a rfid sensor for corrosion

32 IEEE SENSORS JOURNAL, VOL. 16, NO. 1, JANUARY 1, 2016

A RFID Sensor for CorrosionMonitoring in Concrete

Walter D. Leon-Salas, Member, IEEE, and Ceki Halmen

Abstract— A radio frequency identification (RFID) sensorfor corrosion monitoring in concrete is presented. The sensorcan perform linear polarization, open circuit potential, andtemperature measurements. The sensor obtains its power froman external RFID reader, which also functions as a datalogger.The sensor’s electronic circuit comprises an RFID modem,a low-power microcontroller, and a three-electrode low-powerpotentionstat. The electronic circuit and the three electrodes arehoused in a 3D-printed case measuring 11.8 cm × 4 cm × 5.6 cm.An analysis of the inductive coupling between the reader’s andthe sensor’s antennas is carried out to guide the optimizationof the RFID communication link. Tests with a wet and aconcrete electrochemical corrosion cells show that the developedsensor has a performance comparable with costly and bulkierbenchtop potentiostats. An accelerated corrosion test wasconducted by embedding the electrodes in concrete for 24 days.Linear polarization resistance measurements obtained from thedeveloped sensor show the initiation and progression of corrosion.An uncertainty evaluation is carried out showing that thedeveloped RFID sensor has an accuracy compatible withprecision benchtop instruments.

Index Terms— RFID, corrosion, concrete, monitoring.

I. INTRODUCTION

CORROSION of concrete reinforcement is one of themain factors leading to premature deterioration of con-

crete structures worldwide [1]. According to the U.S. FederalHighway Administration (FHWA), the cost of corrosion inthe United States is estimated to be around $276 billion peryear [2]. The direct cost due to corrosion of highway bridgesis estimated to be $8.29 billion, while the indirect costs dueto traffic delays and productivity loss is estimated to be muchhigher [3]. Detecting corrosion in its early stages can informthe scheduling of preventive maintenance measures in order toprolong the service life of the concrete structure and reducethe likelihood of catastrophic failures.

Corrosion monitoring with non-destructive methods arepreferable. In particular, the use of wireless concrete-embedded corrosion sensors is appealing since it avoidsexposed wires that can themselves corrode or break. Given that

Manuscript received June 30, 2015; revised August 19, 2015; acceptedAugust 26, 2015. Date of publication September 4, 2015; date of currentversion December 10, 2015. The associate editor coordinating the review ofthis paper and approving it for publication was Dr. Chang-Soo Kim.

W. D. Leon-Salas is with the School of Engineering Technology, PurdueUniversity, West Lafayette, IN 47907 USA (e-mail: [email protected]).

C. Halmen is with the Department of Civil and Mechanical Engineering,University of Missouri–Kansas City, Kansas City, MO 64110 USA (e-mail:[email protected]).

Color versions of one or more of the figures in this paper are availableonline at http://ieeexplore.ieee.org.

Digital Object Identifier 10.1109/JSEN.2015.2476997

most reinforced concrete structures are designed to have aservice life of 50 to 100 years, corrosion-monitoring sensorscannot rely on batteries for their power. Instead, passivesensors that harvest their energy from their environment, orfrom an external reader through inductive coupling, are themost suitable solution.

Several efforts to develop passive embeddable corrosionsensors have been reported in the literature. In [4]–[6] abinary passive sensor based on radio frequency resonance isreported. The sensor consists of an LC tank whose resonantfrequency is monitored by an external reader. The LC tank hasa steel wire that is exposed to corrosion. As the exposed wirecorrodes, it eventually breaks, changing the capacitance of theLC tank and its resonant frequency, which is detected by theexternal reader. Once the exposed wire breaks, the sensor stopsproviding new information about the corrosion process.

In [7] a binary passive corrosion sensor dubbed SmartPebble is reported. The Smart Pebble monitors the potentialdifference between two electrodes of an electrolytic cell:an ion-selective electrode and a reference electrode. Whenthe potential difference between the electrodes exceeds athreshold, a bit of information is conveyed to an externalRFID reader via a RFID tag (MCRF202 from Microchip) byinverting the tag’s identification code.

Resonant LC tanks embedded in concrete have beenused in [8] and [9] to monitor the corrosion potential ofa steel electrode. The capacitance of the LC tank is setby varactor. The varactor’s capacitance varies in responseto changes in the corrosion potential, thus changing theresonant frequency of the LC tank. The change in resonantfrequency is detected by an external reader. A drawback ofthis technique is that the corrosion potential is affected byother factors besides corrosion, such as limited diffusion ofoxygen, concrete porosity and the presence of highly resistivelayers [1]. Hence, monitoring the steel’s corrosion potentialonly provides the likelihood or tendency of corrosion, butnot the actual corrosion rate. Another drawback of resonantLC tanks is that the transmission of corrosion information isinherently analog, hence, susceptible to noise.

The effect of eddy currents in the response of RFIDtags mounted on corroding steel plates has been investigatedin [10]. It was found that the peak amplitude of the RFID tagresponse is correlated with the atmospheric exposure time ofthe steel plates. A related work by the same research grouppresented in [11] showed that the permittivity of corrodingsteel can be detected using a vector network analyzer and awaveguide.

1530-437X © 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

LEON-SALAS AND HALMEN: RFID SENSOR FOR CORROSION MONITORING IN CONCRETE 33

A well-established technique for determining actualcorrosion rates is the linear polarization resistancemeasurement [12]. Commercially-available sensors suchas the Embedded Corrosion instrument (ECi-2) from VirginiaTechnologies can perform linear polarization resistancereadings as well as temperature, resistivity and chlorideconcentration measurements [13]. The ECi-2 is a digitalperipheral instrument on a local area network embeddedinside the concrete structure. The network communicateswith an external datalogger for corrosion data read-out. Dueto the large scale of the embedded local area network, thissolution is most suitable for new construction projects wherethe sensors can be wired up before the concrete is poured.

In this work we present a RFID-based sensor suitable forcorrosion monitoring in reinforced concrete structures. Thesensor can perform linear polarization, open circuit potentialand temperature measurements. The sensor can also measureits own power supply voltage and relay that information tothe external reader, which can then modify its power outputaccordingly. The sensor obtains its power from an externalRFID reader, which also functions as a datalogger. The sensoremploys a RFID modem that supports the ISO 15693 and theISO 18000 communication standards. Hence, commercially-available RFID readers can be employed to communicate withthe sensor.

Unlike other wireless corrosion sensors, communicationbetween the sensor and the reader is fully digital and employscyclic redundancy checks (CRC) and collision detection,guaranteeing that the data read from the sensor is free fromtransmission errors. Moreover, communication is bidirectionalwhich allows digital commands to be sent to the sensor.This bidirectionality enables the developed sensor to performseveral types of measurements, namely, linear polarization,open circuit voltage, temperature and power supply voltage.Temperature sensing comes at no additional hardware costbecause the microcontroller employed in the sensor designincorporates an on-chip temperature sensor. The developedcorrosion sensor can be installed in new structures beforeconcrete is poured, but it is small enough to be installed inexisting concrete structures via a back-filled core. This workis an extension of our work presented in [14]. Here the sensoris fully characterized and tested in both a wet and a concreteelectrochemical cell.

The rest of this paper is organized as follows: section IIprovides a brief overview of the corrosion process insteel, section III describes the sensor design and operation,section IV presents measurements collected with the developedsensor and with a benchtop potentiostat using a wet and aconcrete-based electrochemical cell. Section V concludes thepaper.

II. CORROSION OVERVIEW

A. Corrosion of Reinforcement in Concrete

Corrosion of reinforcing steel in concrete occurs via coupledanodic and cathodic electrochemical reactions at the steel-concrete interface. The anodic and cathodic reactions takeplace with the help of concrete pore water, which acts as anelectrolyte. In the anodic reaction, steel is oxidized releasing

electrons, which flow into the steel and iron ions that go intosolution. This anodic reaction is expressed as follows:

Fe → Fe+2 + 2e− (1)

The cathodic reaction can take different forms dependingon the availability of oxygen and the pH of the environment.If oxygen is available and if the concrete is highly alkaline,with a pH in the range of 12.5 to 13.5, water is reducedproducing hydroxyl ions as shown in (2):

1/2O2 + H2O + 2e− → 2O H − (2)

The product of the anodic and cathodic reactions (ironions and hydroxyl ions) combine together to produce ferroushydroxide:

Fe+2 + 2O H − → Fe(O H )2 (3)

Ferrous hydroxide in turn can be further oxidized intoproducts producing a stable film that passivates and protectsthe reinforcing steel from further corrosion [15]. This passivefilm, however, can be destroyed and localized corrosion can beinitiated. The main mechanisms that destroy the passive filmare carbonation and chloride ingression [16], [17].

When atmospheric carbon dioxide diffuses into the concrete,it neutralizes the alkalis, reducing the pH of the concretepore solution. Chloride ions, on the other hand, displace O2−from the passive layer leading to initiation of the passive filmdestruction [18]. Typical sources of chloride ions are de-icingsalts and seawater exposure. The volume of corrosion productscan be more than six times larger than the volume of steel [15].Such volume increase inside hardened concrete causes stressesthat lead to cracking, spalling, and serious deterioration ofstructural concrete. Further, corrosion of reinforcing steelresults in a reduction of the steel cross-section, loss of steelductility, and reduced bond strength between the steel andconcrete [19].

B. Corrosion Cell and Linear Polarization

A corrosion cell is a device employed to measure theanodic and cathodic electrochemical reactions of a corrodingmetal. A typical corrosion cell consists of three electrodes(working, reference and counter electrodes) submerged inan electrolyte. The working electrode (WE) is made outof the corroding metal. The reference electrode (RE) is anelectrode with a stable potential that provides a referencefor measurements in the cell. Common reference electrodechemistries are silver/silver chloride, saturated calomel andcopper/copper sulfate. The counter electrode (CE) works asa current source and is made of a material that does notcorrode.

1) Cell Potential: The cell potential, also known asOpen-Circuit Potential (OCP), is the potential differencebetween the WE and the RE. The measurement of theOCP is a common procedure for corrosion inspection ofreinforced concrete structures. This technique is describedin the ASTM C 876, Standard Test Method for Half-cellPotentials of Uncoated Reinforcing Steel in Concrete.According to this standard, if the OCP, measured using a


Fig. 1. Anodic and cathodic currents of a corroding metal as predicted bythe Butler-Volmer equation.

saturated copper/copper sulfate (CSE) reference electrode,is below −0.350 V, the probability of corrosion is more than90% [20]. The OCP measurement technique gives only anestimation of the likelihood of corrosion but does not provideinformation on the actual rate of corrosion.

Further, OCP readings can be below −0.350 V withoutsignificant presence of corrosion. If oxygen access is restricted,the steel becomes active but corrosion does not progressand the measured OCP can fall to values of −1 V [18].Other factors affecting OCP readings are the presence ofhigh resistive layers of concrete, the age of the concreteand the position of the reference electrode [21]–[23]. Hence,OCP readings alone are not sufficient for a reliable measure-ment of the corrosion rate of reinforcing steel.

2) Linear Polarization: Linear polarization is an electro-chemical technique employed to measure the instantaneouscorrosion current, Icorr , of the WE and the dynamics ofthe anodic and cathodic reactions. A popular model thatdescribes the anodic and cathodic reactions is given by theButler-Volmer equation as follows:

Icell = Icorr e2.3(Ecell−EOC )/βa

︸︷︷︸

Ianodic

− Icorr e2.3(EOC −Ecell )/βc

︸︷︷︸

Icathodic

(4)

where, Icell is cell current flowing between the CE and theWE, Ecell is the potential difference between the WE and theRE, EOC is the OCP, and βa and βc are the Tafel constantsfor the anodic and the cathodic reactions, respectively.

In a linear polarization measurement, the corrosion cellpotential Ecell is moved away from EOC disrupting thebalance between the anodic and cathodic reactions. As Ecell

changes, the cell’s current Icell is measured. Figure 1 showsthe anodic and cathodic reactions currents as well as the cell’scurrent Icell as predicted by the Butler-Volmer equation for thecase βa = 2βc. From the figure, it can be seen that the anodicreaction dominates for Ecell � EOC (anodic region) whilethe cathodic reaction dominates for Ecell � EOC (cathodicregion). At Ecell = 0 both reactions are in equilibrium andthe current through the cell is Icell = 0.

The corrosion current Icorr can be estimated from the Ecell

vs. Icell curve using different approaches. A common approachalso taken in this work, involves the calculation of an inter-mediate parameter known as the polarization resistance Rp .

Fig. 2. Conceptual diagram of the operation of the proposed RFID-basedcorrosion sensor.

From Figure 1 we see that in the vicinity of the OPC(within 10 to 30 mV) the potential-current curve is fairly linear.The slope of the curve when Ecell = EOC is defined as thepolarization resistance [24]. From equation (4) the slope ofthe potential-current curve at Ecell = EOC is given by:

Rp = �Ecell

�Icell

∣

∣

∣

∣

Ecell =EOC

= βaβc

2.3Icorr (βa + βc)(5)

From (5) the corrosion current can be solved to obtain:

Icorr = 1

2.3Rp

(

βaβc

βa + βc

)

= B

Rp(6)

where B is a proportionality factor and a function of the Tafelconstants βa and βc. The Tafel constants can be determinedas the slopes of the Ecell vs. log(Icell ) curve in the anodicand cathodic regions [25]. Once B and Rp are known, thecorrosion current is calculated using (6). Of importance incorrosion monitoring is the calculation of thickness reductionof the material per unit time (�s/�t). Thickness reductioncan be determined from Faraday’s equations [26]:

�s

�t= 3268

(

B

Rp A

)(

M

zρ

)

mm/year (7)

where, M is the mol mass of the metal, A is the area of theworking electrode, z is the number of electrons in the reactionequation for the anodic reaction (per atom of the dissolvingmetal) and ρ is the density of the metal.

III. SENSOR DESIGN AND OPERATION

Figure 2 shows a conceptual operation diagram of theproposed RFID-based corrosion sensor. For clarity, in thefigure the sensor is shown above the reinforcement bar (rebar)but in an actual deployment the sensor and the rebar willbe at the same distance from the external surface of theconcrete for sensor readings to correlate with the rebar. Theconcrete structure is naturally exposed to chlorides, oxygenand water which diffuse through concrete and reach the rebar.When chloride ions reach the rebar, the steel’s passive film is


disrupted leaving steel exposed to corrosion. The presence ofoxygen and water start and maintain the corrosion process.

The sensor incorporates a low-power three-electrodepotentiostat, which is used to perform linear polarizationmeasurements. The potentiostat measures the current thatflows between WE and the CE while the potential of the CEwith respect to the RE is varied. The WE is a piece rebarsteel while the CE is made out of a conductive material thatdoes not corrode. Typical materials for the CE are carbon,nickel and stainless steel. In this work a carbon electrodewas used. The function of the RE is to provide a stableand known potential reference. In this work a silver-silverchloride RE was employed. A linear polarization curve isobtained by plotting the CE potential against the WE current.The corrosion state of the WE can be extracted from thepolarization curve. When the corrosion sensor is placed inclose proximity to a rebar, then the corrosion state of the WEis a good estimation of the corrosion state of the rebar.

Since the corrosion sensor is expected to last as longas the concrete structure, it cannot rely on batteries as itsenergy source. Battery leakage and the limited number ofcharge-discharge cycles of batteries will limit the lifetime ofthe sensor. Instead, the proposed sensor relies on inductivecoupling to power itself and for data communications.Inductive coupling is a well established technique employedin RFID systems. The proposed corrosion sensor includesa RFID modem to communicate to commercially-availableRFID readers.

The sensor also comprises a low-power micro-controllerunit (MCU), which controls the potentiostat and readslinear polarization curves. The polarization curves arestored on an electrically erasable programmable read-onlymemory (EEPROM). The EEPROM and the RFID modemare integrated in a single chip. The RFID reader accessesthe EEPROM contents through the RFID modem and readsthe measured linear polarization curve. The data collectedby the RFID reader is sent via a serial port to a computerwhere the polarization curve is analyzed and displayed.

A. Electronic Circuit

Figure 3 shows the schematic circuit of the corrosion sensor.The sensor comprises a low-power mixed-signal MCU (TexasInstruments MSP430F2012), a dual-access EEPROM withintegrated RFID modem (ST Microelectronics M24LR64),a 16-bit digital-to-analog converter (Texas InstrumentsDAC8411) and three micro-power opamps OA1 and OA2(Texas Instruments OPA2369) and the low-offset zero-driftOA3 (Texas Instruments OPA330).

The MCU has an integrated 8-input 10-bit analog-to-digitalconverter (ADC), on-chip voltage reference generation,a temperature sensor, I2C and SPI serial protocol supportand five power-saving modes. The MCU communicates withthe external RFID reader via a RFID modem and a coilantenna. The AC voltage developed at the coil antenna isrectified by the diode bridge and regulated by the low-dropoutregulator (LDO) to generate a stable supply of 3.0 V for thesensor. The capacitor CT is a variable capacitor used to tune

Fig. 3. Schematic diagram of the corrosion sensor circuit.

the antenna to the RFID carrier frequency of 13.56 MHz.The zener diode Z1 protects the LDO from excessively largevoltages. The choke inductor L1 blocks the high carrierfrequency into the LDO to reduce power supply noise. Therole of Cstab is to smooth out the voltage rectified by thediode bridge. Resistors R1 and R2 form a voltage divider thatallows the MCU to monitor the Vdc input voltage to the LDO.The Vdc voltage can then be read by the external RFID readerwhich can adjust its output power level accordingly.

The operational amplifiers OA1 to OA3, the digital-to-analog converter (DAC) and the switch S1 constitute athree-electrode potentionstat which is used to perform linearpolarization measurements. To carry out a linear polarizationmeasurement, the Open Circuit Potential (OCP) of thecorrosion cell must be measured first. To measure the OCP,the MCU first opens switch S1 and generates voltage referenceVre f = 1.5 V (through pin A4). The virtual ground effect ofopamp OA2 sets the WE potential to Vre f . The potential of theRE is read through ADC channel A5. Opamp OA3 in unity-gain buffer configuration isolates the RE from the internalADC sample-and-hold circuit. The OC voltage is readilycomputed by subtracting Vref from the measured RE potential.

To perform a linear polarization measurement the MCUcloses switch S1, outputs voltage reference Vre f and programsthe DAC to output the following voltage:

VD AC ={

OCP − 20 mV if ramp = up

OCP + 20 mV if ramp = down(8)

The variable ramp controls the direction of the linearpolarization measurement. If ramp = up, VD AC is increasedfrom OCP − 20 mV to OCP + 20 mV. If ramp = down,VD AC is decreased from OCP + 20 mV to OCP − 20 mV.The ramp’s rate of increment and the number of points canalso be changed by the user through the RFID interface.


For each point in the ramp, the output of opamp OA2, VO A2, isread by the MCU using its internal ADC. The current throughthe WE, Icell , is measured by reading the voltage generated asIcell flows through R f b. Equation (9) shows this relationship:

Icell = VO A2 − Vref

R f b(9)

The function of capacitor C f b (in parallel with R f b) isto filter out high-frequency noise from the linear polarizationreadings. To reduce noise in the readings even further, everypoint in the polarization curve is read 64 times and the averageis reported. The same approach is followed when measuringthe OCP.

B. RFID Communications

The component that enables RFID communications is thedual-interface EEPROM. This EEPROM has an integratedRFID modem that operates at a frequency of 13.56 MHz andsupports the ISO 15693 and ISO 18000 RFID communicationstandards. Therefore, a commercially-available RFID readercan be employed to read and write the content of theEEPROM. The EEPROM can also be accessed via anI2C port. Using the I2C port the MCU can read and write theEEPROM contents. Hence, the external RFID reader and theMCU can exchange data by reading and writing to specificlocations of the dual-interface EEPROM.

To send a command to the MCU, the RFID reader writesa byte in the EEPROM location 0000h. Every second theMCU reads location 0000h. If a new command has beenwritten, the MCU proceeds to execute it. When commandexecution is finished, the MCU asserts the IDLE bit (bit 3 oflocation 0000h) to signal the reader that the command has beenexecuted. When the MCU is not executing a command, it goesinto a low-power mode (LPM3) to reduce power consumption.

The commands that the MCU can execute includeperforming a linear polarization measurement, reading theOCP, reading temperature and reading the supply voltage Vdc.The measurement results are stored in pre-defined memorylocations which can then be read by the RFID reader. TheRFID reader reads the polarization curve in blocks of 32 byteseach. Each block of data has a CRC and a collision data field.If the CRC check fails or if a collision is detected, the datablock is read again until the CRC check is correct and nocollisions are detected. This error detection mechanism ensuresthat data from the sensor is free from transmission errors.

C. Power Consumption

The current consumption of the different components ofthe corrosion sensor were measured. The sensor consumeson average 100.2 μA when it is in the idle state waitingfor a command from the external reader. The highest averagecurrent consumption is 225.2 μA and occurs when the sensoris performing a linear polarization measurement.

Table I shows the current consumption of individual com-ponents of the corrosion sensor as a percentage of the totalcurrent. The highest current consumption occurs when thereference voltage generation circuit is turned on and its output

TABLE I

MEASURED PERCENTAGE CURRENT CONSUMPTION OF THE

DIFFERENT COMPONENTS OF THE CORROSION SENSOR

Fig. 4. Measured current consumption of sensor during an OCP and linearpolarization reading.

routed outside the MCU. Using a discrete reference voltagegeneration circuit (outside the MCU) could further reduce thecurrent consumption of the sensor.

Figure 4 shows the measured instantaneous current con-sumption of the sensor while it is idle, reading the OCPand performing a linear polarization measurement. Thelarge current peaks are due to the reference circuit beingturned on and off. To reduce the average current consumption,the reference circuit is turned on just before a voltage readingtakes place.

D. Inductive Coupling Analysis

In this section an analysis of the inductive coupling linkbetween the sensor and the RFID reader is carried out. Theanalysis results in a model that will then be used to guidethe optimization of the inductive coupling link. The inductivecoupling link analysis will follow the approach presentedin [27]. Figure 5 shows the circuit modeling the inductivecoupling between the sensor and the RFID reader.

The reader’s antenna is modeled with inductor L1.Rs1 models the resistance of L1 in series with the outputresistance of the voltage source v p . Likewise, the coil antennaof the sensor is modeled with the inductor L2 with seriesresistance Rs2. The capacitor C2 models the capacitances in


Fig. 5. Equivalent circuits: (a) circuit modeling the inductive couplingbetween the corrosion sensor and the RFID reader; (b) circuit showing theequivalent AC load produced by the combination of the full-wave rectifierand the DC load Rdc ; and (c) equivalent circuit with non-coupled inductors.

parallel with the sensor’s antenna which includes the tuningcapacitor and the input capacitance of the RFID modem.

The load resistance Rload models the current consumptionof the sensor’s circuitry. The antennas of the reader and thesensor constitute two mutually coupled coils with mutualinductance M . The series resistance of the linear regula-tor (LDO) and the load resistance can be modeled with a singleD.C load Rdc.

The value of Rdc is given by the following equation [27]:

Rdc ={

Rload if Vdc < Vsupply

Rload

(

VdcVsupply

)

if Vdc ≥ Vsupply(10)

The full-wave rectifier and the DC load Rdc can in turn bemodeled by an equivalent AC load Rac. It can be shown thatthe value of the equivalent AC load Rac is given by [27]:

Rac = Rdc

2

(

1 + 2Vdiode

Vdc

)

(11)

where Vdiode is the voltage drop across the diodes in thevoltage rectifier bridge. Replacing (10) in (11) yields:

Rac = Rload

Vsupply

(

Vdc

2+ Vdiode

)

(12)

To simplify the analysis of the circuit in Figure 5(b), themutually-coupled inductors are replaced with the non-coupled

Listing 1. Iterative procedure to find the values of v1 and v2.

inductors L1(1 − k2) and L2(1 − k) and with two voltage-dependent voltage sources as shown in Figure 5(c).

From the equivalent circuit in Figure 5(c) the followingexpressions for v1 and v2 can be written:

v1 = k

nv2 + jωL1(1 − k2)

(

v p − (k/n)v2

Rs1 + jωL1(1 − k2)

)

(13)

v2 = knv1

(

Z2

Z2 + Rs2 + jωL2(1 − k)

)

(14)

where k is the coupling coefficient between the antennas andis given by:

k = M√L1 L2

(15)

and

n =√

L2

L1(16)

Given a value of v p equations (13) and (14) can be solvediteratively using the procedure outlined in Listing 1. In thelisting, i is the iteration number. The values of v1 and v2converge in less than 50 iterations.

To validate the model, Vdc was measured for different valuesof the reader’s output voltage v p . In the measurements, thereader and the sensor coil antennas were separated by a con-crete block of a thickness of 3 cm. The coupling coefficient kof the measurement setup was estimated using the proceduredescribed in section IV-A and was found to be 0.762. Figure 6shows the measured values of Vdc and the values predicted bythe model for L1 = 400 nH, L2 = 3.9 μH, C2 = 35 pF,Rload = 20 k�, Rs1 = 50.8 �, Rs2 = 0.5 �, Vload = 3.0 Vand ω = 2π ×13.56×106. Notably, the measurements and themodel agree fairly well and show that to obtain a Vdc voltageof 3.1 V or more (the LDO dropout voltage is about 0.1 V),the amplitude of v p should be at least 4.4 Vpp . For a differentconcrete thickness, the coupling coefficient can be measuredagain and the model used to calculate the minimum outputvoltage of reader.


Fig. 6. Measured and modeled values of Vdc as the reader’s output voltage v pis varied.

Fig. 7. Populated printed circuit board of the corrosion sensor.

E. Sensor Board and Sensor Assembly

A printed circuit board (PCB) hosting the electronic com-ponents of the sensor was designed and fabricated. Figure 7shows the PCB with the mounted components. The PCB size(including mounted components) is 5 cm×3 cm×1.8 cm.The PCB includes a three-pin connector to attach the threeelectrodes to the board.

The WE was manufactured from a piece of rebar. The rebarpiece was machined to create a working electrode of dimen-sions 1.3×0.5×1.0 cm3. A wire was attached to the workingelectrode using conductive epoxy. The CE was implementedusing a carbon electrode of dimensions 1.5×0.5× 1.5 cm3.The RE is a rugged silver/silver chloride electrode fabricatedby Cathodic Protection Co. and especially designed to beembedded in concrete [28]. The electrode element of the RE isa 99.9% pure silver wire and the electrode media is a speciallyformulated Ag/AgCl matrix. The electrode is encased in aPVC tube with a cementitous based end plug.

A support structure for the PCB, the electrodes and antennawas designed and fabricated using 3D printing. Figure 8 showsthe 3D printed structure with the different components ofthe corrosion sensor. The working and counter electrodes arepositioned facing each to create a direct current flow path.The surface of the working electrode not facing the counterelectrode was coated with epoxy to avoid non-direct currentcontributions. The exposed area on the working electrode is0.55 cm2. The sensor’s antenna was fabricated by winding28 AWG magnet wire around a 3D printed support structurewith a groove to guide the wire. The inductance of the sensorantenna was measured to be 3.9 μH.

Figure 8(b) shows the assembled sensor along with acover designed to protect the sensor from the environment.

Fig. 8. Assembly of corrosion sensor using 3D printed support structures:(a) sensor board with antenna and electrodes and (b) sensor assembly withcover coated with ferrite sheet.

The cover’s inside is coated with ferrite sheet to isolate thesensor electronics from the field radiated by the RFID readerminimizing induced noise. A two-part epoxy was used to fixthe electrodes to their final positions. The electrodes connectto the PCB via a three-pin connector. The dimensions of thefully assembled sensor are 11.8 cm×4 cm×5.6 cm.

IV. RESULTS AND MEASUREMENTS

The corrosion sensor system was characterized in thelaboratory under different conditions. A procedure to measurethe coupling coefficient k of the coupled reader-sensor antennasystem was performed first. In a second test, linear polarizationmeasurements were performed on a wet electrochemicalcorrosion cell using the corrosion sensor and a benchtoppotentiostat. Finally, in an accelerated test the electrodeswere embedded in concrete and corrosion was monitored forseveral days.

A. Coupling Coefficient Measurement

To measure the coupling coefficient k, the reader and thesensor antennas were placed facing each other on oppositesides of a concrete block of 3 cm thickness. The reader antennawas connected to a function generator while the sensor antennawas connected to an oscilloscope.

The function generator was set to generate a sinusoidof frequency 13.56 MHz and amplitude U1 = 1 V. Theamplitude, U2, of the voltage induced at the sensor antenna


Fig. 9. Linear polarization curve obtained using a benchtop potentiostat anda wet electrochemical corrosion cell.

was measured with an oscilloscope. The coupling coefficientwas then be calculated as follows [29]:

k = γU2

U1

√

L1

L2(17)

where γ is a correction factor that accounts for the parasiticcapacitance of the oscilloscope probe Cprobe and is givenby [29]:

γ = 2 − 1

1 − ω2Cprobe L2(18)

Using equations (17) and (18), Cprobe = 10 pF and ameasured value of U2 = 2.87 V yields a coupling coefficientvalue of k = 0.762. This coupling coefficient value indicatesa strong coupling between the antennas due in part to therelatively large size of the antennas (12 cm × 12 cm for thereader antenna and 4 cm × 10.5 cm for the sensor antenna).This procedure can be repeated for other conditions such asdifferent concrete thicknesses, concrete moisture and antennasizes.

B. Wet Electrochemical Corrosion Cell

A wet electrochemical corrosion cell was employed toverify the operation of the sensor’s potentiostat. In the wetelectrochemical corrosion cell the electrodes are submerged ina liquid electrolyte, a 3% (weight per volume) NaCl solutionin our case.

Linear polarization measurements were performed usingthe wet corrosion cell, the developed sensor and a benchtopprecision potentiostat (VersaSTAT 3 from Princeton AppliedResearch). Figure 9 shows the linear polarization curveobtained with the benchtop potentiostat. The EOC measuredby the benchtop potentiostat is −643 mV. From the linearpolarization curve in Figure 9 the polarization resistance Rp

was estimated to be 0.75 k�.A graphical user interface (GUI) that allows the user to

issue commands to the sensor and read data from the sensorwas developed. The GUI runs on a personal computer (PC)connected to the RFID reader through a serial port. If a CRCor a collision error is detected, the GUI automatically retriesthe last read/write operation assuring that the data collected

Fig. 10. Linear polarization curve obtained using developed corrosion sensorand a wet electrochemical corrosion cell.

from the sensor is free from transmission errors. Figure 10shows the linear polarization curve obtained using the sensorand the developed GUI. The EOC measured by the sensor’spotentiostat is −645 mV, which is in good agreement with theEOC read by the benchtop potentiostat.

As observed in Figure 10 the polarization curve fromthe sensor is somewhat noisy. Noise sources affecting thesensor’s potentiostat include power supply noise due to finiteisolation of the LDO regulator and induced noise from thefield generated by the RFID reader. Although relatively small,the noise in the polarization curve can affect the calculationof the polarization resistance.

To remove the effect of noise, a cubic smoothing splinecurve fitting was performed on the polarization curve [30]. Thefitted curve is shown in Figure 10. Using the fitted polarizationcurve, the polarization resistance was estimated to be 0.71 k�,which is in good agreement with the value estimated from thepolarization data collected with the benchtop potentiostat. Themeasurements performed with the wet corrosion cell validatethe operation of the sensor’s potentiostat and show that itsperformance is comparable to bulkier and more expensivesolutions.

C. Concrete Corrosion Cell

Corrosion initiation in reinforced concrete structures typi-cally occurs several years after deployment. In this work, anaccelerated test was performed to assess the performance ofthe corrosion sensor in a concrete medium in a time span ofdays instead of several years. The test was performed usinga concrete-based corrosion cell. The sensor’s electrodes wereembedded in a 5.7×5.5×2.5 cm3 concrete sample. Pre-mixedconcrete patch (Bondex from DAP) was used to prepare theconcrete sample. The sample was allowed to cure for five daysbefore collecting measurements. The sample was placed in a3% (weight per volume) NaCl solution.

The small size of the concrete sample allows rapiddiffusion of chloride ions to the working electrode initiatingcorrosion within days. Linear polarization measurementswere performed on the concrete-based corrosion cellusing the developed sensor and a benchtop potentiostatfor 24 consecutive days. To perform measurements withthe benchtop potentiosat, the sensor cover was removed,


Fig. 11. Calculated polarization resistance from linear polarization curvesacquired with a benchtop potentiostat and the developed RFID sensor.

Fig. 12. Measured OCP with a benchtop potentiostat and the developedRFID sensor.

the electrodes disconnected from the PCB and connected tothe benchtop potentiostat.

The acquired linear polarization curves were used tocalculate the polarization resistance. Figure 11 shows thepolarization resistance calculated using the polarizationcurves acquired with the benchtop potentiostat and with thedeveloped RFID sensor. Notably, the data from the sensorand the benchtop potentiostat are in good agreement andshow active corrosion by the day eight, which is marked bya large decrease in polarization resistance.

Figure 12 shows the OCP read by the sensor and thebenchtop potentiostat. The figure also shows good agreementbetween both instruments. By day seven the OCP has fallenbelow −0.234 V indicating active corrosion (−0.234 VOCP measured with a silver/silver-chloride is equivalent to−0.350 V measured with a copper/copper sulfate electrode).Hence, according to the ASTM C 876 standard after day seventhe probability of corrosion is higher than 90%. After day 15,however, the OCP starts to increase steadily. When com-plemented with polarization resistance measurements, OCPreadings give a more accurate picture of the corrosion process.Moreover, using the calculated polarization resistance values,and the accepted value of B = 26 mV for actively corrodingsteel, thickness reduction rate can be calculated using (7).

The developed sensor can also measure temperature. Theeffect of temperature on corrosion rate has been establishedin several studies [31]–[33]. The temperature information

TABLE II

AVERAGE AND STANDARD UNCERTAINTY OF OCP MEASUREMENTS

obtained from the sensor is intended to complement the OCPand linear polarization readings and provide a more completeunderstanding of the conditions leading to corrosion inside theconcrete structure. The calibration and characterization of thetemperature sensor is presented below. RFID communicationsbetween the reader and the sensor have been verified througha 6 cm concrete barrier.

D. Statistical Characterization of the Developed Sensor

The measurement uncertainty of the RFID corrosion sensorwas evaluated following the statistical method described inthe “Guide of Expression of Uncertainty in Measurement”(GUM) [34]. Repeated measurements of the OCP, temperatureand cell current were performed with the RFID sensor andthe concrete corrosion cell. From these measurements themean and the standard uncertainty were calculated, wherethe standard uncertainty equals the standard deviation of themeasurements [35].

Table II shows the mean (EOC ) and the standard uncer-tainty (uOC ) of the OCP for 100 measurements taken with boththe RFID sensor and the VersaSTAT benchtop potentiostat.A χ2 goodness-of-fit test performed on the OCP readingsindicates that the OCP readings are normally distributed ata significance level of 5%.

The table also shows the confidence interval EOC ±K ·uOC .In the table, a normal distribution for the OCP measurementsand a coverage factor of K = 2 have been assumed for a95% confidence interval. From the table it is seen that theconfidence intervals of the RFID sensor and the VersaSTATpotentiostat are overlapping. Hence, the RFID sensor andthe VersaSTAT potentionstat are said to be compatible whenmeasuring the OCP [35].

The MSP430F2012 MCU employed in the RFID sensor hasan on-chip temperature sensor connected to one of the MCU’sinternal ADC channels. A temperature read command has beenimplemented in the sensor’s firmware. When this commandis received from the reader, the MCU reads the output of itson-chip temperature sensor and stores the result in a pre-defined location in the EEPROM. The embedded temperaturesensor has a linear response from −50 ◦C to 100 ◦C [36].Hence, the relationship between the temperature sensor’soutput voltage, Vtemp , and the temperature, T , can beexpressed as:

T = G × Vtemp + Tof f (19)

where, G and Tof f are the gain and the offset of the sensor.To calibrate the temperature sensor and find the values ofG and Tof f , temperature was measured at three differenttemperature points using the RFID sensor and the HH804U


TABLE III

AVERAGE AND STANDARD UNCERTAINTY OF

TEMPERATURE MEASUREMENTS

high-accuracy thermometer from OMEGA. Measurementswere collected at room temperature, in a cold chamber witha nominal temperature of 4 (◦C) and in a hot chamber withnominal temperature of 38 (◦C). At each temperature point,100 measurements were collected at an interval of one sampleper second. A least-squares polynomial fit was employedwith the average values of Vtemp (from the sensor) andT (from the high-accuracy thermometer) to find the valuesof G = 281.796 ◦C/V and Tof f = −281.799 ◦C.

Table III shows the average temperature (T ), the standarduncertainty (uT ) and the 95% confidence interval (T ± K ·uT )for the HH804U high-accuracy thermometer and the calibratedtemperature sensor. The overlapping confidence intervals indi-cate that the RFID sensor is compatible with the HH804Uthermometer.

The uncertainty in the Icell measurements can be evaluatedby considering the current’s path through the sensor’selectronic circuitry. Icell is converted to a voltage by thetrans-impedance amplifier OA2. The output of OA2, VO A2,is then quantized and converted to digital by the MCU’sinternal 10-bit ADC. The digital representation of VO A2 isthen transmitted to the reader through the RFID inductivecoupled link. At the reader, Icell is recovered using:

Icell =˜VO A2 − Vref

R f b(20)

where, ˜VO A2 is the quantized version of VO A2. Assuminguncorrelation between ˜VO A2, Vre f and R f b and applying thelaw of propagation uncertainty [34] to (20), results in thefollowing expression for the combined standard uncertaintyof Icell :

u2Icell

= 1

R2f b

u2˜VO A2

+ 1

R2f b

u2Vre f

+(

Vre f − ˜VO A2

R2f b

)2

u2R f b

(21)

where, u˜VO A2

, uVref and u R f b are the standard uncertainties of˜VO A2, Vre f and R f b respectively. The values of the referencevoltage and the feedback resistor were measured using aFluke 179 multimeter resulting in Vre f = 1.522 V andR f b = 18.2 k�. Given that the multimeter has a resolution of0.001 V and 0.001 k�, and an accuracy of 0.09 % of reading+ 2 counts for DC voltage measurements and 0.9 % ofreading + 1 count for resistance measurements, the followingconfidence intervals are calculated: 1.522±0.0021 V for Vre f

and 18.2±0.165 k� for R f b. Assuming uniform distributions,

we obtain the standard uncertainties: uVre f = 0.0021/√

3 =0.0012 V and u R f b = 0.165/

√3 = 0.095 k�. The measured

standard uncertainty of ˜VO A2 is u˜VO A2

= 4.643 × 10−4 V.Replacing values in (21) and considering ˜VO A2 = 0,

the following Icell standard uncertainty upper bound can bedetermined: uIcell ≤ 0.443 μA. According to its manual, theVersaSTAT benchtop potentiostat has an current measurementaccuracy of 0.2% of full scale. Hence, the standard uncertaintyof the benchtop potentiostat is 0.231 μA (for a 200 μA scaleand assuming a uniform distribution). The higher accuracy ofthe benchtop potentiostat is due to its superior shielding andhigher resolution analog-to-digital conversion. However, itshigher cost, size and power requirements preclude embeddingit in concrete structures.

The standard uncertainty of the power supply measurementswas also evaluated. To this end, 100 readings of the sensor’spower supply voltage were acquired through the RFIDinterface resulting. The standard uncertainty of power supplyvoltage measurements was estimated to be 0.014 V.

The sensitivity of sensor’s accuracy with respect to thecircuit parameters R f b and Vre f was evaluated using MonteCarlo simulations. Equation (21) was employed as the sensor’saccuracy model. A variation of 10% around their nominalvalues was considered for R f b and Vre f . The sensitivity wascalculated using linear regression fitting on the results ofthe Monte Carlo simulations. The sensitivity of uIcell withrespect to R f b was found to be −2.529 × 10−11 while thesensitivity of uIcell with respect to Vre f was found to be2.866 × 10−7. Although small, the calculated sensitivitiesindicate that variations in the reference voltage Vre f have agreater impact on the standard uncertainty of the current cellmeasurements.

V. CONCLUSION

A RFID-based sensor suitable for corrosion monitoringin reinforced concrete structures has been presented. Thesensor can perform linear polarization, open circuit potentialand temperature measurements. The sensor obtains its powerfrom an external RFID reader, which also functions as adatalogger. The sensor’s electronic circuit comprises a RFIDmodem, a low-power microcontroller and a three-electrodelow-power potentiostat. The proposed sensor can also measureits own power supply voltage and relay that informationto the external reader which can then modify its poweroutput accordingly. Communication between the sensor andthe reader employ CRC and collision detection guaranteeingthat the data read from the sensor is free from transmissionerrors. The measured power consumption of the sensor whileperforming a linear polarization measurement is 675 μW. Thesensor’s electronic circuit and the three electrodes are housedin a 3D-printed case measuring 11.8 cm×4 cm×5.6 cm.An accelerated corrosion test was conducted by embeddingthe encased sensor in concrete for 24 days. Linear polarizationresistance measurements obtained from the embedded sensorshow the initiation and progression of corrosion and are ingood agreement with a much costlier and bulkier benchtoppotentiostat. The proposed sensor can be installed in newstructures before concrete is poured but it is also small enough


to be installed in existing concrete structures via a back-filledcore. The measurement accuracy of the proposed RFID-basedsensor was characterized. It was found that the RFID-basedsensor has a measurement accuracy comparable to precisionbenchtop instruments.

REFERENCES

[1] H.-W. Song and V. Saraswathy, “Corrosion monitoring of reinforcedconcrete structures—A review,” Int. J. Electrochem. Sci., vol. 2, no. 1,pp. 1–28, 2007.

[2] G. H. Koch, M. P. H. Brongers, N. G. Thompson, Y.P. Virmani, and J. H. Payer. (2002). Corrosion Cost andPreventive Strategies in the United States. [Online]. Available:https://www.nace.org/uploadedFiles/Publications/ccsupp.pdf, accessedJan. 2015.

[3] M. Yunovich, N. G. Thompson, T. Balvanyos, and L. Lave. (2001).Corrosion Costs and Preventive Strategies in the United States, Appen-dix D: Highway bridges, Federal Highway Administration. [Online].Available: http://corrosionda.com/downloads/pdf/main_appendices.pdf,accessed Jan. 2015.

[4] M. M. Andringa, D. P. Neikirk, N. P. Dickerson, and S. L. Wood,“Unpowered wireless corrosion sensor for steel reinforced concrete,”in Proc. IEEE Sensors, Oct./Nov. 2005, pp. 155–158.

[5] A. A. Yousef, “Resonant sensors for detecting corrosion in concretebridges,” Transp. Res. Rec., J. Transp. Res. Board, vol. 2201, no. 1,pp. 19–26, 2010.

[6] M. M. Andringa, J. M. Puryear, D. P. Neikirk, and S. L. Wood, “Low-cost wireless corrosion and conductivity sensors,” Proc. SPIE, vol. 6174,pp. 61740X-1–61740X-8, Apr. 2006.

[7] D. G. Watters et al., “Smart pebble: Wireless sensors for structuralhealth monitoring of bridge decks,” Proc. SPIE, vol. 5057, p. 20,Aug. 2003.

[8] K. Perveen, G. E. Bridges, S. Bhadra, and D. J. Thomson, “Corrosionpotential sensor for remote monitoring of civil structure based on printedcircuit board sensor,” IEEE Trans. Instrum. Meas., vol. 63, no. 10,pp. 2422–2431, Oct. 2014.

[9] S. Bhadra, D. J. Thomson, and G. E. Bridges, “A wireless embeddedpassive sensor for monitoring the corrosion potential of reinforcingsteel,” Smart Mater. Struct., vol. 22, no. 7, p. 075019, 2013.

[10] M. Alamin, G. Y. Tian, A. Andrews, and P. Jackson, “Corrosiondetection using low-frequency RFID technology,” Insight-Non-Destruct.Test. Condition Monitor., vol. 54, no. 2, pp. 72–75, 2012.

[11] H. Zhang, G. Y. Tian, A. Simm, and M. Alamin, “Electromagneticmethods for corrosion under paint coating measurement,” Proc. SPIE,vol. 8759, pp. 875919-1–875919-10, Jan. 2013.

[12] J. P. Broomfield, J. Rodriguez, L. M. Ortega, and A. M. Garcia,“Techniques to assess the corrosion activity of steel reinforced con-crete structures,” Amer. Soc. Test. Mater. Special Tech. Pub. 1276,pp. 91–106, 1996.

[13] Virginia Technologies, Inc. Electronic Corrosion Instruments. [Online].Available: http://corrosioninstrument.com/, accessed Jan. 2015.

[14] W. Leon-Salas, S. Kanneganti, and C. Halmen, “Development of asmart RFID-based corrosion sensor,” in Proc. IEEE Sensors, Oct. 2011,pp. 534–537.

[15] C. Andrade and C. Alonso, “On-site measurements of corrosion rateof reinforcements,” Construction Building Mater., vol. 15, nos. 2–3,pp. 141–145, Mar./Apr. 2001.

[16] C. Alonso, C. Andrade, X. R. Nóvoa, M. Izquierdo, and M. C. Pérez,“Effect of protective oxide scales in the macrogalvanic behaviour ofconcrete reinforcements,” Corrosion Sci., vol. 40, no. 8, pp. 1379–1389,1998.

[17] S. Ahmad, “Reinforcement corrosion in concrete structures, its monitor-ing and service life prediction—A review,” Cement Concrete Compos.,vol. 25, no. 4, pp. 459–471, May/Jul. 2003.

[18] M. F. Montemor, A. M. P. Simões, and M. G. S. Ferreira, “Chloride-induced corrosion on reinforcing steel: From the fundamentals to themonitoring techniques,” Cement Concrete Compos., vol. 25, nos. 4–5,pp. 491–502, May/Jul. 2003.

[19] F. Mansfeld, M. W. Kendig, and S. Tsai, “Recording and analysis ofAC impedance data for corrosion studies,” Corrosion, vol. 38, no. 11,pp. 570–580, 1982.

[20] P. Roberge, Corrosion Engineering: Principles and Practice, 1st ed.New York, NY, USA: McGraw-Hill, 2008.

[21] B. Elsener, C. Andrade, J. Gulikers, R. Polder, and M. Raupach, “Hall-cell potential measurements—Potential mapping on reinforced concretestructures,” Mater. Struct., vol. 36, no. 7, pp. 461–471, Aug. 2003.

[22] M. Stern and A. L. Geary, “Electrochemical polarization I . A theoreticalanalysis of the shape of polarization curves,” J. Electrochem. Soc.,vol. 104, no. 1, pp. 56–63, 1957.

[23] M. Stern, “A method for determining corrosion rates from linearpolarization data,” Corrosion, vol. 14, no. 9, pp. 60–64, 1958.

[24] C. Andrade and J. A. González, “Quantitative measurements of cor-rosion rate of reinforcing steels embedded in concrete using polar-ization resistance measurements,” Mater. Corrosion, vol. 29, no. 8,pp. 515–519, Aug. 1978.

[25] J. R. Scully, “Polarization resistance method for determination of instan-taneous corrosion rates,” Corrosion, vol. 56, no. 2, pp. 199–218, 2000.

[26] E. Bardal, Ed., Corrosion and Protection. New York, NY, USA:Springer, 2004.

[27] K. van Schuylenbergh and R. Puers, Inductive Powering: BasicTheory and Application to Biomedical Systems. New York, NY, USA:Springer-Verlag, 2009.

[28] Cathodic Protection Co Ltd. Permanent Ag/AgCl Reference Electrode.[Online]. Available: http://www.refcells.com/pdfs/AG4.pdf, accessedFeb. 2015.

[29] K. Finkenzeller, RFID Handbook, 3rd ed. New York, NY, USA: Wiley,2012.

[30] P. Craven and G. Wahba, “Smoothing noisy data with spline functions,”Numer. Math., vol. 31, no. 4, pp. 377–403, Dec. 1979.

[31] M. Pour-Ghaz, O. B. Isgor, and P. Ghods, “The effect of temperatureon the corrosion of steel in concrete. Part 1: Simulated polarizationresistance tests and model development,” Corrosion Sci., vol. 51, no. 2,pp. 415–425, Feb. 2009.

[32] J. M. Deus, B. Díaz, L. Freire, and X. R. Nóvoa, “The electrochemicalbehaviour of steel rebars in concrete: An electrochemical impedancespectroscopy study of the effect of temperature,” Electrochim. Acta,vol. 131, no. 10, pp. 106–115, Jun. 2014.

[33] J. Hu, X. Cheng, X. Li, P. Deng, and G. Wang, “The coupled effect oftemperature and carbonation on the corrosion of rebars in the simulatedconcrete pore solutions,” J. Chem., vol. 2015, Mar. 2015, Art. ID 462605.

[34] Evaluation of Measurement Data—Guide for the Expression of Uncer-tainty in Measurement, JCGM 100:2008, 1993.

[35] A. Ferrero and S. Salicone, “Measurement uncertainty,” IEEE Instrum.Meas. Mag., vol. 9, no. 3, pp. 44–51, Jun. 2006.

[36] Texas Instruments. MSP430x2xx Family User’s Guide. [Online].Available: http://www.ti.com/lit/ug/slau144j/slau144j.pdf, accessedJun. 2015.

Walter D. Leon-Salas received the B.S. degree in electronics engineeringfrom the Universidad Nacional de San Agustin, Arequipa, Peru, in 1996, andthe M.S. and Ph.D. degrees in electrical engineering from the University ofNebraska–Lincoln, in 2001 and 2006, respectively. From 1996 to 1998, he waswith Telefonica del Peru. From 2007 to 2012, he was with the University ofMissouri–Kansas City, where he was an Assistant Professor with the ComputerScience Electrical Engineering Department. He is currently with the Schoolof Engineering Technology, Purdue University. His research interests includecorrosion sensing CMOS imaging, data compression, and energy harvesting.He received the CAREER Award from the National Science Foundation andthe Leadership Excellence Achievement Program Award from the MissouriSociety of Professional Engineers in 2011.

Ceki Halmen received the B.S. degree in civil engineering from BogaziciUniversity, Istanbul, Turkey, and the M.S. and Ph.D. degrees from Texas A&MUniversity, College Station, TX. He is currently an Associate Professor withthe University of Missouri–Kansas City. He is a member of the AmericanConcrete Institute. He is also a member of the ACI 222 Corrosion and201 Durability Committees.

a rfid sensor for corrosion

Documents