Tire-Pressure Identification Using Intelligent Tire with Three-Axis Accelerometer

sensors

Article Tire-Pressure Identiﬁcation Using Intelligent Tire with Three-Axis Accelerometer

Bing Zhu , Jiayi Han and Jian Zhao * State Key Laboratory of Automotive Simulation and Control, Jilin University, Changchun 130025, China; [email protected] (B.Z.); [email protected] (J.H.) * Correspondence: [email protected]

Received: 27 April 2019; Accepted: 4 June 2019; Published: 5 June 2019

Abstract: An intelligent tire uses sensors to dynamically acquire or monitor its state. It plays a critical role in safety and maneuverability. Tire pressure is one of the most important status parameters of a tire; it influences vehicle performance in several important ways. In this paper, we propose a tire-pressure identification scheme using an intelligent tire with 3-axis accelerometers. As the primary sensing system, the accelerometers can continuously and accurately detect tire pressure with less electronic equipment mounted in the tire. To identify tire pressure in real time during routine driving, we first developed a prototype for the intelligent tire with three 3-axis accelerometers, and carried out data-acquisition tests under different tire pressures. Then we filtered the data and concentrated on the vibration acceleration of the rim in the circumferential direction. After analysis, we established the relationship between tire pressure and characteristic frequency of the rim. Finally, we verified our identification scheme with actual vehicle data at different tire pressures. The results confirm that the identified tire pressure is very close to the actual value.

Keywords: intelligent tires; tire pressure; accelerometer; wheel vibration

1. Introduction Interactions between a tire and the road exert forces on a vehicle and alter its motion state. The ultimate aim of most vehicle active safety systems, such as the anti-lock braking system (ABS) or the electronic stability program (ESP), is to make the tires controllable [1–3]. The status information of the tire is especially important for all kinds of active safety systems and especially for self-driving vehicles [4,5]. At present, tire information, such as tire force and coefficient of road adhesion, is estimated by modules and filters [6–10]. During estimation, there is much simplification and linearization in the models of the vehicle, wheel, and tires, and the kinematic state of the vehicle is indeterminate. Therefore, application of estimated tire information is always restricted. To acquire more accurate and appropriate tire information, the concept of an intelligent tire system is being researched extensively [11,12]. The intelligent tire, which is equipped with sensors and other electronic equipment, is meant to measure the tire pressure, deformation, force, and other information. Its most important part is the sensing system [13,14]. Intelligent tires can be classified into five major types, based on their sensing methods, which are accelerometers, optical sensors, strain sensors, polyvinylidene fluoride (PVDF) sensors, and other sensors. Yi Xiong et al. used an optical tire sensor-based laser to measure rolling tire deformation [15]. Matsuzaki et al. calculated the in-plane strain and out-of-plane displacement with a single CCD camera fixed on the wheel rim [16]. Optical sensors can achieve noncontact measurement, hence they have no adverse effect on the tire rubber, but these sensors are expensive and consume too much power, and wheel vibration can influence their accuracy.

Sensors 2019, 19, 2560; doi:10.3390/s19112560 www.mdpi.com/journal/sensors Sensors 2019, 19, 2560 2 of 13

Strain sensors and PVDF sensors have gained more attention in tire sensing [17–21]. Zhang et al. embedded a pressure-sensitive, electric conductive rubber sensor inside the tire rubber layer to extract the 3-D forces on the contact patch [17]. Armstrong et al. applied three piezoelectric film sensors, which are a kind of strain sensor, on the inner surface of a tire to detect normal pressure, deflection, and longitudinal strain [18]. Erdogan et al. proposed a method to estimate the tire-road friction coefficient using a sensor consisting of PVDF film attached to the surface of a cantilever beam [19]. However, these sensors occupy a large area of the inner surface of the tire, which will alter the elastic property of the rubber and cause abnormal measurements. In addition, the huge temperature variation of a tire has a strong impact on strain sensors and PVDF sensors. Accelerometers are now used widely for intelligent tires [22–25]. Savaresi et al. presented a direct approach to estimate the instantaneous vertical and longitudinal forces by in-tire acceleration measurements and standard wheel-speed sensors [22]. Matsuzaki et al. proposed an identification scheme for the friction coefficient involving the slip angle, and applied force estimation through the use of a 3-axis accelerometer glued on the inner surface of a tire [23]. Niskanen et al. studied the distorted contact area of a tire in wet conditions using three accelerometers fixed on its inner liner [24]. Accelerometers are compact, small, and weigh just a few grams. They hardly influence a tire when glued to its inner surface. Accelerometers also consume less energy than optical or other sensors. This makes it possible to achieve a passive intelligent tire. In addition, their output signals are easy to read and contain more information compared to PVDF or strain sensors. With all of these attractive features, accelerometers are the most suitable sensors for intelligent tires. Intelligent tires equipped with accelerometers, as described above, are mostly researched to identify and estimate tire deformation, slip angle, vertical or longitudinal force, and tire-road friction coefficients. While this information is vital, there has been little research on tire pressure using intelligent tires equipped with accelerometers. Tire pressure influences several important aspects of vehicle performance [26]. Rolling resistance increases when a tire loses pressure, which reduces fuel economy. Pressure that is too low results in stationary wear, which can cause a tire to burst. When the tire pressure is high, the tire and road contact patch will be non-uniform, which reduces adhesive force. In addition, the tire will harden, which leads to an uncomfortable ride for passengers. The tire pressure also makes sense to vehicle active safety or control applications. The cornering stiffness is influenced by the tire pressure greatly. The cornering stiffness of the front and rear tires determines the steering behavior directly [27,28]. In this consideration, the tire pressure has something to do with the active safety systems, such as ESP. The tire pressure monitoring in real time becomes more and more important for the present and the future. The vehicle control algorithm is more and more complexity, and more variables are taken into account during tire modelling. The influence of the tire pressure can be and must be taken into consideration. For example, Janulevicius et.al. investigated how driving wheels of a front-loaded vehicle interact with the terrain depending on tire pressure [29]. Toma et al. studied the influence of tire pressure on the results of diagnosing brakes and suspension [30]. With the real time tire pressure, the control strategy under different tire pressure can be optimized and enhances the vehicle performance [31]. Therefore, Tire-pressure monitoring should be the fundamental function of an intelligent tire. Tire-pressure monitor systems (TPMSs) are widely used all over the world [32,33]. There are two types of TPMS. One is called indirect TPMS and the other is direct TPMS. The indirect TPMS can work with no need for installing additional sensors. It uses the signal of the wheel speed sensors or other existing sensors [34]. This kind of TPMS is only able to judge whether the tire is lacking of pressure. It cannot provide the precise tire pressure in real time. Therefore, the use of the indirect TPMS is limited. It can only help to acquaint the drivers with the tire situation and it can do nothing with the advanced active safety systems [35]. The direct TPMS is based on a pressure sensor mounted at the tire nozzle. It is practical and durable for the present, but it is a separate system. More electronic circuits or equipment are acquired to combine this kind of TPMS with intelligent tire or advanced active safety systems. This will reduce the reliability and add complexity. If we can estimate the tire Sensors 2019, 19, x 3 of 14 function is developed by equipping accelerometer or other vibration sensors, the tire pressure identificationSensors 2019 method, 19, 2560 proposed in this paper will help to take the place of the present TPMS.3 of 13 The major motivation to install accelerometers in the tire may be to estimate tire deformation, slip angle or frictionalpressure force. precisely In this in real condition, time using if thewe sensors can id installedentify the in thetire intelligent pressure tire, by a the seamless accelerometers tire pressure which is alreadymonitoring installed function in the tire, will bethen achieved. pressure In sensors the future in whenthe tire the will intelligent be unnecessary. tire with multi-function The less electronic equipmentis developed we mount by equipping in the tire, accelerometer the more orpractical other vibration and reliable sensors, an the intelligent tire pressure tire identification system we can achievemethod [5,36]. proposed in this paper will help to take the place of the present TPMS. The major motivation Into this install paper, accelerometers we have tried in the to tire identify may be tothe estimate tire pressure tire deformation, using 3-axis slip angle accelerometers or frictional force.in the tire. In this condition, if we can identify the tire pressure by the accelerometers which is already installed in We designed a development platform for an intelligent tire, and conducted road tests under different the tire, then pressure sensors in the tire will be unnecessary. The less electronic equipment we mount tire pressuresin the tire, thefor more data practical acquisition and reliable using an intelligentthe platform. tire system We we canmathematically achieve [5,36]. processed the accelerometerIn this data paper, and we developed have tried to an identify identificati the tireon pressure scheme using for 3-axis tire accelerometerspressure by inanalyzing the tire. the characteristicsWe designed in athe development frequency platform domain. for anTo intelligent verify tire,our andtire-pressure conducted road identification tests under di scheme,fferent we conductedtire pressures another forroad data test acquisition was carried using out the with platform. different We mathematically tire pressures processed and compared the accelerometer the identified and actualdata andtire developedpressures. an identification scheme for tire pressure by analyzing the characteristics in the frequency domain. To verify our tire-pressure identification scheme, we conducted another road test was carried out with different tire pressures and compared the identified and actual tire pressures. 2. Intelligent Tire with 3-Axis Accelerometers and Data-Acquisition Test 2. Intelligent Tire with 3-Axis Accelerometers and Data-Acquisition Test 2.1. Test Equipment and Method 2.1. Test Equipment and Method To discover the pattern of wheel- and tire-vibration acceleration, we made a prototype for an To discover the pattern of wheel- and tire-vibration acceleration, we made a prototype for an intelligent tire with 3-axis accelerometers; the overall structure is shown in Figure 1. The wheel was intelligent tire with 3-axis accelerometers; the overall structure is shown in Figure1. The wheel was 16 × 6.516 J 6.5and J and the the tire tire waswas a 205a 205/55R16/55R16 (Michelin, (Michelin, Clermont-Ferrand, Clermont-Ferrand, France). The France). wheel part The contained wheel part × containeda wheel a wheel with with connector connector and accelerometers, and accelerometers, and a slip and ring a moduleslip ring to module conduct anto conduct electrical signal.an electrical signal.The The cabin cabin part part contained contained a sensor a adapter, sensor data-acquisition adapter, data-acquisition system, host computer, system, and host inertial computer, system and inertialand system global and navigation global satellitenavigation system satellite (GNSS). system (GNSS).

FigureFigure 1. 1.OverallOverall structure structure ofof intelligent intelligent tire. tire.

Weused three 3-axis accelerometers to measure the vibration acceleration of different parts. The 3-axis We used three 3-axis accelerometers to measure the vibration acceleration of different parts. The accelerometers (SAE30050Q, Yutian, Wuxi, China) were small and light (9.5 mm 9.5 mm 9.5 mm, 6 g), × × 3-axis withaccelerometers measurement ranges(SAE30050Q,500 g and 2~5000Yutian, Hz. AccelerometerWuxi, China) C was ﬁxedwere to thesmall well of theand rim, light ± (9.5 mmand × accelerometers9.5 mm × 9.5 Amm, and 6 B g), were with respectively measurement ﬁxed to theranges internal ±500 surface g and of the2~5000 tread Hz. and theAccelerometer sidewall C was fixedof the to tire. the Thewell setup of the of the rim, accelerometers and accelerometers and the coordinates A and B ofwere the wheelrespectively are shown fixed in Figure to the2. internal surface of the tread and the sidewall of the tire. The setup of the accelerometers and the coordinates of the wheel are shown in Figure 2.

Sensors 2019, 19, x www.mdpi.com/journal/sensors Sensors 2019, 19, x 4 of 14

Sensors 2019, 19, 2560 4 of 13 Sensors 2019, 19, x 4 of 14

Figure 2. Accelerometer setup and tire coordinates.

To export the signal from inside the tire, we drilled on the well and installed a sealed connection socket, as shown in Figure 3a. The maximum sealing pressure of the socket was 60 bars, which exceeded the maximum tire pressure of 3.5 bars. The socket introduced no possibility of leaks from the tire and transmitted electrical signals without loss. Because a wheel constantly spins during driving, we designed a slip ring module to transmit the signal from the wheel to the car body, as shown in Figure 3b. The slip ring was assembled in a circular plate, which was attached to the wheel hub by five joint cylinders, each with a claw at one end to clamp the nut of the hub bolt and a threaded hole at the other end to attach the circular plate. We used integral electronic piezoelectric (IEPE) accelerometers, so a sensor adapter (CT5210, CHENGTEC, Shanghai, China) was needed to supply power and a condition signal, as shown in Figure 3e. An RT3000 (OxTS, Middleton Stoney, United Kingdom) combined inertial and GNSS system was installed in the vehicle, as shown in Figure 3f, to acquire vehicle dynamic states, such as velocity, acceleration, and yaw rate. The velocity accuracy was 0.05 km/h, and the data frequency was 100 Hz. We used a MicroAutoBox II (dSPACE, Paderborn, Germany) data-acquisition system to record the vibration acceleration, vehicle speed, and other data, as shown in Figure 3d. This device has 32 Figure 2. Accelerometer setup and tire coordinates. 16-bit analog input channels with a 200 Ksps conversion rate and −10~10V input voltage range, eight To export the signal from inside the tire, we drilled on the well and installed a sealed connection 16-bitTo analog export output the signal channels, from insideand six the controller tire, we drilled area network on the well (CAN) and channels. installed aThe sealed outputs connection of the socket, as shown in Figure3a. The maximum sealing pressure of the socket was 60 bars, which exceeded accelerometerssocket, as shown were in conditionedFigure 3a. The by themaximum adapter sealanding then pressure input to of the the MicroAutoBox socket was 60 II bars,through which an the maximum tire pressure of 3.5 bars. The socket introduced no possibility of leaks from the tire and analog-to-digitalexceeded the maximum converter. tire Thepressure RT3000 of 3.5and bars. MicroAutoBox The socket introducedII communicated no possibility by CAN of (Controller leaks from transmitted electrical signals without loss. Areathe tire Network) and transmitted bus. The electricaldata were signals acquired without through loss. a CAN channel. Because a wheel constantly spins during driving, we designed a slip ring module to transmit the signal from the wheel to the car body, as shown in Figure 3b. The slip ring was assembled in a circular plate, which was attached to the wheel hub by five joint cylinders, each with a claw at one end to clamp the nut of the hub bolt and a threaded hole at the other end to attach the circular plate. We used integral electronic piezoelectric (IEPE) accelerometers, so a sensor adapter (CT5210, CHENGTEC, Shanghai, China) was needed to supply power and a condition signal, as shown in Figure 3e. An RT3000 (OxTS, Middleton Stoney, United Kingdom) combined inertial and GNSS system Sensorswas installed 2019, 19, xin the vehicle, as shown in Figure 3f, to acquire vehicle dynamic states, such as velocity,5 of 14 (a) (b) (c) acceleration, and yaw rate. The velocity accuracy was 0.05 km/h, and the data frequency was 100 Hz. SensorsWe 2019 used, 19, x a MicroAutoBox II (dSPACE, Paderborn, Germany) data-acquisitionwww.mdpi.com/jou systemrnal/sensors to record the vibration acceleration, vehicle speed, and other data, as shown in Figure 3d. This device has 32 16-bit analog input channels with a 200 Ksps conversion rate and −10~10V input voltage range, eight 16-bit analog output channels, and six controller area network (CAN) channels. The outputs of the accelerometers were conditioned by the adapter and then input to the MicroAutoBox II through an analog-to-digital converter. The RT3000 and MicroAutoBox II communicated by CAN (Controller Area Network) bus. The data were acquired through a CAN channel.

(d) (e) (f)

Figure 3. TestTest setup and equipment: ( a)) Connector Connector on on the the rim rim well; well; ( (b)) slip slip ring ring module; module; ( (c)) connector connector on thethe vehiclevehicle body; body; (d ()d data-collection) data-collection system; system; (e) battery(e) battery and and adapter; adapter; (f) combined (f) combined inertial inertial and global and globalnavigation navigation satellite satellite system system (GNSS) (GNSS) system. system.

2.2. Vehicle Test under Different Tire Pressures (a) (b) (c) To determine how to identify the tire pressure in an intelligent tire, we conducted an experiment inSensors which 2019 the, 19, xtest vehicle traveled on an urban road with different tire pressures.www.mdpi.com/jou The tirernal/sensors pressure was set at 150 kPa, 250 kPa, and 350 kPa. To include a variety of road surfaces and to develop a practical and universal method for tire pressure identification, we chose a road of about 2 km, including asphalt, cement, damaged, and snow-covered road surfaces, and the test vehicle traveled at speeds from 0 to 50 km/h in a natural fashion at each tire pressure. The weather was sunny and the temperature was about −20 °C. A truncated section of the resultant data is shown in Figure 4. The data include nine channels of vibration acceleration from three 3-axis accelerometers, and the velocity, acceleration, yaw rate, and other parameters were acquired synchronously.

Figure 4. Vibration acceleration of rim well, tire tread, and sidewall in three directions.

Sensors 2019, 19, x www.mdpi.com/journal/sensors Sensors 2019, 19, 2560 5 of 13

Because a wheel constantly spins during driving, we designed a slip ring module to transmit the Sensorssignal 2019 from, 19 the, x wheel to the car body, as shown in Figure3b. The slip ring was assembled in a circular5 of 14 plate, which was attached to the wheel hub by ﬁve joint cylinders, each with a claw at one end to clamp the nut of the hub bolt and a threaded hole at the other end to attach the circular plate. We used integral electronic piezoelectric (IEPE) accelerometers, so a sensor adapter (CT5210, CHENGTEC, Shanghai, China) was needed to supply power and a condition signal, as shown in Figure3e. An RT3000 (OxTS, Middleton Stoney, United Kingdom) combined inertial and GNSS system was installed in the vehicle, as shown in Figure3f, to acquire vehicle dynamic states, such as velocity, acceleration, and yaw rate. The velocity accuracy was 0.05 km/h, and the data frequency was 100 Hz. We used a MicroAutoBox II (dSPACE, Paderborn, Germany) data-acquisition system to record the vibration acceleration, vehicle speed, and other data, as shown in Figure3d. This device has 32 16-bit analog input channels with a 200 Ksps conversion rate and 10~10V inputvoltage range, (d) (e) − (f) eight 16-bit analog output channels, and six controller area network (CAN) channels. The outputs of the accelerometersFigure 3. Test setup were and conditioned equipment: by(a) Connector the adapter on andthe rim then well; input (b) slip to the ring MicroAutoBox module; (c) connector II through an analog-to-digitalon the vehicle body; converter. (d) data-collection The RT3000 system; and MicroAutoBox (e) battery andII adapter; communicated (f) combined by CAN inertial (Controller and Areaglobal Network) navigation bus. Thesatellite data system were acquired(GNSS) system. through a CAN channel.

2.2. Vehicle Vehicle Test Test under DiDifferentfferent Tire PressuresPressures To determine how to identify the tire pressure in an intelligent tire, we conducted an experiment experiment in which thethe testtest vehiclevehicle traveled traveled on on an an urban urban road road with with di ffdifferenterent tire tire pressures. pressures. The The tire pressuretire pressure was wasset at set 150 at kPa, 150 250kPa, kPa, 250 andkPa, 350 and kPa. 350 To kPa. include To incl a varietyude a variety of road of surfaces road surfaces and to develop and to adevelop practical a practicaland universal and methoduniversal for tiremethod pressure for identification, tire pressure we identification, chose a road of we about chose 2 km, a including road of asphalt,about 2cement, km, including damaged, asphalt, and snow-covered cement, damaged, road surfaces, and snow-covered and the test vehicleroad surfaces, traveled and at speedsthe test from vehicle 0 to traveled50 km/h at in speeds a natural from fashion 0 to 50 at km/h each in tire a natural pressure. fashion The weather at each tire was pressure. sunny and The the weather temperature was sunny was andabout the 20temperature◦C. was about −20 °C. − A truncated section section of of the the resultant resultant data data is is shown shown in in Figure Figure 4.4 .The The data data in includeclude nine nine channels channels of vibrationof vibration acceleration acceleration from from three three 3-axis 3-axis accelerome accelerometers,ters, and andthe velocity, the velocity, acceleration, acceleration, yaw yawrate, rate,and otherand other parameters parameters were were acquired acquired synchronously. synchronously.

FigureFigure 4. 4. VibrationVibration acceleration acceleration of of rim rim well, well, tire tire tread, and sidewall in three directions.

Sensors 2019, 19, x www.mdpi.com/journal/sensors Sensors 2019, 19, 2560 6 of 13

3. Identification Scheme for Tire Pressure The wheel, which is the combination of the rim and tire, is a kind of spring-damp system. That means a change of stiffness will alter the characteristic frequency of the wheel. According to the spring-loaded ring tire model in the vibration model [37,38], the relationship between torsion vibration natural frequency and tire stiffness can be derived as: s 1 Kθ ftor = , (1) 2π ρbl where ftor is the torsion natural frequency, which can reflect the characteristics of the circumferential vibration; Kθ is the rotational spring constant of the sidewall portion per unit length, representing the tire and wheel stiffness; ρ is the density; and b and l are the thickness and half-length, respectively, of the tread ring. From Equation (1), we can see that the natural frequency increases with the stiffness. Furthermore, the stiffness of the wheel can be influenced by the tire pressure. Another tire model from Akasaka illustrates the relationship between stiffness and tire pressure [39]. The theoretical formula of the rotational spring constant about p is [38]:

Kt = g(rB, rc, rD, ϕD, αD, θ) p + h(h, rB, rD, ϕD, Vr, Gr), (2) · where Kt is the rotational spring constant, which is nearly the same as Kθ in the spring-loaded ring tire mode; p is the tire pressure; rB, rC, and rD are the tire radius parameters related to special tire positions; ϕD, and αD are the angle parameters; θ is the rotational displacement; Vr is content rate; Gr is the rigidity modulus; and h is the thickness. Equation (2) is used to illustrate the relationship between the stiffness and tire pressure, so there is no need to derive it in detail. Generally, the stiffness of the wheel increases with the tire pressure and when the tire pressure continues to increase, the stiffness will remain unchanged or will decrease slightly. According to this principle, we can identify the tire pressure by mining the data provided by the intelligent tire, and we can determine the characteristic frequency. Above all, it is important to determine how the tire pressure influences the stiffness and establish an accurate relationship between characteristic frequency and tire pressure.

3.1. Vibration Analysis of the Tire Previous research has shown that one of the characteristic frequencies of the wheel is between 35 Hz and 80 Hz [40,41]. Therefore, we concentrated on analyzing the amplitude-frequency characteristic in this range. We first analyzed the vibration acceleration data under different tire pressures from accelerometers A and B. A piece of data under a steady speed was selected to show the frequency component of the tire vibration. We used the fast Fourier transform (FFT) algorithm to achieve a time-frequency transformation. FFT is an improved algorithm for the discrete Fourier transform (DFT). The main formulas are as follows: k X(k) = X1(k) + W X2(k) N , (3) X k + N = X (k) Wk X (k) 2 1 − N 2 where: P X (k) = N/2 1 x (r)Wkr 1 r=0 − 1 N/2 P , (4) X (k) = N/2 1 x (r)Wkr 2 r=0 − 2 N/2 where: 2π kn j N kn WN = e− , (5) Sensors 2019, 19, 2560 7 of 13 Sensors 2019, 19, x 7 of 14 where W is the twiddle factor, x1(r) and x2(r) are subsequences of the time series data dividing by the where WW isis the the twiddle twiddle factor, factor, xx11(r()r )and and xx2(2r()r are) are subsequences subsequences of of the the time time series series data data dividing dividing by by the the odevityodevity of of the the ordinal ordinal number. number. XX(k(k) )and and XX(k( k++ N/2N/)2 are) are the the frequency frequency domain domain results. results. TheThe length ofof thethe datadata was was 5 5 seconds, seconds, and and the the sample sample frequency frequency was was 1 kHz. 1 kHz. The The results results are shown are shownin Figures in Figures5 and6. 5 The and semitransparent 6. The semitransparent spectrums spectrums are the original are the original results of results the FFT, of the and FFT, the weightedand the weightedcurves are curves the smoothed are the smoothed results by re Gaussiansults by Gaussian smoothing smoothing filter. The filter. vibrations The vibrations of the tire of tread the tire were treadalmost were alike almost in three alike directions in three under directions different under tire di pressures,fferent tire and pressures, the tire and side the wall tire was side the wall same. was It is the same. It is difficult to distinguish between tire pressures through spectral analysis. We speculated thediffi same.cult to It distinguishis difficult to between distinguish tire between pressures tire through pressures spectral through analysis. spectral We analysis. speculated We speculated that the tire that the tire had filtered the vibration through its elastic property. The changes in tire pressure did thathad the filtered tire had the vibrationfiltered the through vibration its elasticthrough property. its elastic The property. changes The in tire changes pressure in tire did pressure not lead did to the not lead to the obvious differences of the vibration characteristic of the tire. notobvious lead to di fftheerences obvious of thedifferences vibration of characteristicthe vibration ofcharacteristic the tire. of the tire.

(a) (b) (c)

FigureFigure 5. Amplitude-frequencyAmplitude-frequency curve curve of ofaccelerometer accelerometer A A(tire (tire tread) tread) in inthree three directions: directions: ((aa)) Circumferential Circumferential direction; direction; (b ()b axial) axial direction; direction; (c) ( cradial) radial direction. direction.

(a) (b) (c)

FigureFigure 6. Amplitude-frequencyAmplitude-frequency curve curve of ofaccelerometer accelerometer B B(tire (tire sidewall) sidewall) in inthree three directions: directions: ((aa)) Circumferential Circumferential direction; direction; ( (bb) )axial axial direction; direction; (c (c) )radial radial direction. direction. 3.2. Vibration Analysis of the Rim 3.2. Vibration Analysis of the Rim We analyzed the data from accelerometer C, which reflected the vibration of the rim. As a rigid We analyzed the data from accelerometer C, which reflected the vibration of the rim. As a rigid body, the rim will vibrate with a regular pattern when excited by the road surface. The data of body, the rim will vibrate with a regular pattern when excited by the road surface. The data of accelerometer A were processed in the same way. The result is shown in Figure7. It is observed that accelerometer A were processed in the same way. The result is shown in Figure 7. It is observed that the axial and radial vibration amplitude-frequency curves remain the same for accelerometers A and B, the axial and radial vibration amplitude-frequency curves remain the same for accelerometers A and reflecting no distinction. However, the result in the circumferential direction reveals a difference in the B, reflecting no distinction. However, the result in the circumferential direction reveals a difference peak frequency at different tire pressures. The peak frequency rises with the tire pressure. The changes in the peak frequency at different tire pressures. The peak frequency rises with the tire pressure. The in tire pressure are reflected in the peak frequency of the amplitude- frequency curve. Therefore, changes in tire pressure are reflected in the peak frequency of the amplitude- frequency curve.

Sensors 2019, 19, x www.mdpi.com/journal/sensors Sensors 2019, 19, 2560 8 of 13 Sensors 2019, 19, x 8 of 14

Therefore,the circumferential the circumferential vibration vibration of the rim of becomes the rim becomes the key for the tire-pressure key for tire-pressure identiﬁcation. identification. To prove Tothat prove the change that the of change the peak of frequency the peak isfrequency not occasional, is not weoccasional, analyzed we the analyzed relationship the betweenrelationship peak betweenfrequency peak and frequency tire pressure and overtire pressure the whole over vehicle the whole test. vehicle test.

(a) (b) (c)

FigureFigure 7. 7. Amplitude-frequencyAmplitude-frequency curve curve of ofaccelerometer accelerometer C C(rim (rim well) well) in inthree three directions: directions: (a(a) )Circumferential Circumferential direction; direction; (b (b) )axial axial direction; direction; (c ()c )radial radial direction. direction.

3.3.3.3. Relationship Relationship Between Between Characteristic Characteristic Frequency Frequency and and Tire Tire Pressure Pressure TheThe circumferential circumferential accelerationacceleration datadata of of accelerometer accelerometer C underC under tire pressurestire pressures of 150 of kPa, 150 250 kPa, kPa, 250and kPa, 350 and kPa 350 were kPa segmented were segmented every every five seconds, five seconds, and eachand each segment segment was was processed processed through through FFT. FFT.We truncatedWe truncated the amplitude-frequency the amplitude-frequency curves curves from35 from Hz to35 80 Hz Hz, to then 80 Hz, smoothed then smoothed and normalized and normalizedeach curve each and plottedcurve and them plotted in chronological them in chronologi order.cal The order. result The is shown result is in shown Figure in8. Figure The vibration 8. The vibrationacceleration acceleration intensity intensity at different at different frequency frequency is represented is represented by colors. by colors. The intensityThe intensity increases increases from fromblue blue to yellow. to yellow. The The red red weighted weighted straight straight lines lines represent represent thethe average peak peak positions positions at at different different tiretire pressures. pressures.

(a) (b) (c)

FigureFigure 8. 8. SegmentedSegmented amplitude-frequency amplitude-frequency curve curve of of acce accelerometerlerometer A A(rim (rim well) well) inin circumferential circumferential direction:direction: (a ()a )150 150 kPa; kPa; (b (b) 250) 250 kPa; kPa; (c ()c 350) 350 kPa. kPa.

FigureFigure 88 showsshows the the peak peak values values arou aroundnd 45 45 Hz Hz at at a a tire tire pressure pressure of of 150 150 kPa. kPa. As As the the tire tire pressure pressure increased,increased, the the peak peak values values went went to to about about 50 50 Hz Hz wh whenen the the tire tire pressure pressure was was 250 250 kPa. kPa. When When the the tire tire pressurepressure increased increased to to 350 350 kPa, kPa, the the peak peak values values gath gatheredered around around 53 53 Hz. Hz. This This tr trendend is isconsistent consistent with with thethe analysis analysis presented presented atat thethe beginningbeginning of of this this section. section. That That is tois say,to say, the the characteristic characteristic frequency frequency about aboutwhich which the peak the valuespeak values gathered gathered is an inherent is an inherent attribute attribute of the wheel.of the wheel. It will It not will be not aﬀected be affected by external by externalfactors suchfactors as vehicle such as speed, vehicle but increasesspeed, but with increa tire pressureses with withintire pressure a certain within range. Therefore,a certain range. we next Therefore,sought to establishwe next ansought accurate to ruleestablish or relationship an accurate between rule or characteristic relationship frequency between and characteristic tire pressure. frequency and tire pressure.

Sensors 2019, 19, x www.mdpi.com/journal/sensors SensorsSensors 20192019, 19, 19, x, 2560 9 9of of 14 13

To determine the characteristic frequency at different tire pressures, it is practical and effective To determine the characteristic frequency at different tire pressures, it is practical and effective to to combine all segments of amplitude-frequency curves and locate the peak value. However, we first combine all segments of amplitude-frequency curves and locate the peak value. However, we first must must consider how the vehicle speed affects the amplitude of the amplitude-frequency curves, or the consider how the vehicle speed affects the amplitude of the amplitude-frequency curves, or the features features of the amplitude-frequency curves at low vehicle speed will be neglected. Because the of the amplitude-frequency curves at low vehicle speed will be neglected. Because the vibration vibration acceleration intensity is influenced greatly by the speed. The wheel and tire will be excited acceleration intensity is influenced greatly by the speed. The wheel and tire will be excited more more greatly when the speed goes higher [42]. We will use the segmented amplitude-frequency data greatly when the speed goes higher [42]. We will use the segmented amplitude-frequency data at at 350 kPa to illustrate the effect of vehicle speed on peak amplitude. We found that the peak 350 kPa to illustrate the effect of vehicle speed on peak amplitude. We found that the peak amplitude amplitude is directly proportional to the square of the vehicle speed, as shown in Figure 9. is directly proportional to the square of the vehicle speed, as shown in Figure9.

FigureFigure 9. 9. RelationshipRelationship between between peak peak amplitude amplitude and and vehicle vehicle speed. speed.

AccordingAccording to tothis this relationship, relationship, we developed we developed a function a functionto calculate to the calculate weighted the amplitude- weighted frequencyamplitude-frequency curves at diffe curvesrent tire at di pressures:ﬀerent tire pressures: 𝐺 =X∑ 1𝐺 , G ,= G, ,(6) (6) sum,tp 2 tp,i vi where Gsum,tp is the weighted amplitude-frequency curve (tp = 150 kPa, 250 kPa, 350 kPa); Gtp,i is the segmentedwhere Gsum,tp amplitude-frequencyis the weighted amplitude-frequency curve (i = 1,2,3,…); 1/v curvei2 is the (tp =weight150 kPa, function 250 kPa, to eliminate 350 kPa); theGtp,i effectis the 2 ofsegmented vehicle speed; amplitude-frequency and vi is the average curve vehicle (i = 1,2,3, speed... during); 1/vi is the the current weight segment. function to eliminate the eﬀect of vehicleThe weighted speed; and amplitude-frequencyvi is the average vehicle curves speed at duringdifferent the tire current pressures segment. were smoothed and normalizedThe weighted for convenient amplitude-frequency observation. The curves results at are different shown tire in pressuresFigure 10. were The smoothedcharacteristic and frequenciesnormalized at for150convenient kPa, 250 kPa, observation. and 350 kPa The are easi resultsly observed are shown to be in 45.4 Figure Hz, 1050.2. TheHz, characteristicand 53.2 Hz, respectively.frequencies at 150 kPa, 250 kPa, and 350 kPa are easily observed to be 45.4 Hz, 50.2 Hz, Sensors 2019, 19, x 10 of 14 and 53.2 Hz, respectively.

FigureFigure 10. Weighted10. Weighted amplitude-frequency amplitude-frequency curves at at different diﬀerent tire tire pressures. pressures.

We used frequency as the independent variable and tire pressure as the dependent variable, and Sensors 2019used, 19 the, x least square method to fit the relationship between characteristicwww.mdpi.com/jou frequency andrnal/sensors tire pressure. The curve is shown in Figure 11, and the relation function is shown as Equation (7).

Figure 11. Relationship between characteristic frequency and tire pressure.

𝑝 =1.60⋅𝑓 − 132.37 ⋅ 𝑓 + 2856.54, (7)

where ptire is the identified tire pressure and fpeak is the characteristic frequency.

4. Verification and Discussion To verify the scheme developed in Section 3, we conducted another experiment, which was the same as the experiment in Section 2, except that the tire pressures were set at 200 kPa and 300 kPa. We first broke the circumferential vibration acceleration data into small segments of five-seconds duration. We transformed each segment from the time to frequency domain by FFT, smoothed the amplitude-frequency curve, and extracted the peak frequency. Figure 12 shows the result at 200 kPa.

Sensors 2019, 19, x www.mdpi.com/journal/sensors Sensors 2019, 19, x 10 of 14

Sensors 2019, 19, 2560 10 of 13

We used frequency as the independent variable and tire pressure as the dependent variable, and used the least square method to ﬁt the relationship between characteristic frequency and tire Figure 10. Weighted amplitude-frequency curves at different tire pressures. pressure. The curve is shown in Figure 11, and the relation function is shown as Equation (7).

We used frequency as the independent2 variable and tire pressure as the dependent variable, and ptire = 1.60 f 132.37 fpeak + 2856.54, (7) used the least square method to fit the· peak relation− ship· between characteristic frequency and tire pressure. The curve is shown in Figure 11, and the relation function is shown as Equation (7). where ptire is the identiﬁed tire pressure and fpeak is the characteristic frequency.

. Figure 11. Relationship between characteristic frequency and tire pressure. Figure 11. Relationship between characteristic frequency and tire pressure. 4. Veriﬁcation and Discussion

To verify the scheme developed𝑝 =1.60⋅ in Section𝑓3,− we 132.37 conducted ⋅ 𝑓 another+ 2856.54 experiment,, which was the(7) same as the experiment in Section2, except that the tire pressures were set at 200 kPa and 300 kPa. where ptire is the identified tire pressure and fpeak is the characteristic frequency. We ﬁrst broke the circumferential vibration acceleration data into small segments of ﬁve-seconds duration. We transformed each segment from the time to frequency domain by FFT, smoothed the Sensors4. Verification 2019, 19, x and Discussion 11 of 14 amplitude-frequency curve, and extracted the peak frequency. Figure 12 shows the result at 200 kPa. To verify the scheme developed in Section 3, we conducted another experiment, which was the same as the experiment in Section 2, except that the tire pressures were set at 200 kPa and 300 kPa. We first broke the circumferential vibration acceleration data into small segments of five-seconds duration. We transformed each segment from the time to frequency domain by FFT, smoothed the amplitude-frequency curve, and extracted the peak frequency. Figure 12 shows the result at 200 kPa.

Sensors 2019, 19, x www.mdpi.com/journal/sensors

(a) (b)

FigureFigure 12. 12. ResultResult at at 200 200 kPa: kPa: (a (a) )Peak-frequency Peak-frequency extraction; extraction; ( (bb)) peak peak frequency frequency distribution distribution curve. curve.

WeWe can can infer infer that that there there is is no no relation relation between between peak peak frequency frequency and and vehicle vehicle speed. speed. Whatever Whatever speed speed thethe vehicle vehicle travels, travels, the the peak peak frequency frequency is is steady steady an andd easily determined. This This is is consistent consistent with with the the analysisanalysis in in Section Section 33.. The probability of an identiﬁcationidentification result whose error is belowbelow 5%5% comparedcompared withwith the the actual actual tire tire pressure 200 kPa is 85%. The The average average value value of of the the peak peak frequency frequency during during the the experiments was 48.01 Hz. Using the result of Section 3, the identified tire pressure is determined as 195.39 kPa through Equation (4). The identification error is 2.31%. There is no denying that the instantaneous peak-frequency extraction is not accurate enough, so we analyzed the errors, as shown in Figure 12b. The errors are normally distributed. So, we can enhance the identification precision by averaging several continuous peak frequencies. We did the same with the data at 300 kPa. Figure 13a shows the identification of the peak frequency. The probability of an identification result whose error is below 5% compared with the actual tire pressure of 300 kPa is 83%. The average peak frequency was 52.02 Hz, and the identified tire pressure is computed as 300.38 kPa through Equation (4). The identification error is 0.13%. Figure 13b shows the error distribution of the peak-frequency identification. It is much the same as the result at 200 kPa.

(a) (b)

Figure 13. Result at 300 kPa: (a) Peak-frequency extraction; (b) peak frequency distribution curve.

Sensors 2019, 19, x www.mdpi.com/journal/sensors Sensors 2019, 19, x 11 of 14

(a) (b)

Figure 12. Result at 200 kPa: (a) Peak-frequency extraction; (b) peak frequency distribution curve.

We can infer that there is no relation between peak frequency and vehicle speed. Whatever speed the vehicle travels, the peak frequency is steady and easily determined. This is consistent with the analysisSensors 2019 in, 19 Section, 2560 3. The probability of an identification result whose error is below 5% compared11 of 13 with the actual tire pressure 200 kPa is 85%. The average value of the peak frequency during the experiments was 48.01 Hz. Using the result of Section 3, the identified tire pressure is determined as 195.39experiments kPa through was 48.01 Equation Hz. Using (4). The the identification result of Section error3, the is 2.31%. identified tire pressure is determined as 195.39There kPa throughis no denying Equation that (4).the instantaneous The identification peak-fr errorequency is 2.31%. extraction is not accurate enough, so we analyzedThere is nothe denyingerrors, as that shown the instantaneous in Figure 12b. peak-frequency The errors are extractionnormally distributed. is not accurate So, enough,we can enhanceso we analyzed the identification the errors, precision as shown by in averaging Figure 12 b.several The errors continuous are normally peak frequencies. distributed. So, we can enhanceWe thedid identificationthe same with precision the data by at averaging 300 kPa. several Figure continuous13a shows peakthe identification frequencies. of the peak frequency.We did The the sameprobability with the ofdata an identification at 300 kPa. Figure result 13 awhose shows error the identification is below 5% of compared the peakfrequency. with the actualThe probability tire pressure of anof 300 identification kPa is 83%. result The aver whoseage errorpeak isfrequency below 5% was compared 52.02 Hz, with and thethe actualidentified tire tirepressure pressure of 300 is kPacomputed is 83%. Theas 300.38 average kPa peak through frequency Equation was 52.02 (4). Hz,The andidentification the identified error tire is pressure 0.13%. Figureis computed 13b shows as 300.38 the error kPa through distribution Equation of the (4). peak-frequency The identification identification. error is 0.13%. It is much Figure the 13 bsame shows as the result error distribution at 200 kPa. of the peak-frequency identification. It is much the same as the result at 200 kPa.

(a) (b)

Figure 13. ResultResult at at 300 kPa: ( (aa)) Peak-frequency Peak-frequency extraction; extraction; ( (bb)) peak peak frequency frequency distribution distribution curve. curve.

These results prove that the tire pressure identiﬁcation scheme using vibration acceleration we Sensorsdeveloped 2019, 19 is, ablex to estimate the tire pressure accurately. The speciﬁc valuewww.mdpi.com/jou of the tire pressurernal/sensors can be given in real time using this method, rather than a simple tire pressure level.

5. Conclusions To “intelligentize” a tire, which means that the tire itself is capable of sensing and communicating, we designed a prototype of an intelligent tire with three 3-axis accelerometers. To develop a tire-pressure identification scheme, we conducted a data-acquisition experiment using the prototype at different tire pressures and analyzing the wheel vibration in the frequency domain. We discovered a relationship between tire pressure and tire vibration. Generally, the peak frequency rises with the tire pressure. The identification scheme was verified with actual vehicle data at different tire pressures. The results confirm that the identified tire pressure is very close to the actual value. The probability of identification error below 5% is above 80%. The processing complexity and period are reasonable and practical.

Author Contributions: In this article, the experiments, calculations, and analysis were done by B.Z. and J.H. with advice and inspiration from J.Z. The detailed methodology was provided by J.H. The original draft was prepared by J.H. The review and editing were done by B.Z and J.Z. Funding: This work is partially supported by National Key R&D Program of China (2018YFB0105103). National Natural Science Foundation of China (51575225, 51775235, 51475206). Conﬂicts of Interest: The authors declare no conﬂict of interest. Sensors 2019, 19, 2560 12 of 13

References

1. Jalali, M.; Khosravani, S.; Khajepour, A.; Chen, S.K.; Litkouhi, B. Model predictive control of vehicle stability using coordinated active steering and differential brakes. Mechatronics 2017, 48, 30–41. [CrossRef] 2. Lu, Q.; Gentile, P.; Tota, A.; Sorniotti, A.; Gruber, P.; Costamagna, F. Enhancing vehicle cornering limit through sideslip and yaw rate control. Mech. Syst. Signal Proc. 2016, 75, 455–472. [CrossRef] 3. Sivaramakrishnan, S.; Siramdasu, Y.; Taheri, S. A new design tool for tire braking performance evaluations. J. Dyn. Syst. Meas. Control 2015, 137, 071013. [CrossRef] 4. Guo, H.; Yin, Z.; Cao, D.; Chen, H.; Lv, C. A review of estimation for vehicle tire-road interactions toward automated driving. IEEE Trans. Syst. Man Cybern. Syst. 2019, 49, 14–30. [CrossRef] 5. Askari, H.; Hashemi, E.; Khajepour, A.; Khamesee, M.B.; Wang, Z.L. Tire condition monitoring and intelligent tires using nanogenerators based on piezoelectric, electromagnetic, and triboelectric effects. Adv. Mater. Technol. 2019, 4, 1800105. [CrossRef] 6. Li, J.; Zhang, Y.; Yi, J. A Hybrid Physical-Dynamic Tire/Road Friction Model. J. Dyn. Syst. Meas. Control. 2013, 135, 011007. [CrossRef] 7. Rezaeian, A.; Zarringhalam, R.; Fallah, S.; Melek, W. Novel Tire Force Estimation Strategy for Real-Time Implementation on Vehicle Applications. IEEE Trans. Veh. Technol. 2015, 64, 2231–2241. [CrossRef] 8. Liu, Y.H.; Li, T.; Yang, Y.Y.; Ji, X.W.; Wu, J. Estimation of tire-road friction coefficient based on combined APF-IEKF and iteration algorithm. Mech. Syst. Signal Proc. 2017, 88, 25–35. [CrossRef] 9. Jung, H.; Choi, S.B. Real-Time Individual Tire Force Estimation for an All-Wheel Drive Vehicle. IEEE Trans. Veh. Technol. 2018, 67, 2934–2944. [CrossRef] 10. Guo, H.; Liu, H.; Yin, Z.; Wang, Y.; Chen, H.; Ma, Y. Modular scheme for four-wheel-drive electric vehicle tire-road force and velocity estimation. IET Intell. Transp. Syst. 2019, 13, 551–561. [CrossRef] 11. Matsuzaki, R.; Todoroki, A. Wireless Monitoring of Automobile Tires for Intelligent Tires. Sensors 2008, 8, 8123–8138. [CrossRef][PubMed] 12. APOLLO Consortium. Intelligent Tyre for Accident Free Traffic. Available online: http://www.vtt.fi/apollo (accessed on 10 April 2019). 13. Xiong, Y.; Yang, X. A review on in-tire sensor systems for tire-road interaction studies. Sens. Rev. 2018, 38, 231–238. [CrossRef] 14. Sergio, M.; Manaresi, N.; Tartagni, M.; Guerrieri, R. On a road tire deformation measurement system using a capacitive–resistive sensor. Smart Mater. Struct. 2006, 15, 1700–1706. [CrossRef] 15. Xiong, Y.; Tuononen, A. Rolling deformation of truck tires: Measurement and analysis using a tire sensing approach. J. Terramech. 2015, 61, 33–42. [CrossRef] 16. Matsuzaki, R.; Hiraoka, N.; Todoroki, A.; Mizutani, Y. Optical 3D Deformation Measurement Utilizing Non-planar Surface for the Development of an “Intelligent Tire”. J. Solid Mech. Mater. Eng. 2010, 4, 520–532. [CrossRef] 17. Zhang, Y.; Yi, J. Static Tire/Road Stick–Slip Interactions: Analysis and Experiments. IEEE-ASME Trans. Mechatron. 2014, 19, 1940–1950. [CrossRef] 18. Armstrong, E.G.; Sandu, C.; Taheri, S. Investigation into use of piezoelectric sensors in a wheeled robot tire for surface characterization. J. Terramech. 2015, 62, 75–90. [CrossRef] 19. Erdogan, G.; Alexander, L.; Rajamani, R. Estimation of Tire-Road Friction Coefficient Using a Novel Wireless Piezoelectric Tire Sensor. IEEE Sens. J. 2010, 11, 267–279. [CrossRef] 20. Garcia-Pozuelo, D.; Yunta, J.; Olatunbosun, O.; Yang, X.; Diaz, V. A Strain-Based Method to Estimate Slip Angle and Tire Working Conditions for Intelligent Tires Using Fuzzy Logic. Sensors 2017, 17, 874. [CrossRef] 21. Eom, J.; Lee, H.; Choi, B. A study on the tire deformation sensor for intelligent tires. Int. J. Precis. Eng. Manuf. 2014, 15, 155–160. [CrossRef] 22. Savaresi, S.M.; Tanelli, M.; Langthaler, P.; Re, L.D. New Regressors for the Direct Identification of Tire Deformation in Road Vehicles Via “In-Tire” Accelerometers. IEEE Trans. Control Syst. Technol. 2008, 16, 769–780. [CrossRef] 23. Matsuzaki, R.; Kamai, K.; Seki, R. Intelligent tires for identifying coefficient of friction of tire/road contact surfaces using three-axis accelerometer. Smart Mater. Struct. 2015, 24, 25010–25018. [CrossRef] 24. Niskanen, A.J.; Tuononen, A.J. Three 3-axis accelerometers fixed inside the tyre for studying contact patch deformations in wet conditions. Veh. Syst. Dyn. 2014, 52, 287–298. [CrossRef] Sensors 2019, 19, 2560 13 of 13

25. Khaleghian, S.; Taheri, S. Terrain classification using intelligent tire. J. Terramech. 2017, 71, 15–24. [CrossRef] 26. Solmaz, S. A novel method for indirect estimation of tire pressure. J. Dyn. Syst. Meas. Control 2016, 138, 054501. [CrossRef] 27. Liu, Z.; Gao, Q. Analytical investigation on tire dynamics by rigid–elastic coupled tire model with nonlinear sidewall stiffness. J. Braz. Soc. Mech. Sci. Eng. 2018, 40, 80. [CrossRef] 28. Carlson, C.R.; Gerdes, J.C. Consistent nonlinear estimation of longitudinal tire stiffness and effective radius. IEEE Trans. Control Syst. Technol. 2005, 13, 1010–1020. [CrossRef] 29. Janulevicius, A.; Pupinis, G.; Kurkauskas, V. How driving wheels of front-loaded tractor interact with the terrain depending on tire pressures. J. Terramech. 2014, 53, 83–92. [CrossRef] 30. Toma, M.; Andreescu, C.; Stan, C. Influence of tire inflation pressure on the results of diagnosing brakes and suspension. Proced. Manuf. 2018, 22, 121–128. [CrossRef] 31. Shyrokau, B.; Wang, DW.; Savitski, D.; Hoepping, K.; Ivanov, V. Vehicle motion control with subsystem prioritization. Mechatronics 2015, 30, 297–315. [CrossRef] 32. Flatscher, M.; Dielacher, M.; Herndl, T.; Lentsch, T.; Matischek, R.; Prainsack, J. A bulk acoustic wave (baw) based transceiver for an in-tire-pressure monitoring sensor node. IEEE J. Solid State Circuit. 2009, 45, 167–177. [CrossRef] 33. Toghi Eshghi, A.; Lee, S.; Kazem Sadoughi, M.; Hu, C.; Kim, Y.C.; Seo, J.H. Design optimization under uncertainty and speed variability for a piezoelectric energy harvester powering a tire pressure monitoring sensor. Smart Mater. Struct. 2017, 26, 1–18. [CrossRef] 34. Egaji, O.A.; Chakhar, S.; Brown, D. An innovative decision rule approach to tyre pressure monitoring. Exp. Syst. Appl. 2019, 124, 252–270. [CrossRef] 35. Chen, Z.; Xie, Z.; Zhang, J. Measurement of Vehicle-Bridge-Interaction force using dynamic tire pressure monitoring. Mech. Syst. Signal Proc. 2018, 104, 370–383. [CrossRef] 36. Braghin, F.; Brusarosco, M.; Cheli, F.; Cigada, A.; Manzoni, S.; Mancosu, F. Measurement of contact forces and patch features by means of accelerometers fixed inside the tire to improve future car active control. Veh. Syst. Dyn. 2006, 44, 3–13. [CrossRef] 37. Matsubara, M.; Tsujiuchi, N.; Koizumi, T.; Bito, K. Sensitivity Analysis for Natural Frequency of Tire Lateral Bending Mode. In Proceedings of the 12th International Conference on Motion and Vibration Control, Sapporo, Japan, 3–7 August 2014. 38. Matsubara, M.; Tajiri, D.; Horiuchi, M.; Kawamura, S. Evaluation of Spring Properties of Tire Sidewall under Changes in Inflation Pressure. SAE Int. J. Passeng. Cars Mech. Syst. 2015, 8, 825–833. [CrossRef] 39. Akasaka, T.; Yamazaki, S.; Miyamoto, J. An Approximate evaluation of Rotational Stiffness of Radial Tire. Trans. Jpn. Soc. Compos. Mater. 1984, 10, 24–31. 40. Zhao, J.; Su, J.; Zhu, B.; Shan, J. An Indirect TPMS Algorithm Based on Tire Resonance Frequency Estimated by AR Model. SAE Int. J. Passeng. Cars Mech. Syst. 2016, 9, 99–106. [CrossRef] 41. Zhang, Q.; Liu, B.; Liu, G. Design of tire pressure monitoring system based on resonance frequency method. In Proceedings of the IEEE/SAME International Conference on Advanced Intelligent Mechatronics, Singapore, 14–17 July 2009; pp. 781–785. 42. Zhu, B.; Han, J.; Zhao, J. Practical Design of an Energy Harvester Considering Wheel Rotation for Powering Intelligent Tire Systems. J. Electron. Mater. 2017, 46, 2483–2493. [CrossRef]

© 2019 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/).