one model sluggish fermentation and other real grape must fermentation process. This calibration process was carried out in two steps, each with adjustable parameters [16].
3.1.1. Preparation of the model solutions of grape must In order to confirm that piezoelectric resonators are valid for the monitoring of grape must fermentation, two different sets of model solutions were used for pre-investigation purposes. The first set represents a normal fermentation (N1 :N9 ) and the second represents a sluggish fermentation (S1 :S9 ). A variety of reasons may be attributable to this kind of failed or sluggish fermentations, which occasionally occur: the lack of dissolved oxygen, an unbalanced ratio between sugar and nitrogen, a low fermentation temperature, an inadequate rehydration, a thermal shock of yeast, etc. [25]. These model solutions represent different stages of the corresponding fermentation process according to their particular mixture of glucose, fructose, ethanol and glycerol dissolved in water. The composition of the prepared model solutions for a normal and sluggish fermentation is shown in Tables 1 and 2 [11]. In the normal model solution, a decrease in fructose and glucose concentration from 110 & 100 g/l (N1 ) to 2 & 1 g/l (N9 ) occurs. Similarly, the glycerol and ethanol concentration are increasing from zero to 9 g/l and 14% v/v. In the sluggish fermentation the decrease in fructose and glucose concentration stops prematurely and settles around 57 and 21 g/l. Along with this development the increase in the ethanol concentration stops and does not exceed the value of 8% v/v.
Fig. 4. Resonant frequency and quality factor of the resonator immersed in two model solutions of grape must at 20 ◦ C.
3.1.2. Calibration process with model solutions of grape must In the first step of the calibration process the circuit outputs fosc and Gosc were transformed into the key parameters of the resonator, namely the resonant frequency (fr ) and the quality factor (Q-factor). Thanks to the phase shift compensation provided by the PLL, fr is obtained as the oscillation frequency for each liquid. On the other hand, Q-factor is calculated from the circuit output gain, Gosc . The constant Kic , which is not affected by the loading conditions of the resonator, was calculated by fitting the values of fr , Gosc and the experimental data obtained for different liquids [23] (see Table 3). Gosc = Kic Q 2fr
(1)
Due to the high sensitivity of the density and viscosity to the temperature and electronic noise from surrounding equipment at different frequencies, low, but measurable drift effects are present when determining fr and Q-factor. The random fluctuation of fr was obtained directly through the Allan deviation of fosc [26]. However, the random fluctuation of Gosc was translated into the corresponding fluctuation of Q-factor by error propagation in Expression (1) [27]. These fluctuations determine the ultimate resolutions obtained with the sensor; for the model solutions of grape must they were below 100 mHz for fr and 0.7 for Q-factor. Fig. 4 shows the mean values of fr and Q-factor when the resonator is immersed in the model solutions of grape must: normal and slug-
294
J. Toledo et al. / Sensors and Actuators B 254 (2018) 291–298
Table 4 Density and viscosity values estimated from our resonator (MEMS) and measured with the commercial density-viscosity meter (viscometer) at 20 ◦ C for model solutions of grape must and the calibration error. Normal Fermentation Solution
Viscometer values
Sluggish Fermentation MEMS values
Calibration error
[mPa s]
[g/ml]
[mPa s]
1.949 1.974 1.722 1.707 1.770 1.754 1.770 1.809 1.742
1.082 1.063 1.028 1.017 1.003 0.991 0.985 0.984 0.983
2.274 1.090 2.054 1.069 1.797 1.033 1.880 1.018 1.604 1.002 1.598 0.987 1.752 0.986 1.662 0.979 1.679 0.977 Mean error [%]
1 2 3 4 5 6 7 8 9
[g/ml]
fr =
1−
1+
Q =
2
g1 = C1
1+
1 2Q 2
Lg1 m
[mPa s]
[g/ml]
[mPa s]
16.7 4.0 4.3 10.1 9.4 8.9 1.0 8.1 3.6 7.36
0.7 0.6 0.5 0.1 0.1 0.4 0.1 0.5 0.6 0.38
2.032 1.969 1.819 1.724 1.774 1.770 1.676 1.714 1.777
1.081 1.063 1.049 1.029 1.024 1.022 1.021 1.019 1.017
2.073 1.085 2.092 1.070 2.110 1.045 2.154 1.039 1.892 1.029 1.847 1.024 1.832 1.022 1.763 1.019 1.743 1.016 Mean error [%]
(2)
fr + C2 C
4
P [g/ml]
ε [%]
ε [%]
2.0 6.2 16.0 25.0 6.7 4.4 9.3 2.9 2.0 8.3
0.4 0.7 0.3 1.0 0.5 0.2 0.1 0.1 0.1 0.38
to distinguish between ordinary and sluggish fermentations at an early stage of the fermentation process in artificial grape must. 3.1.3. Calibration error and sensor resolution Once the density and viscosity for each liquid are known, both the error associated with the calibration process and the resolution related to the random fluctuations can be determined. The error terms are calculated as the deviation between the values of viscosity and density estimated from our calibration (MEMS) and the values measured with the commercial instrument (viscometer), obtaining a mean error term for the sluggish fermentation around 8.3% for the viscosity and 0.38% for the density. The resolutions were obtained from the measured random fluctuations in fr and Q-factor by Expressions (6) and (7):
=
|
=
∂ |fr ∂fr
|
2
∂ |fr ∂fr
+
2
|
∂ |Q ∂Q
+
|
2
∂ |Q ∂Q
(6)
2 (7)
Where and are the estimated resolutions for density and viscosity, respectively, and the partial derivatives are computed from Expressions (2)–(5). The values obtained for all the model solutions are below 500 g/ml and 17 Pa s, for density and viscosity. The resolutions and the calibration errors obtained are better for density than for viscosity. This occurs because the Q-factor, determined from the measurement of Gosc in the interface circuit, presents a worse resolution than the frequency measurement. 3.2. Monitoring grape must fermentation
fo,vac
√
g2 = C3 +
Calibration error
ε [%]
Lg2 m Lg2 m
MEMS values
ε [%]
gish fermentation process. The measurements of fr and Q-factor were performed with a temperature control unit [16]. Nevertheless, the impact of temperature fluctuations in the resolution of the sensor during the measurement process is minimal compared with the continuous evolution of the real grape must fermentation. As can be seen in Fig. 4, the significant changes in the composition of the model solutions affects the fr and the Q-factor of the resonator [13]. The combination of unfermented sugars and less than expected ethanol concentration, resulted into a flat curve of fr for the sluggish fermentation. In the second step of the calibration procedure, the physical properties of the liquid, density and viscosity, were obtained from the key parameters of the resonator, fr and Q-factor. These are related to the mechanical properties of the resonators, i.e. resonator equivalent mass (m) and natural frequency in vacuum (fo,vac ), and the fluid conditions, i.e. distributed damping associated to the liquid (g1 ) and the distributed equivalent added mass (g2 ), both per unit length (L). A model defined in previous works [28–30] allows for the determination of these variables separately as in Eqs. (2) and (3). Nevertheless, the dependence of g1 and g2 on the viscosity and density of the fluid, as a closed-form analytical expression, is only available for simple geometries and particular cases [31]. For this reason, we used a Taylor series expansion of g1 and g2 in terms of the viscosity, density and their product (see Expressions (4) and (5)) [32,33]. fo,vac
Viscometer values
√
(3)
(4) (5)
fr
The model was solved iteratively to obtain the values of the coefficients C1 , C2 , C3 and C4 (Table 3) that fit best to the experimental data for the normal model solution [13]. The results of the estimated density values using our MEMS device are compared to those measured by the density-viscosity meter Anton Paar DMA4100M (viscometer) at 20 ◦ C in Table 4. As can be seen, an almost linear decrease in the density occurs for the normal model solution and a premature stop in the density decrease for the sluggish solution. These results show the potential
In the previous section, the theoretical background for using the density and viscosity as valid parameters of the fermentation process was checked. In addition, the performance of the resonator was verified and validated with two model solutions of grape must. Nevertheless, our main goal is the monitoring of a real grape must fermentation process. For this reason, we applied the same procedure that was developed for the model solutions. In this case, through the measurements of fosc and Gosc in the interface circuit and using the coefficients Ci obtained for the normal model solution (see Table 3), we were able to monitor the density and the viscosity during the fermentation. The performance and reliability of the sensor in a real grape must were validated by three different fermentation processes and carried out by different procedures. In the first process, designated as off-line, grape must samples were manually extracted from a commercial fermenter and measured with the resonator under static
J. Toledo et al. / Sensors and Actuators B 254 (2018) 291–298
Fig. 5. Resonator wire bonded in a 24-pin DIP (dual inline package) and immersed in a red grape must. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
immersion. In the second set-up, denominated as in-line, a homemade fermenter was implemented. In this case, the grape must was automatically injected into the cell containing the resonator with a peristaltic pump. After each measurement, the sensor was cleaned with water. In both set-ups the measurements were made with the grape must in static conditions and verifying that the resonator characteristics were maintained after each measurement and cleaning step. However, in the third set-up, denominated as flow-through, there was a continuous flow of grape must through the cell with the resonator. In this case, the sensor was not cleaned after each measurement as grape must was continuously circulating. 3.2.1. Off-line monitoring of a grape must fermentation Grape must was completed using 25% (w/v) of red grape skin from Cencibel variety. Afterwards, the must was inoculated with 0.2 g/ml of Saccharomyces cerevisiae strain (UCLM S325, FouldSpringer) previously rehydrated following the supplier’s guidelines. A commercial cylindrical fermenter (mini-Bioreactor Applikon) was filled with 3 L of inoculated must and grape skins, where the temperature was controlled at 28 ◦ C. In the first method, the fermentation monitoring was carried out by means of the extraction and analysis of 7 mL-samples every 5 h approximately during 6 days. After the extraction the samples were centrifuged for one minute at 1000 rpm (Universal 32R Hettich) and kept refrigerated until analysis. In this case, the resonator was immersed in the grape must samples (Fig. 5), cleaned with water and dried with N2 after
295
Fig. 6. Off-line monitoring of the density-viscosity values estimated from our resonator (MEMS) and measured with the commercial density-viscosity meter (viscometer) at 20 ◦ C.
each measurement. Once the measurements were completed, the samples were kept frozen at −20 ◦ C, in order to stop the evolution of the fermentation process. Following the same procedure as in the previous section with the model solutions, the evolution of density and viscosity was measured as shown in Fig. 6. As can be observed, the density follows the expected evolution for a normal fermentation [11,13] and viscosity values follow a rather similar trend. Both parameters decrease with the evolution of the fermentation since the ethanol concentration increases. The viscosity is not a measured variable in the monitoring of a fermentation. However, in our case it is obtained simultaneously with the density providing further information of the process. The error terms and resolutions were calculated as in Section 3.1.3. The mean error term obtained was around 2.4% for the viscosity and 0.3% for the density being the viscosity resolution below 7 Pa s. In the case of the real grape must the obtained density resolution was below 100 g/ml instead of the 1200 g/ml value obtained in some previous works using a high-resolution low-cost optoelectronic instrument [11,12]. By comparing to the measurements with the benchtop instrument, the errors obtained confirmed that the calibration process involving the normal model solutions is also valid for other liquids in a similar range of density and viscosity.
Fig. 7. Schematic of the fermenter and external fluid circuit designed for the in-line monitoring of grape must fermentation process.
296
J. Toledo et al. / Sensors and Actuators B 254 (2018) 291–298
Fig. 8. In-line monitoring of the density-viscosity values estimated from our resonator (MEMS) and measured with the commercial density-viscosity meter (viscometer) at 20 ◦ C.
Fig. 9. Flow-through monitoring of the density-viscosity values estimated from our resonator (MEMS) and measured with the commercial density-viscosity meter (viscometer) at 20 ◦ C.
3.2.2. In-line monitoring of a grape must fermentation In the second method, a home-made fermenter, with a volume of 3 L, and a flow cell for the resonator, were designed and implemented in a fluidic circuit. The fermenter was immersed in a thermostatic bath at 29 ◦ C and connected to a peristaltic pump, which enabled the recirculation of the grape must between the fermenter and the cell resonator (Fig. 7). Two filters, with a pore size of 500 and 100 m, were introduced in the tube sampling the grape must. These filters prevents the circulation of grape skin, clots or sediments through the flow cell. An air pump provides oxygen for the adequate yeast fermentation. The cell flow, fabricated with a 3D printer, enabled the input/output of the grape must and the housing of the electronic package enclosing the resonator. In addition, the cell has an aperture to evacuate the air that enters into the cell from the fermenter. The systematic procedure for the measurements is as follows. First of all, the grape must was injected in the flow cell with the peristaltic pump until its temperature reached 20 ◦ C (around one minute). This temperature stabilization was assisted by a thermoelectric controller in contact with the resonator. The measurement of the circuit outputs, fosc and Gosc , was carried out following the same procedure as in the previous sections obtaining the density and viscosity values of Fig. 8. Once completed, the grape must was extracted and measured with the commercial density-viscosity meter. After that, the valves in Fig. 7 were switched to allow water injection into the cell, cleaning the resonator during one minute. In order to check that the cleaning process was successful and effective, the value of fr and Q-factor was registered before and after each measurement in water, obtaining a value of 560,000 ± 10 Hz and 140 ± 2, respectively. The mean error term obtained was around 2% for the viscosity and 0.6% for the density being the resolution below 1 Pa s and 88 g/ml, respectively.
3.2.3. Flow-through monitoring of a grape must fermentation Once the previous in-line set-up was proved as a valid configuration for the continuous monitoring, a third procedure, designated as flow-through, was carried out. In this scheme, all the measurements were taken in a continuous flow-through circulation of the grape must as the fermentation progresses, without cleaning the resonator with water. Unlike the previous procedures, the circuit outputs, fosc and Gosc , were registered every minute, and density and viscosity deduced with the same calibration constants as before. The measurements are summarized in Fig. 9. In order to check the correct evolution of the process and resonator behaviour, different grape must samples were taken approximately every six hours and measured with the commercial density-viscosity meter. This final procedure presents two main advantages compared to the ones previously described. On the one hand, we developed a real time monitoring system allowing the detection of a sluggish fermentation or any other problem during the whole process span. On the other hand, all the measurements were carried out without any cleaning process. In the flow-through measurements, the mean error term was around 1.55% for the viscosity and 0.23% for the density, being the resolution around 20 Pa s and 1 mg/ml, respectively. These values are similar to those of the off-line set-up, demonstrating the performance of piezoelectric resonators as sensors for a real in-line monitoring of a fermentation process. It is worth to remark that the density and viscosity resolutions were slightly better in the previous cases, probably due to the dynamics of the flow in the continuous flow-through of grape must and the eventual formation of air bubbles in the cell (see Table 5). However, the calibration error is slightly lower than those obtained for the other two measurement schemes. This error was determined from the difference between the data measured with the commercial instrument and those taken with the MEMS sensor every six hours. In addition, we have estimated the average error considering all the measurements taken during the fermenta-
Table 5 Summary of the sensor resolution ( and ) and calibration error (ε and ε). Fermentation process Off-line
In-line
Flow-through
[Pa s]
[g/ml]
ε [%]
ε [%]
[Pa s]
[g/ml]
ε [%]
ε [%]
[Pa s]
[mg/ml]
ε [%]
ε [%]
7
100
2.4
0.3
1
88
2
0.6
20
1
1.55
0.23
J. Toledo et al. / Sensors and Actuators B 254 (2018) 291–298
tion (one per minute). For that, we took the difference between the fitting curve plotted in Fig. 9, obtained with the values of the commercial instrument, and those estimated with the MEMS sensor, resulting an error of 4% for the viscosity and 0.7% for the density. This is similar to the values obtained in the two other fermentation processes and within the interval determined in the initial calibration (see Table 4). 3.2.4. Temperature dependence In the experiments described above, the temperature of the grape must was stabilized to 20 ◦ C ± 0.05 ◦ C as a reference in the evolution of the process, to neutralize the temperature fluctuations outside the fermentation tank and their influence on the density and viscosity. Nevertheless, our sensor has been validated for other temperatures of interest in the fermentation industry. Two grape must samples, representing the beginning and ending of the fermentation process, were analysed at 15 and 30 ◦ C. The density and viscosity values were determined at these temperatures with our MEMS sensor, with the same calibration of Section 3.1, and compared with the commercial density-viscosity meter, obtaining an error lower than 1 and 2% for the density and viscosity, respectively. As this error is similar to that previously obtained in the calibration at 20 ◦ C, we can conclude that the sensor responds accurately to the changes in density or viscosity caused by changes in temperature, in this range. The density values obtained at 15 or 30 ◦ C can be used to extrapolate the density to a reference temperature of 20 ◦ C [34], the common reference in winemaking, obtaining a mean density error of 0.5% with respect to the value directly measured at this temperature. These low errors demonstrate the possibility to replace the temperature control unit by a basic temperature monitoring system, where the density at 20 ◦ C can be inferred at any stage of the process from its value at a different temperature. 4. Conclusion This work presents a novel approach for the monitoring of a grape must fermentation process with a piezoelectric MEMS resonator excited in the 4th-order of the roof tile-shaped mode. Two important parameters of the resonator were measured using a PLL-based oscillator circuit: the quality factor and the resonant frequency. Once these two parameters are known, the viscosity and density of the liquid can be determined. In order to measure these parameters in a real grape must, a calibration procedure of the resonator was performed using model solutions of artificial grape must representing an ordinary (for calibration) and a stuck or sluggish (for validation) fermentation process and a commercial density-viscosity meter. Furthermore, an in-line flow-through monitoring of a grape must fermentation was carried out. Our results demonstrate the high performance of MEMS resonators to detect the decrease in sugar and the increase in ethanol concentrations during the grape must fermentation with a resolution of 1 mg/ml and 20 Pa s for the density and viscosity, respectively. Acknowledgments This work was supported by Spanish Ministerio de Economía y Competitividad project TEC2015-67470-P and the Austrian Research Promotion Agency within the “Austrian COMET-Program” in the frame of K2 XTribology (project no. 849109). Javier Toledo acknowledge financial support from FPI-BES-2013-063743 grant. References [1] Ronald S. Jackson, Wine Science, Principles and Applications, third ed., Elsevier, 2008.
297
[2] R.B. Boulton, V.L. Singleton, L.F. Bisson, R.E. Kunkee, Principles and Practices of Winemaking, Springer, US, 1999, http://dx.doi.org/10.1007/978-1-47576255-6. [3] N. El Haloui, D. Picque, G. Corrieu, Alcoholic fermentation in winemaking: on-line measurement of density and carbon dioxide evolution, J. Food Eng. 8 (1988) 17–30, http://dx.doi.org/10.1016/0260-8774(88)90033-7. [4] K. Koukolitschek, G. Krieg, W. Maier, Verfahren und Vorrichtung zur präzisen Bestimmung der Alkoholkonzentration in Flüssigkeiten, 1991, pp. 4138419. [5] E.T. Nerantzis, P. Tataridis, I.A. Sianoudis, X. Ziani, E. Tegou, Winemaking process engineering: on line fermentation monitoring – sensors and equipment, E-J. Sci. Technol. (2007) 29–36. ˜ C. Torre, B. García, C.E. Carleos, R. Crespo, L.M. Cárcel, [6] M.A. Pérez, R. Muniz, Impedance spectrometry for monitoring alcoholic fermentation kinetics under wine-making industrial conditions, XIX IMEKO World Congr. Fundam. Appl. Metrol. (2009) 2574–2578 http://www.imeko2009.it.pt/Papers/FP 228. pdf. [7] R. Crespo, L.M. Cárcel, M.A. Pérez, I. Nevares, M. del Álamo, Suitable at-line turbidity sensor for wine fermentation supervision, Int. Conf. Food Innov. (2010) 1–4. [8] N. Lamberti, L. Ardia, D. Albanese, M. Di Matteo, An ultrasound technique for monitoring the alcoholic wine fermentation, Ultrasonics 49 (2009) 94–97, http://dx.doi.org/10.1016/j.ultras.2008.06.003. [9] P. Resa, L. Elvira, F.M. De Espinosa, Concentration control in alcoholic fermentation processes from ultrasonic velocity measurements, Food Res. Int. 37 (2004) 587–594, http://dx.doi.org/10.1016/j.foodres.2003.12.012. ˜ Experiences in [10] C. Quintáns, J.M. Acevedo, A.M. Cao y Paz, M.J. Grana, measuring density by fiber optic sensors in the grape juice fermentation process, XIX IMEKO World Congr. Fundam. Appl. Metrol. (2009) 2579–2582. [11] F. Jiménez-Márquez, J. Vázquez, J. Úbeda, J.L. Sánchez-Rojas, High-resolution low-cost optoelectronic instrument for supervising grape must fermentation, Microsyst. Technol. 20 (2014) 769–782, http://dx.doi.org/10.1007/s00542013-2033-3. [12] F. Jiménez-Márquez, J. Vázquez, J. Úbeda, J.L. Sánchez-Rojas, Low-cost and portable refractive optoelectronic device for measuring wine fermentation kinetics, Sens. Actuators B Chem. 178 (2013) 316–323, http://dx.doi.org/10. 1016/j.snb.2012.12.091. [13] J. Toledo, F. Jiménez-Márquez, J. Úbeda, V. Ruiz-Díez, G. Pfusterschmied, U. Schmid, J.L. Sánchez-Rojas, Piezoelectric MEMS resonators for monitoring grape must fermentation, J. Phys. Conf. Ser. 757 (2016) 12020, http://dx.doi. org/10.1088/1742-6596/757/1/012020. [14] B. Jakoby, M.J. Vellekoop, Physical sensors for liquid properties, IEEE Sens. J. 11 (2011) 3076–3085, http://dx.doi.org/10.1109/JSEN.2011.2167716. [15] J. Toledo, T. Manzaneque, J. Hernando-García, J. Vázquez, A. Ababneh, H. Seidel, M. Lapuerta, J.L. Sánchez-Rojas, Application of quartz tuning forks and extensional microresonators for viscosity and density measurements in oil/fuel mixtures, Microsyst. Technol. 20 (2014) 945–953, http://dx.doi.org/ 10.1007/s00542-014-2095-x. [16] J. Toledo, T. Manzaneque, V. Ruiz-Díez, F. Jiménez-Márquez, M. Kucera, G. Pfusterschmied, E. Wistrela, U. Schmid, J.L. Sánchez-Rojas, Comparison of in-plane and out-of-plane piezoelectric microresonators for real-time monitoring of engine oil contamination with diesel, Microsyst. Technol. 22 (2016) 1781–1790, http://dx.doi.org/10.1007/s00542-016-2833-3. [17] T. Manzaneque, V. Ruiz, J. Hernando-Garcia, A. Ababneh, H. Seidel, J.L. Sanchez-Rojas, Characterization and simulation of the first extensional mode of rectangular micro-plates in liquid media, Appl. Phys. Lett. 101 (2012), http://dx.doi.org/10.1063/1.4758470. [18] J.L. Sanchez-Rojas, J. Hernando, a Donoso, J.C. Bellido, T. Manzaneque, a Ababneh, H. Seidel, U. Schmid, Modal optimization and filtering in piezoelectric microplate resonators, J. Micromech. Microeng. 20 (2010) 55027, http://dx.doi.org/10.1088/0960-1317/20/5/055027. [19] J. Toledo, T. Manzaneque, V. Ruiz-Díez, M. Kucera, G. Pfusterschmied, E. Wistrela, W. Steindl, U. Schmid, J.L. Sánchez-Rojas, Piezoelectric mems resonators for density and viscosity sensing in engine oil with diesel fuel, Transducers 2591 (2015) 436–439, http://dx.doi.org/10.1109/TRANSDUCERS. 2015.7180954. [20] G. Pfusterschmied, M. Kucera, J. Toledo, W. Steind, V. Ruiz-Diez, A. Bittner, M. Schneider, J.L. Sanchez-Rojas, U. Schmid, Wine fermentation monitoring using piezoelectric MEMS resonators, in: 20th Int Conf. Miniaturized Syst. Chem. Life Sci., MicroTAS, 2016. [21] A.W. Leissa, Vibration of Plates, Scientific and Technical Information Division, National Aeronautics and Space Administration, 1969, NASA SP-160. [22] M. Kucera, F. Hofbauer, E. Wistrela, T. Manzaneque, V. Ruiz-Díez, J.L. Sánchez-Rojas, A. Bittner, U. Schmid, Lock-in amplifier powered analogue Q-control circuit for self-actuated self-sensing piezoelectric MEMS resonators, Microsyst. Technol. 20 (2014) 615–625, http://dx.doi.org/10.1007/ s00542-013-1997-3. [23] J. Toledo, T. Manzaneque, V. Ruiz-Díez, M. Kucera, G. Pfusterschmied, E. Wistrela, U. Schmid, J.L. Sánchez-Rojas, Piezoelectric resonators and oscillator circuit based on higher-order out-of-plane modes for density-viscosity measurements of liquids, J. Micromech. Microeng. (2016) 84012, http://dx. doi.org/10.1088/0960-1317/26/8/084012. [24] V. Ferrari, D. Marioli, A. Taroni, Improving the accuracy and operating range of quartz microbalance sensors by a purposely designed oscillator circuit, IEEE Trans. Instrum. Meas. 50 (2001) 1119–1122, http://dx.doi.org/10.1109/19. 963169.
298
J. Toledo et al. / Sensors and Actuators B 254 (2018) 291–298
[25] P. Ribéreau-Gayon, Y. Glories, A. Maujean, D. Dubourdieu, Handbook of Enology, 2006, http://dx.doi.org/10.1002/0470010398. [26] F.L. Walls, D.W. Allan, Measurements of frequency stability, Proc. IEEE 74 (1986) 162–168, http://dx.doi.org/10.1109/PROC.1986.13429. [27] T. Manzaneque, V. Ruiz-díez, J. Hernando-garcía, E. Wistrela, M. Kucera, U. Schmid, J.L. Sánchez-rojas, Piezoelectric MEMS resonator-based oscillator for density and viscosity sensing, Sens. Actuators A Phys. 220 (2014) 305–315, http://dx.doi.org/10.1016/j.sna.2014.10.002. [28] C. Castille, I. Dufour, C. Lucat, Longitudinal vibration mode of piezoelectric thick-film cantilever-based sensors in liquid media, Appl. Phys. Lett. 96 (2010) 1–4, http://dx.doi.org/10.1063/1.3387753. [29] A. Maali, C. Hurth, R. Boisgard, C. Jai, T. Cohen-Bouhacina, J.P. Aimé, Hydrodynamics of oscillating atomic force microscopy cantilevers in viscous fluids, J. Appl. Phys. 97 (2005) 1–6, http://dx.doi.org/10.1063/1.1873060. [30] M. Youssry, N. Belmiloud, B. Caillard, C. Ayela, C. Pellet, I. Dufour, A straightforward determination of fluid viscosity and density using microcantilevers: from experimental data to analytical expressions, Sens. Actuators A Phys. 172 (2011) 40–46, http://dx.doi.org/10.1016/j.sna.2011.02. 025. [31] L.D. Landau, E.M. Lifshitz, Fluid Mechanics, Elsevier, 1987, http://dx.doi.org/ 10.1007/b138775. [32] C. Riesch, E.K. Reichel, F. Keplinger, B. Jakoby, Characterizing vibrating cantilevers for liquid viscosity and density sensing, J. Sens. 2008 (2008) 1–9, http://dx.doi.org/10.1155/2008/697062. [33] I. Dufour, E. Lemaire, B. Caillard, H. Debéda, C. Lucat, S.M. Heinrich, F. Josse, O. Brand, Effect of hydrodynamic force on microcantilever vibrations: applications to liquid-phase chemical sensing, Sens. Actuators B Chem. 192 (2014) 664–672, http://dx.doi.org/10.1016/j.snb.2013.10.106. [34] Compendium of International Methods of Analysis of Wines and Musts, 2015.
Biographies Javier Toledo Serrano received his Engineer’s degree from the University of CastillaLa Mancha (Spain) in 2010, and a master’s degree in microelectronics, design and applications of micro/nano systems from the University of Sevilla He is a four-year doctoral student at the Microsystems, Actuators and Sensors group from the Universidad de Castilla-La Mancha, Ciudad Real (Spain). His research interests are in MEMS with liquid media application.
Víctor Ruiz-Díez received his Engineer’s degree from the Universidad de Castilla-La Mancha (Spain) in 2010, and a master’s degree in Physics and Mathematics from the Universidad de Castilla-La Mancha and Universidad de Granada (Spain) in 2011. He is a third-year doctoral student at the Microsystems, Actuators and Sensors group from the Universidad de Castilla-La Mancha, Ciudad Real (Spain). His research interests are in MEMS with liquid media applications. Georg Pfusterschmied started studies in electrical engineering in 2007 and receivedhis Dipl. Ing. (M.Sc.) degree in 2013 from Vienna University of Technology. In 2010, he joined the Institute for Sensor- and Actuator Systems where he performed hisdiploma thesis in the field of MEMS Devices for Sensing and Harvesting Applica-tions. Since Feb. 2014 he is pre-Doc for SiC-based MEMS at Vienna University of Technology. Ulrich Schmid started studies in physics and mathematics in 1992. He performed his diploma work at the research laboratories of the Daimler-Benz AG on the electrical characterization of silicon carbide microelectronic devices for high temperature applications. In 1999, he joined the research laboratories of Daimler Chrysler AG (now EADS Deutschland GmbH) in Ottobrunn/Munich. He developed a robust fuel injection rate sensor for automotive diésel engines and received his Ph.D. degree in 2003. From 2003 to 2008, he was post-doc at the Chair of Micromechanics, Microfluidics/Microactuators at Saarland University. Since October 2008, he is full professor for Microsystems Technology at the Vienna University of Technology. José Luis Sánchez-Rojas received the B.E., M.S., and Ph.D. degrees in telecommunication engineering from the Universidad Politécnica de Madrid, Spain. In 1997, he was an Invited Postdoctoral Scientist with Cornell University, Ithaca, NY, and Colorado University, Boulder. He is currently a Full Professor with the Universidad de Castilla-La Mancha, Ciudad Real, Spain. His research interests cover optical and electrical characterization and modeling of semiconductor devices and MEMS/NEMS sensors.
Contents lists available at ScienceDirect
Sensors and Actuators B: Chemical journal homepage: www.elsevier.com/locate/snb
Flow-through sensor based on piezoelectric MEMS resonator for the in-line monitoring of wine fermentation J. Toledo a,∗ , V. Ruiz-Díez a , G. Pfusterschmied b , U. Schmid b , J.L. Sánchez-Rojas a a b
Microsystems, Actuators and Sensors Group, Universidad de Castilla-La Mancha, Ciudad Real, Spain Institute of Sensor and Actuator Systems, TU Wien, Vienna, Austria
a r t i c l e
i n f o
Article history: Received 18 January 2017 Received in revised form 13 July 2017 Accepted 14 July 2017 Available online 17 July 2017 Keywords: Microresonators Piezoelectric Oscillator circuit Density In-line Grape must fermentation
a b s t r a c t The traditional procedure followed by winemakers for monitoring grape must fermentation is not automated, has not enough accuracy or has only been tested in discrete must samples. In order to contribute to the automation and improvement of the wine fermentation process, we have designed an AlNbased piezoelectric microresonator, serving as a density sensor, resonantly excited in the 4th-order roof tile-shaped vibration mode. Furthermore, conditioning circuits were designed to convert the one-port impedance of the resonator into a resonant two-port transfer function. This allowed us to design a Phase Locked Loop-based oscillator circuit, implemented with a commercial lock-in amplifier with an oscillation frequency determined by the resonance mode. We measured the fermentation kinetics by simultaneously tracking the resonance frequency and the quality factor of the microresonator. The device was first calibrated with an artificial model solution of grape must and then applied for the in-line monitoring of real grape must fermentation. Our results demonstrate the high potential of MEMS resonators to detect the decrease in sugar and the increase in ethanol concentrations during the grape must fermentation with a resolution of 1 mg/ml and 20 Pa s as upper limits for the density and viscosity, respectively. © 2017 Elsevier B.V. All rights reserved.
1. Introduction The fermentation of grape must involves the interaction between yeasts, bacteria, fungi and viruses. A correct bio-chemical process is a necessary condition but not sufficient by itself to determine the final quality of the wine, whose assessment relies on a comprehensive analysis of its chemical components, as the basic flavour of wine depends on at least 20 compounds [1]. Therefore, winemakers must carefully supervise the wine fermentation process to ensure a wine of the expected quality. One of the key parameters monitored during wine fermentation is the fermentation kinetics. This provides essential information about the steady transformation of grape must into wine due to the decrease of glucose and fructose that leads to the formation of ethanol, glycerol and carbon dioxide along with biomass, as a result of yeast metabolism [2]. This process is traditionally monitored by enologists, who manually extract and analyse discrete samples at least twice a day using instruments such as the aerometer, spectrophotometer or colorimeter. They essentially determine the density and its rate of change since these parameters provide informa-
∗ Corresponding author. E-mail address: [email protected] (J. Toledo). http://dx.doi.org/10.1016/j.snb.2017.07.096 0925-4005/© 2017 Elsevier B.V. All rights reserved.
tion about the evolution of the fermentation as a result of yeast metabolism. In an industrial fermentation process, temperature fluctuations would influence to some extent the fermentation process and thus the density. Nevertheless, it is the variations that help enologists determine the current fermentation status, and whether or not corrective measures are required as it progresses. Besides, a variety of alternative procedures have been used for monitoring the grape must fermentation such as density determination based either on differential pressure measurements [3] or on the use of flexural oscillators [4], monitoring of CO2 released during the process due to the gradual loss of mass [3,5], determination of yeast cell population evolution by means of impedance techniques [6] and turbidity measurements [7], ultrasound measurements conducted to determine the propagation velocity in grape musts [8,9], refractive techniques based on fibre optics [10], optoelectronic device based on measurements of the refractive index [11,12] and an off-line monitoring based on a piezoelectric resonator [13]. Overall, these different approaches have not enough resolution or have only been tested in discrete must samples. In this context, the simultaneous determination of the density and viscosity of liquids, through measurement of the resonant frequency (fr ) and quality factor (Q-factor) of a mechanical resonator, has already been reported [14–16]. This approach presents several advantages with
292
J. Toledo et al. / Sensors and Actuators B 254 (2018) 291–298
Fig. 1. (a) Top-view micrograph of the resonator. (b) Modal shape measured (15-mode) with a laser Doppler vibrometer.
respect to traditional methods, such as real-time analysis, on-line configuration and low liquid volumes. Piezoelectric resonators with in-plane vibration modes [17], flexural and torsional modes [18] have already been tested. However, it is well-established that a different family, such as the roof tile-shape modes, present better quality factors at moderate frequencies [19,20]. The present work evaluates the performance of this vibration mode, within micromachined self-actuated and self-sensing aluminium nitride (AlN)-based cantilever sensor for the in-line monitoring of grape must fermentation in flow-through configuration. 2. Material and methods 2.1. Resonator design and characterization The resonator was designed for the 4th order roof-tile shaped vibration mode. It resulted in a cantilever resonator with optimized electrode layout featuring five nodal lines in one direction, and 1 nodal line in the perpendicular direction. Considering Leissa’s nomenclature, the vibration mode is named as 15-mode [21]. The resonator was designed with a length of L = 2524 m, width of W = 1274 m and thickness T of 20 m. It was fabricated from a SOI wafer with a 20 m-thick device layer covered with a 650 nm-thin AlN piezoelectric film [22]. The top metallization has four striped electrodes that allow a selective excitation of the vibration modes and act as a filter for higher modes [18]. This device has already been tested in a previous work obtaining a quality factor of 116 in 2-Propanol [23]. The considered modal shape and the fabricated resonator are shown in Fig. 1 [13].
Fig. 2. Open loop response for the 15-mode measured in a liquid test (2-Propanol) with and without the dummy compensation.
2.2. PLL-based oscillator The simultaneous determination of density and viscosity of liquids, through measurement of the resonant frequency and the quality factor of a mechanical resonator, is challenging due to the low quality factors and parasitic effects present in liquid media [24]. For this reason, our approach is based on roof tile-shaped resonators in two-port configuration and a parasitic compensating device. In the two-port scheme, one of the electrodes was used for actuation (+) and the other for sensing (−) (see Fig. 1). However, our results show that a capacitive crosstalk (Cft ) across the actuation and sensing ports has a significant contribution to the output current. To minimize this parasitic component, a compensating device was introduced. This dummy device reproduces the structure of the resonator, but without the backside release, thus preventing any vibration. For this reason, we designed an interface circuit, used in a previous work [16], to subtract the dummy response from that of the resonator. Since the materials and dimensions in the
Fig. 3. Schematics of the PLL-based oscillator and the interface circuit.
dummy and resonator devices are the same, they are expected to show identical electrical behaviour with respect to their parasitic effects. This results in a clear resonance, with low baseline (see Fig. 2) and remarkable phase transition [13]. However, a phase shift (about −12◦ ) is introduced by the instrumentation amplifier (Fig. 2) at the maximum gain, representing the differences in electrical performance between the dummy, the resonator device and Cft . In order to compensate this phase shift, the resonator and interface circuit were finally included in an oscillator based on a Phase Locked Loop (PLL) instrument (Fig. 3), instead of an oscillator circuit based on discrete components [16]. The phase compensation
J. Toledo et al. / Sensors and Actuators B 254 (2018) 291–298 Table 1 Composition of model solutions representing a normal fermentation process (N1 :N9 ).
293
Table 2 Composition of model solutions representing a sluggish fermentation process (S1 :S9 ).
Solution
Fructose [g/l]
Glucose [g/L]
Glycerol [g/L]
Ethanol [% v/v]
Solution
Fructose [g/l]
Glucose [g/L]
Glycerol [g/L]
Ethanol [% v/v]
N1 N2 N3 N4 N5 N6 N7 N8 N9
110 90 70 60 40 20 8 5 2
100 80 30 20 10 2 2 2 1
0 0 5 5 6 7 7 7 9
0 1 6 8 9 12 13 13 14
S1 S2 S3 S4 S5 S6 S7 S8 S9
110 90 80 70 66 64 62 58 57
100 80 60 30 27 26 24 22 21
0 0 3 5 5 5 6 6 7
0 1 3 6 7 7 7 7 8
allowed to make the oscillation frequency approximately equal to the natural frequency of the resonator. The PLL instrument used (Zurich HF2LI) provides software-based voltage controlled oscillator (VCO), proportional-integral controller (PI) and phase detector (PD) blocks that form a control system able to track the resonant frequency when the resonator is affected by the surrounding medium. The configuration parameters used for each of the blocks were the proportional gain, Kp = 2.63 Hz/deg, time constant of the PI controller, tc = 16.84 ms, and the bandwidth of the system, BW = 100 Hz.
Table 3 Values of the coefficients obtained in the calibration process for the normal model solution.
Normal model solution
Kic [s]
C1 [m]
C2 []
C3 [m2 ]
C4 [m]
3.58 × 10−10
0.12
−20.99
2.11
−103.22
3. Results and discussion 3.1. Sensor calibration With the goal of estimating the density () and viscosity (), during the fermentation process, two variables were measured from our interface circuit: the oscillation frequency (fosc ) and the gain of the interface circuit module at operation (Gosc ), given by the ratio Vout /Vin . Furthermore, for the application of the resonator as a density-viscosity sensor, a calibration procedure was developed with a model solution of artificial grape must [20] and finally applied to the monitoring of one model sluggish fermentation and other real grape must fermentation process. This calibration process was carried out in two steps, each with adjustable parameters [16].
3.1.1. Preparation of the model solutions of grape must In order to confirm that piezoelectric resonators are valid for the monitoring of grape must fermentation, two different sets of model solutions were used for pre-investigation purposes. The first set represents a normal fermentation (N1 :N9 ) and the second represents a sluggish fermentation (S1 :S9 ). A variety of reasons may be attributable to this kind of failed or sluggish fermentations, which occasionally occur: the lack of dissolved oxygen, an unbalanced ratio between sugar and nitrogen, a low fermentation temperature, an inadequate rehydration, a thermal shock of yeast, etc. [25]. These model solutions represent different stages of the corresponding fermentation process according to their particular mixture of glucose, fructose, ethanol and glycerol dissolved in water. The composition of the prepared model solutions for a normal and sluggish fermentation is shown in Tables 1 and 2 [11]. In the normal model solution, a decrease in fructose and glucose concentration from 110 & 100 g/l (N1 ) to 2 & 1 g/l (N9 ) occurs. Similarly, the glycerol and ethanol concentration are increasing from zero to 9 g/l and 14% v/v. In the sluggish fermentation the decrease in fructose and glucose concentration stops prematurely and settles around 57 and 21 g/l. Along with this development the increase in the ethanol concentration stops and does not exceed the value of 8% v/v.
Fig. 4. Resonant frequency and quality factor of the resonator immersed in two model solutions of grape must at 20 ◦ C.
3.1.2. Calibration process with model solutions of grape must In the first step of the calibration process the circuit outputs fosc and Gosc were transformed into the key parameters of the resonator, namely the resonant frequency (fr ) and the quality factor (Q-factor). Thanks to the phase shift compensation provided by the PLL, fr is obtained as the oscillation frequency for each liquid. On the other hand, Q-factor is calculated from the circuit output gain, Gosc . The constant Kic , which is not affected by the loading conditions of the resonator, was calculated by fitting the values of fr , Gosc and the experimental data obtained for different liquids [23] (see Table 3). Gosc = Kic Q 2fr
(1)
Due to the high sensitivity of the density and viscosity to the temperature and electronic noise from surrounding equipment at different frequencies, low, but measurable drift effects are present when determining fr and Q-factor. The random fluctuation of fr was obtained directly through the Allan deviation of fosc [26]. However, the random fluctuation of Gosc was translated into the corresponding fluctuation of Q-factor by error propagation in Expression (1) [27]. These fluctuations determine the ultimate resolutions obtained with the sensor; for the model solutions of grape must they were below 100 mHz for fr and 0.7 for Q-factor. Fig. 4 shows the mean values of fr and Q-factor when the resonator is immersed in the model solutions of grape must: normal and slug-
294
J. Toledo et al. / Sensors and Actuators B 254 (2018) 291–298
Table 4 Density and viscosity values estimated from our resonator (MEMS) and measured with the commercial density-viscosity meter (viscometer) at 20 ◦ C for model solutions of grape must and the calibration error. Normal Fermentation Solution
Viscometer values
Sluggish Fermentation MEMS values
Calibration error
[mPa s]
[g/ml]
[mPa s]
1.949 1.974 1.722 1.707 1.770 1.754 1.770 1.809 1.742
1.082 1.063 1.028 1.017 1.003 0.991 0.985 0.984 0.983
2.274 1.090 2.054 1.069 1.797 1.033 1.880 1.018 1.604 1.002 1.598 0.987 1.752 0.986 1.662 0.979 1.679 0.977 Mean error [%]
1 2 3 4 5 6 7 8 9
[g/ml]
fr =
1−
1+
Q =
2
g1 = C1
1+
1 2Q 2
Lg1 m
[mPa s]
[g/ml]
[mPa s]
16.7 4.0 4.3 10.1 9.4 8.9 1.0 8.1 3.6 7.36
0.7 0.6 0.5 0.1 0.1 0.4 0.1 0.5 0.6 0.38
2.032 1.969 1.819 1.724 1.774 1.770 1.676 1.714 1.777
1.081 1.063 1.049 1.029 1.024 1.022 1.021 1.019 1.017
2.073 1.085 2.092 1.070 2.110 1.045 2.154 1.039 1.892 1.029 1.847 1.024 1.832 1.022 1.763 1.019 1.743 1.016 Mean error [%]
(2)
fr + C2 C
4
P [g/ml]
ε [%]
ε [%]
2.0 6.2 16.0 25.0 6.7 4.4 9.3 2.9 2.0 8.3
0.4 0.7 0.3 1.0 0.5 0.2 0.1 0.1 0.1 0.38
to distinguish between ordinary and sluggish fermentations at an early stage of the fermentation process in artificial grape must. 3.1.3. Calibration error and sensor resolution Once the density and viscosity for each liquid are known, both the error associated with the calibration process and the resolution related to the random fluctuations can be determined. The error terms are calculated as the deviation between the values of viscosity and density estimated from our calibration (MEMS) and the values measured with the commercial instrument (viscometer), obtaining a mean error term for the sluggish fermentation around 8.3% for the viscosity and 0.38% for the density. The resolutions were obtained from the measured random fluctuations in fr and Q-factor by Expressions (6) and (7):
=
|
=
∂ |fr ∂fr
|
2
∂ |fr ∂fr
+
2
|
∂ |Q ∂Q
+
|
2
∂ |Q ∂Q
(6)
2 (7)
Where and are the estimated resolutions for density and viscosity, respectively, and the partial derivatives are computed from Expressions (2)–(5). The values obtained for all the model solutions are below 500 g/ml and 17 Pa s, for density and viscosity. The resolutions and the calibration errors obtained are better for density than for viscosity. This occurs because the Q-factor, determined from the measurement of Gosc in the interface circuit, presents a worse resolution than the frequency measurement. 3.2. Monitoring grape must fermentation
fo,vac
√
g2 = C3 +
Calibration error
ε [%]
Lg2 m Lg2 m
MEMS values
ε [%]
gish fermentation process. The measurements of fr and Q-factor were performed with a temperature control unit [16]. Nevertheless, the impact of temperature fluctuations in the resolution of the sensor during the measurement process is minimal compared with the continuous evolution of the real grape must fermentation. As can be seen in Fig. 4, the significant changes in the composition of the model solutions affects the fr and the Q-factor of the resonator [13]. The combination of unfermented sugars and less than expected ethanol concentration, resulted into a flat curve of fr for the sluggish fermentation. In the second step of the calibration procedure, the physical properties of the liquid, density and viscosity, were obtained from the key parameters of the resonator, fr and Q-factor. These are related to the mechanical properties of the resonators, i.e. resonator equivalent mass (m) and natural frequency in vacuum (fo,vac ), and the fluid conditions, i.e. distributed damping associated to the liquid (g1 ) and the distributed equivalent added mass (g2 ), both per unit length (L). A model defined in previous works [28–30] allows for the determination of these variables separately as in Eqs. (2) and (3). Nevertheless, the dependence of g1 and g2 on the viscosity and density of the fluid, as a closed-form analytical expression, is only available for simple geometries and particular cases [31]. For this reason, we used a Taylor series expansion of g1 and g2 in terms of the viscosity, density and their product (see Expressions (4) and (5)) [32,33]. fo,vac
Viscometer values
√
(3)
(4) (5)
fr
The model was solved iteratively to obtain the values of the coefficients C1 , C2 , C3 and C4 (Table 3) that fit best to the experimental data for the normal model solution [13]. The results of the estimated density values using our MEMS device are compared to those measured by the density-viscosity meter Anton Paar DMA4100M (viscometer) at 20 ◦ C in Table 4. As can be seen, an almost linear decrease in the density occurs for the normal model solution and a premature stop in the density decrease for the sluggish solution. These results show the potential
In the previous section, the theoretical background for using the density and viscosity as valid parameters of the fermentation process was checked. In addition, the performance of the resonator was verified and validated with two model solutions of grape must. Nevertheless, our main goal is the monitoring of a real grape must fermentation process. For this reason, we applied the same procedure that was developed for the model solutions. In this case, through the measurements of fosc and Gosc in the interface circuit and using the coefficients Ci obtained for the normal model solution (see Table 3), we were able to monitor the density and the viscosity during the fermentation. The performance and reliability of the sensor in a real grape must were validated by three different fermentation processes and carried out by different procedures. In the first process, designated as off-line, grape must samples were manually extracted from a commercial fermenter and measured with the resonator under static
J. Toledo et al. / Sensors and Actuators B 254 (2018) 291–298
Fig. 5. Resonator wire bonded in a 24-pin DIP (dual inline package) and immersed in a red grape must. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
immersion. In the second set-up, denominated as in-line, a homemade fermenter was implemented. In this case, the grape must was automatically injected into the cell containing the resonator with a peristaltic pump. After each measurement, the sensor was cleaned with water. In both set-ups the measurements were made with the grape must in static conditions and verifying that the resonator characteristics were maintained after each measurement and cleaning step. However, in the third set-up, denominated as flow-through, there was a continuous flow of grape must through the cell with the resonator. In this case, the sensor was not cleaned after each measurement as grape must was continuously circulating. 3.2.1. Off-line monitoring of a grape must fermentation Grape must was completed using 25% (w/v) of red grape skin from Cencibel variety. Afterwards, the must was inoculated with 0.2 g/ml of Saccharomyces cerevisiae strain (UCLM S325, FouldSpringer) previously rehydrated following the supplier’s guidelines. A commercial cylindrical fermenter (mini-Bioreactor Applikon) was filled with 3 L of inoculated must and grape skins, where the temperature was controlled at 28 ◦ C. In the first method, the fermentation monitoring was carried out by means of the extraction and analysis of 7 mL-samples every 5 h approximately during 6 days. After the extraction the samples were centrifuged for one minute at 1000 rpm (Universal 32R Hettich) and kept refrigerated until analysis. In this case, the resonator was immersed in the grape must samples (Fig. 5), cleaned with water and dried with N2 after
295
Fig. 6. Off-line monitoring of the density-viscosity values estimated from our resonator (MEMS) and measured with the commercial density-viscosity meter (viscometer) at 20 ◦ C.
each measurement. Once the measurements were completed, the samples were kept frozen at −20 ◦ C, in order to stop the evolution of the fermentation process. Following the same procedure as in the previous section with the model solutions, the evolution of density and viscosity was measured as shown in Fig. 6. As can be observed, the density follows the expected evolution for a normal fermentation [11,13] and viscosity values follow a rather similar trend. Both parameters decrease with the evolution of the fermentation since the ethanol concentration increases. The viscosity is not a measured variable in the monitoring of a fermentation. However, in our case it is obtained simultaneously with the density providing further information of the process. The error terms and resolutions were calculated as in Section 3.1.3. The mean error term obtained was around 2.4% for the viscosity and 0.3% for the density being the viscosity resolution below 7 Pa s. In the case of the real grape must the obtained density resolution was below 100 g/ml instead of the 1200 g/ml value obtained in some previous works using a high-resolution low-cost optoelectronic instrument [11,12]. By comparing to the measurements with the benchtop instrument, the errors obtained confirmed that the calibration process involving the normal model solutions is also valid for other liquids in a similar range of density and viscosity.
Fig. 7. Schematic of the fermenter and external fluid circuit designed for the in-line monitoring of grape must fermentation process.
296
J. Toledo et al. / Sensors and Actuators B 254 (2018) 291–298
Fig. 8. In-line monitoring of the density-viscosity values estimated from our resonator (MEMS) and measured with the commercial density-viscosity meter (viscometer) at 20 ◦ C.
Fig. 9. Flow-through monitoring of the density-viscosity values estimated from our resonator (MEMS) and measured with the commercial density-viscosity meter (viscometer) at 20 ◦ C.
3.2.2. In-line monitoring of a grape must fermentation In the second method, a home-made fermenter, with a volume of 3 L, and a flow cell for the resonator, were designed and implemented in a fluidic circuit. The fermenter was immersed in a thermostatic bath at 29 ◦ C and connected to a peristaltic pump, which enabled the recirculation of the grape must between the fermenter and the cell resonator (Fig. 7). Two filters, with a pore size of 500 and 100 m, were introduced in the tube sampling the grape must. These filters prevents the circulation of grape skin, clots or sediments through the flow cell. An air pump provides oxygen for the adequate yeast fermentation. The cell flow, fabricated with a 3D printer, enabled the input/output of the grape must and the housing of the electronic package enclosing the resonator. In addition, the cell has an aperture to evacuate the air that enters into the cell from the fermenter. The systematic procedure for the measurements is as follows. First of all, the grape must was injected in the flow cell with the peristaltic pump until its temperature reached 20 ◦ C (around one minute). This temperature stabilization was assisted by a thermoelectric controller in contact with the resonator. The measurement of the circuit outputs, fosc and Gosc , was carried out following the same procedure as in the previous sections obtaining the density and viscosity values of Fig. 8. Once completed, the grape must was extracted and measured with the commercial density-viscosity meter. After that, the valves in Fig. 7 were switched to allow water injection into the cell, cleaning the resonator during one minute. In order to check that the cleaning process was successful and effective, the value of fr and Q-factor was registered before and after each measurement in water, obtaining a value of 560,000 ± 10 Hz and 140 ± 2, respectively. The mean error term obtained was around 2% for the viscosity and 0.6% for the density being the resolution below 1 Pa s and 88 g/ml, respectively.
3.2.3. Flow-through monitoring of a grape must fermentation Once the previous in-line set-up was proved as a valid configuration for the continuous monitoring, a third procedure, designated as flow-through, was carried out. In this scheme, all the measurements were taken in a continuous flow-through circulation of the grape must as the fermentation progresses, without cleaning the resonator with water. Unlike the previous procedures, the circuit outputs, fosc and Gosc , were registered every minute, and density and viscosity deduced with the same calibration constants as before. The measurements are summarized in Fig. 9. In order to check the correct evolution of the process and resonator behaviour, different grape must samples were taken approximately every six hours and measured with the commercial density-viscosity meter. This final procedure presents two main advantages compared to the ones previously described. On the one hand, we developed a real time monitoring system allowing the detection of a sluggish fermentation or any other problem during the whole process span. On the other hand, all the measurements were carried out without any cleaning process. In the flow-through measurements, the mean error term was around 1.55% for the viscosity and 0.23% for the density, being the resolution around 20 Pa s and 1 mg/ml, respectively. These values are similar to those of the off-line set-up, demonstrating the performance of piezoelectric resonators as sensors for a real in-line monitoring of a fermentation process. It is worth to remark that the density and viscosity resolutions were slightly better in the previous cases, probably due to the dynamics of the flow in the continuous flow-through of grape must and the eventual formation of air bubbles in the cell (see Table 5). However, the calibration error is slightly lower than those obtained for the other two measurement schemes. This error was determined from the difference between the data measured with the commercial instrument and those taken with the MEMS sensor every six hours. In addition, we have estimated the average error considering all the measurements taken during the fermenta-
Table 5 Summary of the sensor resolution ( and ) and calibration error (ε and ε). Fermentation process Off-line
In-line
Flow-through
[Pa s]
[g/ml]
ε [%]
ε [%]
[Pa s]
[g/ml]
ε [%]
ε [%]
[Pa s]
[mg/ml]
ε [%]
ε [%]
7
100
2.4
0.3
1
88
2
0.6
20
1
1.55
0.23
J. Toledo et al. / Sensors and Actuators B 254 (2018) 291–298
tion (one per minute). For that, we took the difference between the fitting curve plotted in Fig. 9, obtained with the values of the commercial instrument, and those estimated with the MEMS sensor, resulting an error of 4% for the viscosity and 0.7% for the density. This is similar to the values obtained in the two other fermentation processes and within the interval determined in the initial calibration (see Table 4). 3.2.4. Temperature dependence In the experiments described above, the temperature of the grape must was stabilized to 20 ◦ C ± 0.05 ◦ C as a reference in the evolution of the process, to neutralize the temperature fluctuations outside the fermentation tank and their influence on the density and viscosity. Nevertheless, our sensor has been validated for other temperatures of interest in the fermentation industry. Two grape must samples, representing the beginning and ending of the fermentation process, were analysed at 15 and 30 ◦ C. The density and viscosity values were determined at these temperatures with our MEMS sensor, with the same calibration of Section 3.1, and compared with the commercial density-viscosity meter, obtaining an error lower than 1 and 2% for the density and viscosity, respectively. As this error is similar to that previously obtained in the calibration at 20 ◦ C, we can conclude that the sensor responds accurately to the changes in density or viscosity caused by changes in temperature, in this range. The density values obtained at 15 or 30 ◦ C can be used to extrapolate the density to a reference temperature of 20 ◦ C [34], the common reference in winemaking, obtaining a mean density error of 0.5% with respect to the value directly measured at this temperature. These low errors demonstrate the possibility to replace the temperature control unit by a basic temperature monitoring system, where the density at 20 ◦ C can be inferred at any stage of the process from its value at a different temperature. 4. Conclusion This work presents a novel approach for the monitoring of a grape must fermentation process with a piezoelectric MEMS resonator excited in the 4th-order of the roof tile-shaped mode. Two important parameters of the resonator were measured using a PLL-based oscillator circuit: the quality factor and the resonant frequency. Once these two parameters are known, the viscosity and density of the liquid can be determined. In order to measure these parameters in a real grape must, a calibration procedure of the resonator was performed using model solutions of artificial grape must representing an ordinary (for calibration) and a stuck or sluggish (for validation) fermentation process and a commercial density-viscosity meter. Furthermore, an in-line flow-through monitoring of a grape must fermentation was carried out. Our results demonstrate the high performance of MEMS resonators to detect the decrease in sugar and the increase in ethanol concentrations during the grape must fermentation with a resolution of 1 mg/ml and 20 Pa s for the density and viscosity, respectively. Acknowledgments This work was supported by Spanish Ministerio de Economía y Competitividad project TEC2015-67470-P and the Austrian Research Promotion Agency within the “Austrian COMET-Program” in the frame of K2 XTribology (project no. 849109). Javier Toledo acknowledge financial support from FPI-BES-2013-063743 grant. References [1] Ronald S. Jackson, Wine Science, Principles and Applications, third ed., Elsevier, 2008.
297
[2] R.B. Boulton, V.L. Singleton, L.F. Bisson, R.E. Kunkee, Principles and Practices of Winemaking, Springer, US, 1999, http://dx.doi.org/10.1007/978-1-47576255-6. [3] N. El Haloui, D. Picque, G. Corrieu, Alcoholic fermentation in winemaking: on-line measurement of density and carbon dioxide evolution, J. Food Eng. 8 (1988) 17–30, http://dx.doi.org/10.1016/0260-8774(88)90033-7. [4] K. Koukolitschek, G. Krieg, W. Maier, Verfahren und Vorrichtung zur präzisen Bestimmung der Alkoholkonzentration in Flüssigkeiten, 1991, pp. 4138419. [5] E.T. Nerantzis, P. Tataridis, I.A. Sianoudis, X. Ziani, E. Tegou, Winemaking process engineering: on line fermentation monitoring – sensors and equipment, E-J. Sci. Technol. (2007) 29–36. ˜ C. Torre, B. García, C.E. Carleos, R. Crespo, L.M. Cárcel, [6] M.A. Pérez, R. Muniz, Impedance spectrometry for monitoring alcoholic fermentation kinetics under wine-making industrial conditions, XIX IMEKO World Congr. Fundam. Appl. Metrol. (2009) 2574–2578 http://www.imeko2009.it.pt/Papers/FP 228. pdf. [7] R. Crespo, L.M. Cárcel, M.A. Pérez, I. Nevares, M. del Álamo, Suitable at-line turbidity sensor for wine fermentation supervision, Int. Conf. Food Innov. (2010) 1–4. [8] N. Lamberti, L. Ardia, D. Albanese, M. Di Matteo, An ultrasound technique for monitoring the alcoholic wine fermentation, Ultrasonics 49 (2009) 94–97, http://dx.doi.org/10.1016/j.ultras.2008.06.003. [9] P. Resa, L. Elvira, F.M. De Espinosa, Concentration control in alcoholic fermentation processes from ultrasonic velocity measurements, Food Res. Int. 37 (2004) 587–594, http://dx.doi.org/10.1016/j.foodres.2003.12.012. ˜ Experiences in [10] C. Quintáns, J.M. Acevedo, A.M. Cao y Paz, M.J. Grana, measuring density by fiber optic sensors in the grape juice fermentation process, XIX IMEKO World Congr. Fundam. Appl. Metrol. (2009) 2579–2582. [11] F. Jiménez-Márquez, J. Vázquez, J. Úbeda, J.L. Sánchez-Rojas, High-resolution low-cost optoelectronic instrument for supervising grape must fermentation, Microsyst. Technol. 20 (2014) 769–782, http://dx.doi.org/10.1007/s00542013-2033-3. [12] F. Jiménez-Márquez, J. Vázquez, J. Úbeda, J.L. Sánchez-Rojas, Low-cost and portable refractive optoelectronic device for measuring wine fermentation kinetics, Sens. Actuators B Chem. 178 (2013) 316–323, http://dx.doi.org/10. 1016/j.snb.2012.12.091. [13] J. Toledo, F. Jiménez-Márquez, J. Úbeda, V. Ruiz-Díez, G. Pfusterschmied, U. Schmid, J.L. Sánchez-Rojas, Piezoelectric MEMS resonators for monitoring grape must fermentation, J. Phys. Conf. Ser. 757 (2016) 12020, http://dx.doi. org/10.1088/1742-6596/757/1/012020. [14] B. Jakoby, M.J. Vellekoop, Physical sensors for liquid properties, IEEE Sens. J. 11 (2011) 3076–3085, http://dx.doi.org/10.1109/JSEN.2011.2167716. [15] J. Toledo, T. Manzaneque, J. Hernando-García, J. Vázquez, A. Ababneh, H. Seidel, M. Lapuerta, J.L. Sánchez-Rojas, Application of quartz tuning forks and extensional microresonators for viscosity and density measurements in oil/fuel mixtures, Microsyst. Technol. 20 (2014) 945–953, http://dx.doi.org/ 10.1007/s00542-014-2095-x. [16] J. Toledo, T. Manzaneque, V. Ruiz-Díez, F. Jiménez-Márquez, M. Kucera, G. Pfusterschmied, E. Wistrela, U. Schmid, J.L. Sánchez-Rojas, Comparison of in-plane and out-of-plane piezoelectric microresonators for real-time monitoring of engine oil contamination with diesel, Microsyst. Technol. 22 (2016) 1781–1790, http://dx.doi.org/10.1007/s00542-016-2833-3. [17] T. Manzaneque, V. Ruiz, J. Hernando-Garcia, A. Ababneh, H. Seidel, J.L. Sanchez-Rojas, Characterization and simulation of the first extensional mode of rectangular micro-plates in liquid media, Appl. Phys. Lett. 101 (2012), http://dx.doi.org/10.1063/1.4758470. [18] J.L. Sanchez-Rojas, J. Hernando, a Donoso, J.C. Bellido, T. Manzaneque, a Ababneh, H. Seidel, U. Schmid, Modal optimization and filtering in piezoelectric microplate resonators, J. Micromech. Microeng. 20 (2010) 55027, http://dx.doi.org/10.1088/0960-1317/20/5/055027. [19] J. Toledo, T. Manzaneque, V. Ruiz-Díez, M. Kucera, G. Pfusterschmied, E. Wistrela, W. Steindl, U. Schmid, J.L. Sánchez-Rojas, Piezoelectric mems resonators for density and viscosity sensing in engine oil with diesel fuel, Transducers 2591 (2015) 436–439, http://dx.doi.org/10.1109/TRANSDUCERS. 2015.7180954. [20] G. Pfusterschmied, M. Kucera, J. Toledo, W. Steind, V. Ruiz-Diez, A. Bittner, M. Schneider, J.L. Sanchez-Rojas, U. Schmid, Wine fermentation monitoring using piezoelectric MEMS resonators, in: 20th Int Conf. Miniaturized Syst. Chem. Life Sci., MicroTAS, 2016. [21] A.W. Leissa, Vibration of Plates, Scientific and Technical Information Division, National Aeronautics and Space Administration, 1969, NASA SP-160. [22] M. Kucera, F. Hofbauer, E. Wistrela, T. Manzaneque, V. Ruiz-Díez, J.L. Sánchez-Rojas, A. Bittner, U. Schmid, Lock-in amplifier powered analogue Q-control circuit for self-actuated self-sensing piezoelectric MEMS resonators, Microsyst. Technol. 20 (2014) 615–625, http://dx.doi.org/10.1007/ s00542-013-1997-3. [23] J. Toledo, T. Manzaneque, V. Ruiz-Díez, M. Kucera, G. Pfusterschmied, E. Wistrela, U. Schmid, J.L. Sánchez-Rojas, Piezoelectric resonators and oscillator circuit based on higher-order out-of-plane modes for density-viscosity measurements of liquids, J. Micromech. Microeng. (2016) 84012, http://dx. doi.org/10.1088/0960-1317/26/8/084012. [24] V. Ferrari, D. Marioli, A. Taroni, Improving the accuracy and operating range of quartz microbalance sensors by a purposely designed oscillator circuit, IEEE Trans. Instrum. Meas. 50 (2001) 1119–1122, http://dx.doi.org/10.1109/19. 963169.
298
J. Toledo et al. / Sensors and Actuators B 254 (2018) 291–298
[25] P. Ribéreau-Gayon, Y. Glories, A. Maujean, D. Dubourdieu, Handbook of Enology, 2006, http://dx.doi.org/10.1002/0470010398. [26] F.L. Walls, D.W. Allan, Measurements of frequency stability, Proc. IEEE 74 (1986) 162–168, http://dx.doi.org/10.1109/PROC.1986.13429. [27] T. Manzaneque, V. Ruiz-díez, J. Hernando-garcía, E. Wistrela, M. Kucera, U. Schmid, J.L. Sánchez-rojas, Piezoelectric MEMS resonator-based oscillator for density and viscosity sensing, Sens. Actuators A Phys. 220 (2014) 305–315, http://dx.doi.org/10.1016/j.sna.2014.10.002. [28] C. Castille, I. Dufour, C. Lucat, Longitudinal vibration mode of piezoelectric thick-film cantilever-based sensors in liquid media, Appl. Phys. Lett. 96 (2010) 1–4, http://dx.doi.org/10.1063/1.3387753. [29] A. Maali, C. Hurth, R. Boisgard, C. Jai, T. Cohen-Bouhacina, J.P. Aimé, Hydrodynamics of oscillating atomic force microscopy cantilevers in viscous fluids, J. Appl. Phys. 97 (2005) 1–6, http://dx.doi.org/10.1063/1.1873060. [30] M. Youssry, N. Belmiloud, B. Caillard, C. Ayela, C. Pellet, I. Dufour, A straightforward determination of fluid viscosity and density using microcantilevers: from experimental data to analytical expressions, Sens. Actuators A Phys. 172 (2011) 40–46, http://dx.doi.org/10.1016/j.sna.2011.02. 025. [31] L.D. Landau, E.M. Lifshitz, Fluid Mechanics, Elsevier, 1987, http://dx.doi.org/ 10.1007/b138775. [32] C. Riesch, E.K. Reichel, F. Keplinger, B. Jakoby, Characterizing vibrating cantilevers for liquid viscosity and density sensing, J. Sens. 2008 (2008) 1–9, http://dx.doi.org/10.1155/2008/697062. [33] I. Dufour, E. Lemaire, B. Caillard, H. Debéda, C. Lucat, S.M. Heinrich, F. Josse, O. Brand, Effect of hydrodynamic force on microcantilever vibrations: applications to liquid-phase chemical sensing, Sens. Actuators B Chem. 192 (2014) 664–672, http://dx.doi.org/10.1016/j.snb.2013.10.106. [34] Compendium of International Methods of Analysis of Wines and Musts, 2015.
Biographies Javier Toledo Serrano received his Engineer’s degree from the University of CastillaLa Mancha (Spain) in 2010, and a master’s degree in microelectronics, design and applications of micro/nano systems from the University of Sevilla He is a four-year doctoral student at the Microsystems, Actuators and Sensors group from the Universidad de Castilla-La Mancha, Ciudad Real (Spain). His research interests are in MEMS with liquid media application.
Víctor Ruiz-Díez received his Engineer’s degree from the Universidad de Castilla-La Mancha (Spain) in 2010, and a master’s degree in Physics and Mathematics from the Universidad de Castilla-La Mancha and Universidad de Granada (Spain) in 2011. He is a third-year doctoral student at the Microsystems, Actuators and Sensors group from the Universidad de Castilla-La Mancha, Ciudad Real (Spain). His research interests are in MEMS with liquid media applications. Georg Pfusterschmied started studies in electrical engineering in 2007 and receivedhis Dipl. Ing. (M.Sc.) degree in 2013 from Vienna University of Technology. In 2010, he joined the Institute for Sensor- and Actuator Systems where he performed hisdiploma thesis in the field of MEMS Devices for Sensing and Harvesting Applica-tions. Since Feb. 2014 he is pre-Doc for SiC-based MEMS at Vienna University of Technology. Ulrich Schmid started studies in physics and mathematics in 1992. He performed his diploma work at the research laboratories of the Daimler-Benz AG on the electrical characterization of silicon carbide microelectronic devices for high temperature applications. In 1999, he joined the research laboratories of Daimler Chrysler AG (now EADS Deutschland GmbH) in Ottobrunn/Munich. He developed a robust fuel injection rate sensor for automotive diésel engines and received his Ph.D. degree in 2003. From 2003 to 2008, he was post-doc at the Chair of Micromechanics, Microfluidics/Microactuators at Saarland University. Since October 2008, he is full professor for Microsystems Technology at the Vienna University of Technology. José Luis Sánchez-Rojas received the B.E., M.S., and Ph.D. degrees in telecommunication engineering from the Universidad Politécnica de Madrid, Spain. In 1997, he was an Invited Postdoctoral Scientist with Cornell University, Ithaca, NY, and Colorado University, Boulder. He is currently a Full Professor with the Universidad de Castilla-La Mancha, Ciudad Real, Spain. His research interests cover optical and electrical characterization and modeling of semiconductor devices and MEMS/NEMS sensors.