sensors Article

A Blade Tip Timing Method Based on a Microwave Sensor Jilong Zhang 1 , Fajie Duan 1, *, Guangyue Niu 1 , Jiajia Jiang 1 and Jie Li 2 1 2

*

State Key Laboratory of Precision Measuring Technology and Instruments; Tianjin University, Tianjin 300072, China; [email protected] (J.Z.); [email protected] (G.N.); [email protected] (J.J.) China Gas Turbine Establishment; Chengdu 610500, China; [email protected] Correspondence: [email protected]; Tel.: +86-131-3205-3038

Academic Editor: Vittorio M. N. Passaro Received: 22 March 2017; Accepted: 8 May 2017; Published: 11 May 2017

Abstract: Blade tip timing is an effective method for blade vibration measurements in turbomachinery. This method is increasing in popularity because it is non-intrusive and has several advantages over the conventional strain gauge method. Different kinds of sensors have been developed for blade tip timing, including optical, eddy current and capacitance sensors. However, these sensors are unsuitable in environments with contaminants or high temperatures. Microwave sensors offer a promising potential solution to overcome these limitations. In this article, a microwave sensor-based blade tip timing measurement system is proposed. A patch antenna probe is used to transmit and receive the microwave signals. The signal model and process method is analyzed. Zero intermediate frequency structure is employed to maintain timing accuracy and dynamic performance, and the received signal can also be used to measure tip clearance. The timing method uses the rising and falling edges of the signal and an auto-gain control circuit to reduce the effect of tip clearance change. To validate the accuracy of the system, it is compared experimentally with a fiber optic tip timing system. The results show that the microwave tip timing system achieves good accuracy. Keywords: blade tip timing; blade vibration measurement; microwave sensor

1. Introduction Blades are the key components of large rotating machines such as aero engines and gas turbines. Measuring the vibration parameters of blades under working conditions is crucial for analyzing dynamic behavior and damage and stress failure mechanisms, as well as for estimating blade lifespan [1,2]. The blade tip timing (BTT) method has been widely used in blade vibration measurement and analysis instead of the conventional strain gauge system, because it is non-intrusive and is able to measure the vibrations of all blades in real time [3,4]. The BTT system measures the time of arrival (TOA) of a blade tip at the BTT sensor. The TOA changes because of blade vibrations. A once-per-revolution sensor is mounted near the shaft to generate a reference timing signal, as shown in Figure 1. The TOA data is converted to the vibration displacement, and parameters such as vibration frequency and mode can be analyzed through different identification algorithms [5–7]. Tip clearance—the gap between blade tip and engine casing—is an important parameter to guarantee the efficiency and safety of the turbine engines [8].

Sensors 2017, 17, 1097; doi:10.3390/s17051097

www.mdpi.com/journal/sensors

Sensors 2017, 17, 1097 Sensors 2017, 17, 1097

2 of 11 2 of 11

Figure 1. 1. Blade Blade tip tip timing timing method. method. Figure

The TOAcan canbebeacquired acquired different kinds of sensors such as optical capacitive The TOA by by different kinds of sensors such as optical sensors,sensors, capacitive sensors, sensors, eddysensors currentand sensors and magnetoresistive sensors.sensors Opticalhave sensors have achieved time eddy current magnetoresistive sensors. Optical achieved good timegood accuracy accuracy and spatial resolution, with small size and enough bandwidth, using the amplitude of the and spatial resolution, with small size and enough bandwidth, using the amplitude of the reflected reflected or laser Doppler technology [9–13].time Goodaccuracy time accuracy hasbeen alsoobserved been observed in signal or signal laser Doppler technology [9–13]. Good has also in eddy eddy current sensors [14–16] and magnetoresistive sensors [17]. However, high operating current sensors [14–16] and magnetoresistive sensors [17]. However, high operating temperatures, temperatures, as particles well as and smoke particles debris inmake the gas medium opticalfor sensors as well as smoke debris in the and gas medium optical sensorsmake unsuitable many unsuitable many aero engines and gassensors turbines. sensors can be environments; used in high aero enginesfor and gas turbines. Capacitance can Capacitance be used in high temperature temperature environments; however, they are sensitive to the variation gas permittivity, and their however, they are sensitive to the variation of gas permittivity, and theirof bandwidth is limited [18,19]. bandwidth is limited Microwave sensors are a promising technology to overcome these Microwave sensors are [18,19]. a promising technology to overcome these limitations. Microwave sensors are limitations. sensors are and not have sensitive to a polluted and Microwave have a sufficiently large not sensitiveMicrowave to a polluted medium a sufficiently largemedium bandwidth. tip clearance bandwidth. Microwave tip clearance have been reported to[20]. work in conditions up to 2000 °C [20]. sensors have been reported to worksensors in conditions up to 2000 ◦ C Different kinds of sensors have have been Different kinds of microwave microwave sensors been developed developed for for blade blade tip tip clearance clearance and and blade blade tip tip timing measurements [20–24]. The waveguide systems use a metal waveguide as an antenna, and the timing measurements [20–24]. The waveguide systems use a metal waveguide as an antenna, and the operating needs an an additional additional reference reference channel channel operating temperature temperature can can be be very very high high [21,22]. [21,22]. However, However, it it needs for for the the tip tip clearance clearance measurement, measurement, and and aa higher higher frequency frequency or or dielectric dielectric filling filling is is needed needed to to control control the the diameter of the sensor. The coaxial resonator sensor achieves good spatial resolution for tip timing, diameter of the sensor. The coaxial resonator sensor achieves good spatial resolution for tip timing, and [24]. The and is is unsuited unsuited for for tip tip clearance clearance measurement measurement [24]. The patch patch antenna antenna can can achieve achieve tip tip timing timing and and tip tip clearance simultaneously without without an an additional additional reference reference channel. channel. clearance measurement measurement simultaneously In the present present article, article, aa microwave microwave system system for for BTT BTT measurement measurement is is introduced. introduced. A microwave In the A microwave sensor antenna is used to send and and receive microwave signals. The sensor can realize sensor based basedon ona apatch patch antenna is used to send receive microwave signals. The sensor can BTT and tip clearance measurement simultaneously. The form of the received signal and realize BTT and tip clearance measurement simultaneously. The form of the received signal and the the processing method is is analyzed. analyzed. A processing method A timing timing method method for for the the microwave microwave sensor sensor is is proposed proposed to to reduce reduce the the effect tip clearance clearance change change on on the the BTT BTT measurement. measurement. Experiments the effect of of tip Experiments are are conducted conducted to to compare compare the microwave verify the the accuracy accuracy of the microwave microwave microwave BTT BTT system system with with an an existing existing fiber fiber optic optic system system to to verify of the system. clearance measurements system. The The system system is is designed designed to to realize realize simultaneous simultaneous BTT BTT and and tip tip clearance measurements in in the the same paper is is focused focused on on the the BTT BTT measurement measurement aspect. aspect. same system; system; the the present present paper 2. Microwave BTT System 2.1. System System Overview Overview and and Signal Signal Analysis Analysis 2.1. The BTT BTTmeasurement measurementsystem systemacquires acquires TOA of engine blades using microwave sensors. The thethe TOA of engine blades using microwave sensors. The The system structure is shown in Figure 2, with the amplifiers and filters omitted. The phase-locked system structure is shown in Figure 2, with the amplifiers and filters omitted. The phase-locked frequency source source outputs outputs aa precise precise microwave signal, the the signal signal is is amplified to the the frequency microwave signal, amplified and and transmitted transmitted to probe through the circulator, and thus generates the orthogonal local oscillator signals. The probe probe through the circulator, and thus generates the orthogonal local oscillator signals. The probe projects and receives receives the The received received signal signal is is projects the the microwave microwave signal signal to to the the blade blade tip tip and the reflected reflected signal. signal. The filtered and by by thethe mixers and and a paira of intermediate frequency (IF) signals filtered andquadrature quadraturedown-converted down-converted mixers pair of intermediate frequency (IF) is generated. The sum of their squares is calculated in circuits to generate an amplitude signal, and the signals is generated. The sum of their squares is calculated in circuits to generate an amplitude signal,

and the two IF signals are also sampled by analog-to-digital converter (ADC) for tip clearance measurement. The amplitude signal is amplified through the automatic gain control (AGC) circuit

Sensors 2017, 17, 1097

3 of 11

Sensors 2017, are 17, 1097 3 of 11 two IF signals also sampled by analog-to-digital converter (ADC) for tip clearance measurement. The amplitude signal is amplified through the automatic gain control (AGC) circuit and the timing and the timing pulse is generated by the comparator. The timing pulse signals of different channels pulseare is generated by the comparator. The timing pulse signals of different channels are acquired and acquired and the TOAs of all the blades are calculated and uploaded to the host by the acquisition the TOAs of all the blades are calculated and uploaded to the host by the acquisition module. module.

Figure 2. Structure ofof the tiptiming timingand and clearance system. Figure 2. Structure themicrowave microwave blade blade tip tiptip clearance system.

Assuming the transmitting signal generated by the source is X0 = Ascos(ωst + φ0), the received

Assuming the transmitting signal generated by the source is X0 = As cos(ω s t + ϕ0 ), the received signal is: signal is:  A((tw)cos( 0 ϕl+ ((tt)) wst(wslt)+ , ϕ l ), (1) (1) Y = A ( tY ) cos ))+ArAcos( s t0+ l ϕ s t +wϕ r cos l isaccumulated the accumulated phase in the transmissionpath pathand and ϕ(t) φ(t) is is the the phase wherewhere ϕl is φ the phase in the transmission phasedifference differencecaused caused by by the change of tip clearance. Arcos(ω st + φl) expresses the reflected signal from the probe itself, and the change of tip clearance. Ar cos(ω t + ϕ ) expresses the reflected signal from the probe itself, and its s l its amplitude changes slowly with the drift of the antenna reflection coefficient. This signal should amplitude changes slowly with the drift of the antenna reflection coefficient. This signal should be be removed in the signal processing. According to the Friis transmission equation, the sensor received removed in the signal processing. According to the Friis transmission equation, the sensor received power of the reflected microwave signal can be expressed as: power of the reflected microwave signal can be expressed as: Pt Gt Ae Pr P G (2) σ4 , t(4t A ) 2eR Pr = , (2) (4π )2 R4

where Pt is the transmitting power, Gt is the antenna gain of the sensor, Ae is the effective area of the σ is thepower, radar cross-section (RCS) of theof blade, and R isAthe between theof the wherereceiving Pt is theantenna, transmitting Gt is the antenna gain the sensor, the effective area e isdistance sensor and the blade tip. receiving antenna, σ is the radar cross-section (RCS) of the blade, and R is the distance between the When the blade sensor and the blade tip.is perpendicular to the face of the sensor, assume the maximum amplitude of the received signal is A0 and the clearance between the sensor and blade tip is d0. So, according to When the blade is perpendicular to the face of the sensor, assume the maximum amplitude of Equation (2), the amplitude of the received signal can be expressed as:

the received signal is A0 and the clearance between the sensor and blade tip is d0 . So, according to 2 d can  (texpressed ) Equation (2), the amplitude of the receivedA(signal be as: t)  0  A , (3)

0 s 0 d0 2 RCS, σrespectively, (t) where d(t) and σ(t) are the clearance A(t)and = the · · A0 , when the blade is in different 2 d (t) needsσ0to be down-converted. If the heterodyne positions. The received radio-frequency signal

d 2 (t )

(3)

method is adopted, the form of the IF signal would be:

where d(t) and σ(t) are the clearance and the RCS, respectively, when the blade is in different positions. YIF  A (t)cos(to wti be down-converted. (4) 0  l  (t ))  Ar cos( i  l) . The received radio-frequency signal needs Ifwt the heterodyne method is adopted, the form of the IF signal would The reflected signal frombe: the probe itself cannot be eliminated by the filter. Also, the phase factor is retained, so the amplitude and the shape of the signal are affected by phase variation. Thus, the IF YIF (t) cos wi t + ϕ0the + ϕproper t)) + A + ϕThe r cos( wi t[23]. wi must be adjusted in = realAtime to (maintain waveform l + ϕ (signal l ). change of IF also (4) affects BTT accuracy, so the zero-IF quadrature demodulation method is employed in the signal The reflected signal fromsignal the probe itselfwith cannot eliminated the filter. Also, phase processing. The received is mixed the be local oscillatorby signal Aicos(ω st + the φi) and thefactor is retained, so the amplitude theAshape are affected by phase variation. the IF wi orthogonal local oscillatorand signal isin(ωsof t + the φi), signal respectively, and subsequently passes theThus, low-pass

must be adjusted in real time to maintain the proper signal waveform [23]. The change of IF also affects BTT accuracy, so the zero-IF quadrature demodulation method is employed in the signal processing.

Sensors 2017, 17, 1097

4 of 11

The received signal is mixed with the local oscillator signal Ai cos(ω s t + ϕi ) and the orthogonal local oscillator signal Ai sin(ω s t + ϕi ), respectively, and subsequently passes the low-pass filter. The reflected signal from the probe itself Ar Ai cos(ϕl − ϕi ) usually changes slowly, and is filtered together with the other direct current component of the circuit. The result is: Sensors 2017, 17, 1097

4 of 11 s d0 2 σ(t) S I =from · · A0 Ai cos ( ϕ0l −+φi)ϕusually − ϕi ), slowly, and is filtered (5) l + ϕ ( t )changes filter. The reflected signal icos(φ σ0 itself ArA d2 (t)the probe

together with the other direct current component of the circuit. The result is:

and the orthogonal signal SQ . Squaring d 2SI and  (t )SQ and adding them together gives: SI 

0

d 2 (t )



0

 A0 Ai cos( 0  l   (t )  i ) ,

(5)

d0 4 σ ( t ) 2 · A20 A S(t) = 4 SQ and · i and the orthogonal signal SQ. Squaring SI and them together gives: d (t) σadding 0

(6)

d 0 4  (t ) 2 2 The phase component is removed andS (the so the signal after processing t )  signal  edge  Ais (6) is i . 0 Asharper, d 4 (t )  0 suitable for BTT. Considering that the change in clearance is small when the blade tip passes the sensor, The phase component removed signal is sharper, so the signal after the main factor is the change ofis the bladeand tip the RCS. Theedge maximum amplitude of S(t) is processing also inversely is suitable for BTT. Considering that the change in clearance is small when the blade tip proportional to d0 . The received signal is modulated by the blade passing the probe.passes The the system sensor, the main factor is the change of the blade tip RCS. The maximum amplitude of S(t) is also bandwidth is mainly determined by circuit devices, because the carrier frequency is high enough. inversely proportional to d0. The received The bandwidth of the proposed system is 10 signal MHz.is modulated by the blade passing the probe. The system bandwidth is mainly determined by circuit devices, because the carrier frequency is high For a flat plate whose area is A, σ = 4πA2 /λ2 when the beam is perpendicular to the plate. enough. The bandwidth of the proposed system is 10 MHz. The angle of incidence has little change when the2 blade passes, so σ(t) mainly reflects the change of the For a flat plate whose area is A, σ = 4πA /λ2 when the beam is perpendicular to the plate. The projected area of the microwave beam on the blade tip. For so a free blade with a flat the of projected angle of incidence has little change when the blade passes, σ(t) mainly reflects the tip, change the area of the microwave on thebeam blade betip. used approximate A projected small sensor projected area of thebeam microwave ontip the can blade For to a free blade with athe flatRCS. tip, the diameter beam width mean a short rising of the or good spatial resolution, which is area and of the microwave beam on the blade tiptime can be usedsignal, to approximate the RCS. A small sensor diameter and beam width mean a short rising time of the signal, or good spatial resolution, which is preferred for BTT. preferred The phasefor inBTT. Equation (5) can be obtained from Si and Sq by arctangent or other algorithms. phase in the Equation (5) can be path obtained Si and Ssuch q by arctangent or other algorithms. The The phaseThe drift ϕl in transmission duefrom to factors as temperature can be obtained by phase drift φ l in the transmission path due to factors such as temperature can be obtained by tuning tuning the frequency out of the sensor bandwidth. Thus, tip clearance can be acquired from ϕ(t) after the frequency out of the sensor bandwidth. Thus, tip clearance can be acquired from φ(t) after calibration. This is the principle of tip clearance measurement based on a microwave sensor. Thus, it is calibration. This is the principle of tip clearance measurement based on a microwave sensor. Thus, it feasible to integrate BTT and tip clearance measurements into one system. The tip timing pulse can is feasible to integrate BTT and tip clearance measurements into one system. The tip timing pulse can help to pick of the blade tiptip in in the twomeasurement measurement functions share help to out pick part out part of the blade thephase phasesignal. signal. The The two functions share the the driverdriver circuit, acquisition card, and use the same kind of microwave sensor, as shown in the lower circuit, acquisition card, and use the same kind of microwave sensor, as shown in the lower part of Figure part 2. of Figure 2.

2.2. Microwave Sensor for BTT 2.2. Microwave Sensor for BTT A 24-GHz microwave probe basedon onan anend-fire end-fire patch patch antenna in in thethe proposed BTT BTT A 24-GHz microwave probe based antennais isused used proposed system. The circular patch antenna is mounted in a metal case with a ceramic cap in front of it, system. The circular patch antenna is mounted in a metal case with a ceramic cap in front ofasit, as in Figure 3. The outer diameter theprobe probeisis 88 mm. mm. shownshown in Figure 3. The outer diameter ofofthe

Figure Structureof of the the sensor sensor probe. Figure 3. 3.Structure probe.

The temperature in the turbine engine changes over a wide range during operation, and the

The temperature in the turbine engine changes over a wide range during operation, and the probe pass band varies with temperature in a certain range. The main effect of temperature on the probeprobe pass is band variesin with temperature in a certain range. The main the effect of temperature on the the change the relative permittivity of ceramic. Figure 4 shows simulation result of the probereflection is the change in the relative permittivity ceramic. Figuresubstrate 4 shows the simulation result of the coefficient when the permittivityof of the ceramic increases by 20%. The reflection coefficient when the permittivity of the ceramic substrate increases by 20%. The permittivity permittivity of the simulation curve rises by a 7% step. The result shows that the permittivity of the of theceramic simulation curve riseseffect by aon 7% result shows thatprobe the permittivity of aluminathe ceramic has a significance thestep. sensorThe characteristic. A recent prototype used ceramic. The on permittivity pure aluminaAchanges greatlyprototype with temperature, but the has a based significance effect the sensorofcharacteristic. recent probe used alumina-based permittivity of alumina ceramic can remain nearly constant through adding materials with negative temperature coefficients of the dielectric constant. The output frequency of the signal source can be adjusted according to the amplitude of the received signal acquired by the AGC circuit; the variation

Sensors 2017, 17, 1097

5 of 11

ceramic. The permittivity of pure alumina changes greatly with temperature, but the permittivity of alumina ceramic can remain nearly constant through adding materials with negative temperature coefficients of the dielectric constant. The output frequency of the signal source can be adjusted Sensors 2017, 17, 1097 5 of 11 according to the amplitude of the received signal acquired by the AGC circuit; the variation is within the adjustable range of range the driver In addition, the design of theofsensor is also adjusted to is within the adjustable of thecircuit. driver circuit. In addition, the design the sensor is also adjusted increase thethe bandwidth appropriately andand reduce the the impact of permittivity. The The prototype probe can to increase bandwidth appropriately reduce impact of permittivity. prototype probe ◦ C. Silicon nitride has good thermal shock resistance; its thermal conductivity operate at up to 600 can operate at 17, up1097 to 600 °C. Silicon nitride has good thermal shock resistance; its thermal conductivity Sensors 2017, 5 of 11 and so it structural reliability. reliability. and coefficient coefficient of of thermal thermal expansion expansion is is lower lower than than alumina, alumina, so it benefits benefits the the structural is within the range of the driver circuit. In addition, the design of the sensor is alsonitride-based adjusted Furthermore, its temperature coefficient of constant is Thus, silicon Furthermore, itsadjustable temperature coefficient ofthe thedielectric dielectric constant is lower. lower. Thus, silicon nitride-based to increase the bandwidth appropriatelyprobes and reduce the impact of permittivity. The prototype probe ceramic is a better choice for microwave used in high temperatures. ceramic is a better choice for microwave probes used in high temperatures. can operate at up to 600 °C. Silicon nitride has good thermal shock resistance; its thermal conductivity and coefficient of thermal expansion is lower than alumina, so it benefits the structural reliability. Furthermore, its temperature coefficient of the dielectric constant is lower. Thus, silicon nitride-based ceramic is a better choice for microwave probes used in high temperatures.

Figure 4. 4. Reflection Reflection coefficient coefficient of of the the sensor sensor with with the the permittivity permittivityincreased increasedby by20%. 20%. Figure Figure Reflection coefficient of the sensor with the permittivityby increased by 20%. The near and far4.field of an antenna is usually distinguished the Fraunhofer distance d = The near and far field of an antenna is usually distinguished by the Fraunhofer distance d = 2D2 /λ, 2D /λ, where D is the antenna diameter. The 24-GHz sensor works mainly in the near field. The where D is thenear antenna 24-GHz sensordistinguished works mainly near field. The radiation The and fardiameter. field of anThe antenna is usually byin thethe Fraunhofer distance d= radiation patterns of the near field change with distance from the antenna, and are more complex 2D2/λ, where D field is thechange antennawith diameter. Thefrom 24-GHz works in the near field. patterns of the near distance the sensor antenna, andmainly are more complex thanThe those of than those of the far field. The feed point of the patch antenna is usually not in the center of the radiation patterns of the of near change withisdistance the antenna, areantenna, more complex the far field. The feed point thefield patch antenna usuallyfrom not in the center and of the in order to antenna, in order to achieve impedance matching. Assume the direction of the line between the feed than those of the far field. The feed point of the patch antenna is usually not in the of center the achieve impedance matching. Assume the direction of the line between the feed pointcenter and the is pointantenna, and theincenter is X, and the direction Y is perpendicular to X and the axial direction of the probe. order to achieve impedance matching. Assume the direction of the line between the feed X, and the direction Y is perpendicular to X and the axial direction of the probe. The simulation results The simulation ofX,the electric field strengths along these twothe directions at different distances point and theresults center is and the direction Y is perpendicular to X and axial direction of the probe. of the electric field strengths along these two directions at different distances are shown in Figure 5. are shown in Figure 5. The of the lines in the along simulation aredirections all 20 mm. The simulation results of length the electric field strengths these two at different distances The length of the lines in5. the allin 20the mm. are shown in Figure Thesimulation length of theare lines simulation are all 20 mm. 2

Figure 5. Electric field strengths along the two directions at different distances.

Figure 5. Electric field strengths along the two directions at different distances. Figure 5. Electric field strengths along the two directions at different distances. The results show that the field intensity distribution is not symmetrical in the near field due to the pointshow offset. Thethe beam is intensity obviously distribution wider in direction than that in direction Y, and Thefeed results that field is notXsymmetrical in the near fieldthedue to differences decrease with the distance. The electric field strengths along the two directions are similar the feed point offset. The beam is obviously wider in direction X than that in direction Y, and the with each other when is 10The mm, withinfield the far field. So, the probe be installedare with differences decrease withthe thedistance distance. electric strengths along themight two directions similar the direction of the feed point consistent with the length direction of the blade tip, in order to reduce

with each other when the distance is 10 mm, within the far field. So, the probe might be installed with

Sensors 2017, 17, 1097

6 of 11

The results show that the field intensity distribution is not symmetrical in the near field due to the feed point offset. The beam is obviously wider in direction X than that in direction Y, and the differences decrease with the distance. The electric field strengths along the two directions are similar with each other when the distance is 10 mm, within the far field. So, the probe might be installed with the direction of the feed point consistent with the length direction of the blade tip, in order to reduce the signal rise time. The simulation shows that the large difference of the field distribution is6mainly in Sensors 2017, 17, 1097 of 11 the region close to the probe, within about 3 mm. Usually the probe is mounted in a small recess in the casing for rise safety, a recess of 2 shows to 3 mm used to avoid thefield region and reduce the effect. the signal time.soThe simulation thatcan the be large difference of the distribution is mainly in the region to the probe, within about 3 mm. Usually the probe is mounted a small recess Considering thatclose the change of the blade twist angle is usually small, the effectincan be minimized. in thesensor casingbased for safety, so apatch recess antenna of 2 to 3 mm can be used avoid microwave the region and reducetothe effect. The BTT on the is designed to to project signals the blade tip Considering that the change of the blade twist angle is usually small, the effect can be minimized. steadily, so it is also a tip clearance sensor, which is a reason to choose this structure. The opposite way The BTT sensor based on the patch antenna is designed to project microwave signals to the blade tip is to make a strong coupling between the sensor and the blade, thereby the passband differs greatly, steadily, so it is also a tip clearance sensor, which is a reason to choose this structure. The opposite depending on whether the blade is in front of the sensor or not. The signal reflection due to impedance way is to make a strong coupling between the sensor and the blade, thereby the passband differs mismatch shows the arrival of thethe blades greatly, depending on whether blade[24]. is in front of the sensor or not. The signal reflection due to impedance mismatch shows the arrival of the blades [24].

2.3. Timing Method

2.3. TheTiming TOA Method of every blade needs to be extracted from the blade tip signal. There are two main

methods The to generate the BTT pulse. One involves comparing rising or falling TOA of every blade needs to method be extracted from the blade tipthe signal. There are twoedge mainof the generate thevoltage BTT pulse. One method involves comparing the rising or falling of involves the signalmethods with a to fixed trigger in order to generate a pulse sequence for timing. Theedge other signal with a fixed trigger voltage in order to generate a pulse sequence for timing. The other involves finding the time of signal peaks; the zero-crossing point of the resulting differential signal is the peak thesignal. time of signal peaks; the zero-crossing point of the resulting differential signal is the peak of thefinding original of the original signal. Blade tip clearance varies during operation and affects the timing results. Because the microwave Blade tip clearance varies during operation and affects the timing results. Because the beam width changes with tip clearance, as Figure 6 shows, both the amplitude and the rise time of the microwave beam width changes with tip clearance, as Figure 6 shows, both the amplitude and the tip signal change with tip clearance. rise time of the tip signal change with tip clearance.

Figure 6. 6. Sketch signalunder underdifferent different clearances. Figure Sketchofofreceived received tip signal tiptip clearances.

The peak of the sensor signal corresponds to the blade tip position with the largest RCS, and it The peak of the sensor signal corresponds to the blade tip position with the largest RCS, and it is usually invariable with tip clearance change. Thus, the zero-crossing timing method is accurate, is usually invariable with tip clearance change. Thus, the zero-crossing timing method is accurate, even when the tip clearance changes. However, this method requires that the signal has a single peak evenwhich whenrises the tip clearance changes. However, thishave method requires has a single relatively quickly. Some modern blades cavities on the that bladethe tip,signal and for some peakblades whichthe rises relatively modernThe blades have cavities the blade and for signal peak is quickly. flat or hasSome fluctuations. zero-crossing timingon method is nottip, suitable forsome blades theapplications. signal peak is flat or has fluctuations. The zero-crossing timing method is not suitable for such Thus, an improved edge timing method is adopted in the proposed system. An AGC circuit is such applications. used theimproved signal processing circuitmethod to stabilize the sensorinsignal before thesystem. comparator, thereby Thus,inan edge timing is adopted the proposed An AGC circuit reducing the timing error caused by tip clearance change. The AGC circuit includes a peak detector is used in the signal processing circuit to stabilize the sensor signal before the comparator, thereby and variable gain amplifiers, and is controlled by the microprogrammed control unit. Because the reducing the timing error caused by tip clearance change. The AGC circuit includes a peak detector signal width is changed with tip clearance, BTT errors introduced by tip clearance variation cannot and variable gain amplifiers, and is controlled by the microprogrammed control unit. Because the be eliminated simply by levelling the signal amplitude. signal width is changed with tip clearance, BTT errors introduced by tip clearance variation cannot be Hence, the proposed method compares both edges of the sensor signal with a fixed trigger eliminated by the levelling signal voltage,simply and takes averagethe time of theamplitude. two edges as the TOA. The impact of the clearance change on both edges is similar, so the TOA change caused by the rise time change is reduced by using a mean value of the TOA obtained from both edges. This method is equivalent to using the median of the waveform instead of one edge to trigger the timing pulse, as Figure 6 shows. The difference between the two medians is very small after amplification to the same level. In addition, the median

Sensors 2017, 17, 1097

7 of 11

Hence, the proposed method compares both edges of the sensor signal with a fixed trigger voltage, and takes the average time of the two edges as the TOA. The impact of the clearance change on both edges is similar, so the TOA change caused by the rise time change is reduced by using a mean value of the TOA obtained from both edges. This method is equivalent to using the median of the waveform instead of one edge to trigger the timing pulse, as Figure 6 shows. The difference between the7 two Sensors 2017, 17, 1097 of 11 medians is very small after amplification to the same level. In addition, the median is close to vertical when the between the rise and fall time small, makes the TOA the is close todifference vertical when the difference between theisrise and which fall time is small, whichinsensitive makes thetoTOA amplitude insensitivechange. to the amplitude change. Another sample the the signal with with high-speed analog-to-digital converters and process Another method methodisisto to sample signal high-speed analog-to-digital converters and it in the host computer. The signal waveform in a certain range of amplitude is fitted and the timing process it in the host computer. The signal waveform in a certain range of amplitude is fitted and the sequence is calculated with different methods, such as such comparing with thewith trigger or level finding timing sequence is calculated with different methods, as comparing the level trigger or the peak.the However, this requires high sampling and large transfer bandwidth, for finding peak. However, thisa requires a highrate sampling rate and large transferespecially bandwidth, high rotational speeds. The proposed mainly system uses hardware process the signals, thereby especially for high rotational speeds. system The proposed mainly to uses hardware to process the improving the integration the resource consumption of theconsumption host computer, is signals, thereby improvingand thereducing integration and reducing the resource of which the host beneficial for making it an integral part of the engine health monitoring system. computer, which is beneficial for making it an integral part of the engine health monitoring system. 3. 3. Experiments Experiments To To assess assess the the accuracy accuracy of of the the microwave microwave BTT BTT measurement measurement system, system, it it was was compared compared with with an an existing BTT system based on a fiber optic sensor. The fiber optic BTT system used has good existing BTT system based on a fiber optic sensor. The fiber optic BTT system used has good timing timing resolution methods in in previous previous experiments. experiments. For For the the rotor, rotor, resolution and and has has been been compared compared with with strain strain gauge gauge methods the finite element model of which is known, the relative error of the dynamic strain is less than the finite element model of which is known, the relative error of the dynamic strain is less than 10% 10% and and the the relative relative error error of of the the frequency frequency is is less less than than11Hz, Hz,so soititcan canbe beused usedas asaaproper properreference. reference. As As shown shown in in Figure Figure 7, 7, the the experiment experiment was was conducted conducted on on aa rotor rotor test test rig. rig. The The radius radius of of the the rotor rotor was 95 mm, with 16 straight blades. The thickness of the blades was 2 mm. A low rotational was 95 mm, with 16 straight blades. The thickness of the blades was 2 mm. A low rotational speed speed of of 1000 from blade vibration. TheThe effect of bending and and torsion was 1000 rpm rpmwas waschosen chosentotoavoid avoidinterference interference from blade vibration. effect of bending torsion negligible because the shaft was very short.short. A once-per-revolution sensor with with angleangle divisions was was negligible because the shaft was very A once-per-revolution sensor divisions installed near the bearing to minimize the errors in rotational speed. The clearance between the blade was installed near the bearing to minimize the errors in rotational speed. The clearance between the tip andtip microwave sensorsensor was adjusted using ausing displacement platform, and the radial position of the blade and microwave was adjusted a displacement platform, and the radial position fiber optic sensor was adjusted by shims in the fixture. of the fiber optic sensor was adjusted by shims in the fixture.

Figure 7. Test rig and fiber optic blade tip timing sensor (inset). Figure 7. Test rig and fiber optic blade tip timing sensor (inset).

The blade tip of the straight blade used in the experiment was a flat plane, so the projected area The blade tip ofbeam the straight blade used the experiment flatthe plane, so tip. the projected area of of the microwave could be used to in approximate the was RCSaof blade Simulation was the microwave beam could be used to approximate the RCS of the blade tip. Simulation was conducted conducted using the area of the projected region on the blade tip as the RCS. The rotational speed using theblade area of the projected region onas thethe blade tip as theFigure RCS. The rotational speed and curves the blade and the dimension was the same experiment. 8 shows the simulation by dimension was the same as the experiment. Figure 8 shows the simulation curves by Equation (6) Equation (6) and the measured signal, and the simulation result is similar to the measured signal and the measured signal, and thesimulation simulationcurves result is waveform. The amplitude of the is similar leveled to to the the measured measured signal signal waveform. waveform. The The amplitude of the simulation curves is leveled to the measured signal waveform. The diameter of diameter of the projected region of the microwave beam dp was changed from 5 mm to 7 mm in the simulation. The result shows that the width of the signal edge was mainly affected by the projected area of the microwave beam, which was determined by the size and field distribution of the probe antenna and the tip clearance.

Sensors 2017, 17, 1097

8 of 11

the projected region of the microwave beam dp was changed from 5 mm to 7 mm in the simulation. The result shows that the width of the signal edge was mainly affected by the projected area of the microwave beam, which was determined by the size and field distribution of the probe antenna and Sensors 2017, 88ofof11 the tip clearance. Sensors 2017,17, 17,1097 1097 11

Figure The simulation curves and the measured signal. Figure The simulation curves and the measured signal. Figure 8.8.8. The simulation curves and the measured signal.

The Thepulse pulsesignal signalafter afterprocessing processingby bythe themicrowave microwavesensor sensorwas wascollected collectedthe thesame sameway wayas asthe the The pulse signal after processing by the microwave sensor was collected the same way as the optical opticalsensors, sensors,and andthe therising risingand andfalling fallingedges edgesof ofthe thesignal signalwere wereacquired. acquired.The Theangle anglebetween betweenthe the optical sensors, and the rising and falling edges of the signal were acquired. The angle between the microwave Blades were numbered from 1 to 16. microwavesensor sensorand andthe thefiber fiberoptic opticsensor sensorwas was25°. 25°. Blades were numbered from 1 to 16. microwave sensor and the fiber optic sensor was 25◦ . Blades were numbered from 1 to 16. The The TOAs TOAs of of the the blades blades obtained obtained by by the the two two methods methods were were recorded recorded and and compared. compared. The The The TOAs of the blades obtained by the two methods were recorded and compared. The clearance clearance clearancebetween betweenthe theblade bladeand andeach eachsensor sensorwas wasset setto to22mm; mm;there therewas wasan anadditional additional22mm mmbetween between between the blade and each sensor was set to 2 mm; there was an additional 2 mm between the the themicrowave microwavesensor sensorand andthe theend endof ofthe thefixture. fixture.Figure Figure99shows showsthe thevibration vibrationdisplacements displacementsof ofblade blade microwave sensor and the end of the fixture. Figure 9 shows the vibration displacements of blade 1 11obtained by the microwave sensor and the fiber optic sensor, and the differences between them, obtained by the microwave sensor and the fiber optic sensor, and the differences between them, obtained by the microwave sensor and the fiber optic sensor, and the differences between them, measured measuredover over2000 2000revolutions. revolutions.The Thevibration vibrationdisplacement displacementisisthe theproduct productof ofthe theTOAs TOAsand andthe thelinear linear measured over 2000 revolutions. The vibration displacement is the product of the TOAs and the linear velocity. velocity.The Thestandard standarddeviation deviationof ofthe thevibration vibrationdisplacements displacementsobtained obtainedby bythe thefiber fiberoptic opticsensor sensorisis velocity. The standard deviation of the vibration displacements obtained by the fiber optic sensor is 0.00203 0.00203mm, mm,and andthat thatof ofthe themicrowave microwavesensor sensorisis0.00226 0.00226mm. mm.The Thestandard standarddeviation deviationshows showsthat thatthe the 0.00203 mm, and that of the microwave sensor is 0.00226 mm. The standard deviation shows that the two twoBTT BTTsystems systemshave havesimilar similarrandom randommeasurement measurementerrors. errors. two BTT systems have similar random measurement errors.

Figure Vibration displacements blade by the microwave sensor and the fiber optic Figure Vibration displacements of blade 1obtained obtained by the microwave sensor and the fiber optic Figure 9.9.9. Vibration displacements ofof blade 1 1obtained by the microwave sensor and the fiber optic sensor, and the differences between them, measured over 2000 revolutions. sensor, and the differences between them, measured over 2000 revolutions. sensor, and the differences between them, measured over 2000 revolutions.

Usually Usuallythe thetip tipclearance clearanceof ofan anaero aeroengine engineisiswithin within33mm, mm,and andititisispreferred preferredthat thatthe theclearance clearance Usually the tip clearance of an aero engine is within 3 mm, and it is preferred that the clearance between the sensor and the blade be larger to ensure that the blade will not hit the sensor accidentally, between the sensor and the blade be larger to ensure that the blade will not hit the sensor accidentally, between the sensor and the blade be larger to ensure that the blade will not hit the sensor accidentally, so sothe thetip tipclearance clearanceof ofthe themicrowave microwavesensor sensorwas wasset setto to2,2,33and and44mm; mm;the theposition positionof ofthe thefiber fiberoptic optic so the tip clearance of the microwave sensor was set to 2, 3 and 4 mm; the position of the fiber sensor sensorremained remainedunchanged. unchanged.The Thevibration vibrationdisplacements displacementsof ofblade blade11were wererecorded; recorded;the themeans meansand and optic sensor remained unchanged. The vibration displacements of blade 1 were recorded; the means standard standard deviations deviations of of the the differences differences between between the the vibration vibration displacements displacements obtained obtained by by the the two two systems systemsare areshown shownin inTable Table1.1.

Sensors 2017, 17, 1097

9 of 11

and standard deviations of the differences between the vibration displacements obtained by the two systems are shown in Table 1. Sensors 2017, 17, 1097

9 of 11

Table 1. Differences between the vibration displacements obtained by the two systems under different tip clearances. Table 1. Differences between the vibration displacements obtained by the two systems under different tip clearances. Tip Clearance (mm) Mean of Difference (mm) Standard Deviation of Difference (mm)

Tip Clearance 2 (mm) 23 34 4

Mean of2.711 Difference × 10−4 (mm) 2.711 3.159 ××10 10−4−4 3.826 ××10 10−4−4 3.159 3.826 × 10−4

Standard Deviation 0.00127of Difference (mm) 0.00127 0.00149 0.00191 0.00149 0.00191

The results show that the mean of differences between the vibration displacements of the fiber The results showofthat the mean ofsensor differences the vibration the fiber optic sensor and those the microwave remainbetween about the same when displacements the tip clearanceofchanges. optic sensor and those of the microwave sensor remain about the same when the tip clearance The standard deviation is small and changes little with tip clearance. The increase of the standard changes.isThe standard is of small and changes little withwith tip the clearance. increase the deviation mainly due todeviation the decline signal-to-noise ratio (SNR) receivedThe power. Thus,ofthe standard deviation is mainly due to the decline of signal-to-noise ratio (SNR) with the received tip clearance change has little effect on the accuracy of the microwave BTT system. The rise time of power. Thus,sensor the tipsignal clearance change has little on the sensor, accuracy of theindicates microwave BTTspatial system. the fiber optic is less than that of theeffect microwave which a better The rise time of the fiber optic sensor signal is less than that of the microwave sensor, which indicates resolution. However, the accuracy of the microwave sensor is not significantly lower due to the timing a betterdescribed spatial resolution. the accuracy of the microwave sensor is not significantly lower method in chapterHowever, one. dueFigure to the 10 timing method described in chapter one. shows the vibration displacement of blade 1 measured by the microwave BTT system, Figure 10 shows the vibration displacement ofrpm. bladeThe 1 measured byregions the microwave BTT system, when the rotating speed is raised from 1000 to 5000 resonance can be found in the when the rotating speed is raised from 1000 to 5000 rpm. The resonance regions can be found in the picture, and the blade parameters such as natural frequency can be calculated by the resonance points. picture, andthe theengine blade orders parameters as natural points frequency can in bethe calculated the resonance For example, of thesuch two resonance marked figure arebyidentified by a points. For example, the engine orders of the two resonance points marked in the figure are identified single parameter method. The rotating speed of the resonance point at the first mark is 66.7 Hz and byorder a single parameter method. The speed of is the71.9 resonance at theisfirst mark is 66.7 Hz the is 14, the rotating speed of rotating the second mark Hz andpoint the order 13. The resonance and the order is 14, the rotating speed of the second mark is 71.9 Hz and the order is 13. The resonance frequency of the first-order bending vibration is 935.66 Hz according to the FEM model of the rotor, frequency of the first-order bending vibration 935.66 Hz of according the FEM model of the rotor, and the difference between the simulation resultisand results the BTT to system are small. and the difference between the simulation result and results of the BTT system are small.

Figure Figure10. 10.Vibration Vibrationdisplacement displacementofofblade blade1.1.

Theresults resultsshow showthat thatthe theproposed proposedmicrowave microwavesensor sensorBTT BTTmethod methodhas hasgood goodprecision, precision,and andthe the The effectofoftip tipclearance clearancechange changeononthe theBTT BTTaccuracy accuracyisissmall. small.Thus, Thus,the themicrowave microwaveBTT BTTsystem systemisisa a effect feasible and effective solution for monitoring blade vibrations in turbine engines. However, the feasible and effective solution for monitoring blade vibrations in turbine engines. However, the spatial spatial resolution of the optical sensorthan is better than the microwave sensor due to the width, small beam resolution of the optical sensor is better the microwave sensor due to the small beam and width, and the temperature stability of the microwave probe may also affect the measurement the temperature stability of the microwave probe may also affect the measurement accuracy. accuracy. 4. Conclusions 4. Conclusions A BTT system based on a microwave sensor is proposed. The system uses a 24-GHz microwave BTTan system on a of microwave is proposed. The system a 24-GHz sensorAwith outerbased diameter 8 mm andsensor a system signal bandwidth ofuses 10 MHz. The microwave proposed sensor with an outer diameter of 8 mm and a system signal bandwidth of 10 MHz. The proposed microwave BTT system is compared with a conventional BTT system based on a fiber optic sensor. The results show that the microwave BTT system can achieve good accuracy. Thus, the system is feasible for use in applications with a polluted medium and high temperatures, where other kinds of sensor are inappropriate. The improved timing method reduced the effect of tip clearance on BTT

Sensors 2017, 17, 1097

10 of 11

microwave BTT system is compared with a conventional BTT system based on a fiber optic sensor. The results show that the microwave BTT system can achieve good accuracy. Thus, the system is feasible for use in applications with a polluted medium and high temperatures, where other kinds of sensor are inappropriate. The improved timing method reduced the effect of tip clearance on BTT measurement. The system has enough bandwidth for multiple blades and high rotation speeds. The proposed BTT system and probe can also be used to measure tip clearance. Acknowledgments: This work was supported in part by the National High Technology Research and Development Program of China (2013AA102402), in part by National Natural Science Foundations of China under Grant No. 61501319, in part by the open project (MOMST2015-7) of Key Laboratory of Micro Opto-electro Mechanical System Technology, Tianjin University, in part by the National Marine economy innovation development area demonstration project No. cxsf2014-2 and Tianjin Natural Science Foundations of China (17JCQNJC01100). Author Contributions: Jilong Zhang and Guangyue Niu performed the experiments and analyzed the data. Fajie Duan, Jiajia Jiang and Jie Li conceived and supervised the experiments; Jilong Zhang prepared the manuscript. Conflicts of Interest: The authors declare no conflict of interest.

References 1. 2. 3. 4. 5.

6.

7. 8. 9. 10. 11.

12.

13. 14.

15.

Kumar, S.; Roy, N.; Ganguli, R. Monitoring low cycle fatigue damage in turbine blade using vibration characteristics. Mech. Syst. Signal Process. 2007, 21, 480–501. [CrossRef] Rao, A.R.; Dutta, B.K. Vibration analysis for detecting failure of compressor blade. Eng. Fail. Anal. 2012, 25, 211–218. [CrossRef] Salhi, B.; Lardies, J.; Berthillier, M.; Voinis, P.; Bodel, C. Modal parameter identification of mistuned bladed disks using tip timing data. J. Sound Vib. 2008, 314, 885–906. [CrossRef] Hanachi, H.; Liu, J.; Banerjee, A.; Koul, A.; Liang, M.; Alavi, E. Bladed disc crack diagnostics using blade passage signals. Meas. Sci. Technol. 2012, 23, 125003. [CrossRef] Diamond, D.H.; Heyns, P.S.; Oberholster, A.J. A comparison between three blade tip Timing algorithms for estimating synchronous turbomachine blade vibration. In Proceedings of the 9th World Congress on Engineering Asset Management (WCEAM), Pretoria, South Africa, 28–31 October 2014; pp. 215–225. Gallego-Garrido, J.; Dimitriadis, G.; Wright, J.R. A class of methods for the analysis of blade tip timing data from bladed assemblies undergoing simultaneous resonances—Part I: theoretical development. Int. J. Rotating Mach. 2007, 2007, 27247. [CrossRef] Hu, Z.; Lin, J.; Chen, Z.-S.; Yang, Y.-M.; Li, X.-J. A non-uniformly under-sampled blade tip-timing signal reconstruction method for blade vibration monitoring. Sensors 2015, 15, 2419–2437. [CrossRef] [PubMed] Wiseman, M.W.; Guo, T.H. An investigation of life extending control techniques for gas turbine engines. In Proceedings of the American Control Conference, Arlington, VA, USA, 25–27 July 2001; pp. 3706–3707. García, I.; Zubia, J.; Durana, G.; Aldabaldetreku, G.; Illarramendi, M.A.; Villatoro, J. Optical fiber sensors for aircraft structural health monitoring. Sensors 2015, 15, 15494–15519. [CrossRef] [PubMed] Guo, H.T.; Duan, F.J.; Cheng, Z. Numerical analysis of the blade tip-timing signal of a fiber bundle sensor probe. Opt. Eng. 2015, 54, 034103. [CrossRef] García, I.; Beloki, J.; Zubia, J.; Aldabaldetreku, G.; Illarramendi, M.A.; Jiménez, F. An Optical Fiber Bundle Sensor for Tip Clearance and Tip Timing Measurements in a Turbine Rig. Sensors 2013, 13, 7385–7398. [CrossRef] [PubMed] Pfister, T.; Büttner, L.; Czarske, J.; Krain, H.; Schodl, R. Turbo machine tip clearance and vibration measurements using a fibre optic laser doppler position sensor. Meas. Sci. Technol. 2006, 17, 1693–1705. [CrossRef] Michael, T. Method and Apparatus for Tracking a Rotating Blade Tip for Blade Vibration Monitor Measurements. Patent US7836772B2, 23 November 2010. Chana, K.S.; Cardwell, D.N. The use of eddy current sensor based blade tip timing for FOD detection. In Proceedings of the ASME Turbo Expo 2008: Power for Land, Sea, and Air, Berlin, Germany, 9–13 June 2008; pp. 169–178. [CrossRef] Cory, R.; Andreas, H.V. Passive Eddy Current Blade Detection Sensor. Patent US6927567B1, 9 August 2005.

Sensors 2017, 17, 1097

16. 17. 18. 19. 20.

21.

22.

23.

24.

11 of 11

Cory, R.; Andreas, H.V.; Benjam, L. Blade Detection Sensor Having an Active Cooling System. Patent US7170284B2, 30 January 2007. Prochazka, P.; Vanek, F. New methods of noncontact sensing of blade vibrations and deflections in turbomachinery. IEEE Trans. Instrum. Meas. 2014, 63, 1583–1592. [CrossRef] Lawson, C.P.; Ivey, P.C. Tubomachinery blade vibration amplitude measurement through tip timing with capacitance tip clearance probes. Sensor Actuators A Phys. 2005, 118, 14–24. [CrossRef] Gilles, B.; Samim, A. Capactive Sensor Including a Coaxial Cable and a Probe. Patent US5973502, 26 October 1999. Geisheimer, J.L.; Billington, S.A.; Holst, T.; Burgess, D.W. Performance testing of a microwave tip clearance sensor. In Proceedings of the AIAA Joint Propulsion Conference, Tucson, AZ, USA, 10–13 July 2005. [CrossRef] Ryszard, S.; Radosław, P.; Jarosław, S.; Edward, R.; Krzysztof, K.; Paweł, M. Application of Blade-Tip Sensors to Blade-Vibration Monitoring in Gas Turbines. In Thermal Power Plants, 1st ed.; Mohammad, R., Ed.; InTech: Rijeka, Croatia, 2012; Volume 8, pp. 145–176. Alexander, M.; Mikhail, B.; Maksim, S. Microwave blade tip clearance measurements: principles, current practices and future opportunities. In Proceedings of the ASME Turbo Expo 2012, Copenhagen, Denmark, 11–15 June 2012. [CrossRef] Wagner, M.; Schulze, A.; Vossiek, M.; Stephelbauer, C.; Weigel, R.; Vortmeyer, N.; Heide, P. Novel microwave vibration monitoring system for industrial power generating turbines. In Proceedings of the IEEE MTT-S International Microwave Symposium Digest, Baltimore, MD, USA, 7–12 June 1998; pp. 143–146. [CrossRef] Violetti, M.; Xu, Q.; Hochreutiner, O.; Skrivervik, A.K. New microwave sensor for on-line blade tip timing in gas and steam turbines. In Proceedings of the 2012 Asia-Pacific Microwave Conference (APMC 2012), Kaohsiung, Taiwan, 4–7 December 2012; pp. 1055–1057. [CrossRef] © 2017 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).