IEEE SENSORS JOURNAL, VOL. 5, NO. 6, DECEMBER 2005 1281

Complex Permittivity and Moisture Measurements ofOil Palm Fruits Using an Open-Ended Coaxial SensorZulkifly Abbas, Member, IEEE, You Kok Yeow, Abdul Halim Shaari, Kaida Khalid, Jumiah Hassan, and Elias Saion

Abstract—An open-ended coaxial sensor for the determinationof complex permittivity and moisture content of oil palm fruitsis presented in this paper. The measurement system consisting ofthe sensor and a PC-controlled vector network analyzer have beentested successfully on a range of oil palm fruits of various degreesof ripeness. The initial values of the complex permittivity were es-timated using the admittance model of the sensor. The amount ofmoisture content was found by matching the values of permittivityfrom the quasistatic model with the permittivity of a dielectric mix-ture model using the moisture content values obtained from thestandard oven drying method.

Index Terms—Coaxial sensor, microwave, moisture content, net-work analyzer, permittivity.

I. INTRODUCTION

THE conventional method to determine the ripeness of theoil palm bunch is by using the number or percentage of

detached fruits per bunch [1]. However, this method has the dis-advantage that it does not relate to oil quality and quantity. Pre-vious work [2] has shown a close relationship between oil con-tent and moisture content during fruit development. Thus, theoil content and subsequently the time to harvest the fruit bunchcan be determined from moisture measurements. It was found[3] that the amount of moisture content is higher at early stageof fruit development. The need to have excess water is reducedin the ripe fruit as the oil accumulates in the mesocarp. Up to ap-proximately 14–15 weeks after anthesis, the amount of moisturecontent in fresh mesocarp is about 80% and decreases rapidly toabout 30%–40% in the ripe fruit at 20–30 weeks after anthesis.The decrease in moisture content from 15 to 20 weeks after an-thesis is almost the same rate with the accumulation of oil in themesocarp. Therefore, the close relationship between the mois-ture and oil contents in mesocarp gives a possibility of usingpercentage of moisture content or fresh mesocarp as a param-eter to gauge ripeness.

The measurement of moisture content using microwavemethod is widely known to be accurate and rapid. However,not all microwave techniques are suitable for single fruit mea-surements due to small sample size. The microstrip sensor

Manuscript received September 30, 2003; revised October 8, 2004. This workwas supported in part by the Ministry of Science, Technology, and Environmentof Malaysia under project number IRPA 09-02-04-0460-EA001. Y. K. Yeowwas supported in part by the National Science Fellowship award. The associateeditor coordinating the review of this paper and approving it for publication wasDr. Mona Zaghloul.

Z. Abbas, Y. K. Yeow, A. H. Shaari, J. Hassan, and E. Saion are with theINSPEM, Universiti Putra Malaysia, 43400 UPM Serdang, Malaysia.

K. Khalid is with the Department of Physics, Universiti Putra Malaysia,43400 UPM Serdang, Malaysia (e-mail: za@fsas.upm.edu.my).

Digital Object Identifier 10.1109/JSEN.2005.859249

based on attenuation measurement has been successfully usedto determine the amount of moisture content in the oil palmfruits [4]. Unfortunately, the sensor requires laborious samplepreparation where the fresh mesocarp of the oil palm fruit hasto be separated from the nut and cut into small pieces andcrumbled to form a semi-solid sample.

In this work, a 4.1-mm outer-diameter (OD) coaxial sensorhas been developed for the determination of moisture contentin fresh mesocarp of the oil palm fruits. The quasistatic admit-tance model of the open-ended coaxial line is used to estimatethe complex permittivity of the oil palm fruits in the frequencyrange between 2–4 GHz. Then, applying the dielectric mixturemodel, improved values of permittivity can be obtained itera-tively when the calculated values of moisture content agreeswith the experimental values of moisture content found fromstandard oven drying method.

II. BASIC PRINCIPLES

A. Admittance Model

The relationship between the normalized admittance andreflection coefficient of open-ended coaxial sensor can bewritten in the form

(1)

The normalized admittance [5] consists of the normalizedconductance and susceptance

(2)

where

(3)

(4)

1530-437X/$20.00 © 2005 IEEE

1282 IEEE SENSORS JOURNAL, VOL. 5, NO. 6, DECEMBER 2005

where is the dielectric constant of the material filling thecoaxial line, is the dielectric constant in the external medium,

and are the inner and outer radius, respectively, and is thefree space propagation constant, is the Bessel function of ze-roth order, and is the sine integral. Equations (3) and (4) canbe approximated by the first terms of the series expansion [6]

(5)

(6)

where and are the series expan-sion coefficient of normalized conductance and normalized sus-ceptance, respectively. In this study, the aspect ratio

. To take the higher order modes into the formulation,the expansion constants and [7] where

(7)

(8)

where , , and parameters were added to provide more de-grees of freedom for the optimization. The parameters , , and

are optimized until the admittances calculated using (2) givethe best match to the measured admittances of different ma-terials of known permittivities in the range between 1 and 80.In this work, the materials are air, Teflon, methanol, and watergiving , , and in the fre-quency range between 2–4 GHz.

The complex permittivity can be estimated from (2) byminimizing the difference between the measured orand the calculated value of or using error function

(9)

where and are the real and imaginary parts of themeasured reflection coefficient, respectively. In (9), the real andimaginary parts of calculated reflection coefficient are repre-sented by and , respectively.

B. Dielectric Mixture Model

The dielectric properties of the fruit mesocarp can be pre-dicted using the mixture theory [8]

(10)

where , , and are the volume fractions of water, fiber,and oil, respectively, and , , and are the correspondingcomplex permittivities. The values of are obtained from the

Fig. 1. Open-ended coaxial sensor placed against the surface of the fruitmesocarp. (a) Fruit measurement (b) The PTFE stub of the SMA panel wasmachined flat to be used as the coaxial sensor.

Cole–Cole model [9]. As the oil mesocarp consists of only threecomponents water, fiber and oil, then

(11)

which assumes a constant fiber percentage is 16% of a devel-oping fruit [10]. Since is constant, the volume fraction of oilcan be obtained if is known. The relative moisture content,

of the oil palm mixture, in percentage (wet basis), is de-fined as [11]

(12)

where , , and are the mass of water, mass of oil andmass of the fibrous material in the mixture, respectively. Equa-

ABBAS et al.: COMPLEX PERMITTIVITY AND MOISTURE MEASUREMENTS 1283

Fig. 2. Computational procedure to determine the complex permittivity of the oil palm fruit.

Fig. 3. (a) Relationship between predicted and true values of moisture content. (b) Comparison between predicted values and true m.c. for a different batch offruit samples.

tion (12) can be written in terms of density and volume asfollows:

(13a)

(13b)

The substitution of (11) in (13b) yields

(14)

The calculated values of samples moisture content were pre-dicted by least-squares method. Similarly, with previous work,

1284 IEEE SENSORS JOURNAL, VOL. 5, NO. 6, DECEMBER 2005

Fig. 4. Variation in " with moisture content of 26 C.

that minimizes the difference between the raw complex permit-tivity obtained using the quasistatic model of (2) with the (10).The criterion solution based on general absolute error function

(15)

III. MATERIALS AND METHODS

The sensor was constructed from a 12.7 12.7-mm squareflange, SMA stub contact panel. The stub was machined flatand polished. The inner and outer diameters of the PTFE filled,coaxial line sensor are 1.3 and 4.1 mm, respectively. The fourholes of the square flange was filled with silver loaded epoxy tominimize fringing fields. Sampling was done on fruits of variousdegree of ripeness from the various bunches. Twenty bunchesof tenera variety from different oil palms ten years old were se-lected from the university farm for this study. To avoid inter-fering with the normal maturation process, the proper choice ofsampling would be from the outer fruits around the equatorialof the bunch. In addition, abnormal, not fully developed, dry,and rotten fruits were not considered. The surface of the fruit

was wiped dry to free excess surface moisture. Part of the freshmesocarp of each fruit was sliced in the longitudinal directionto ensure good contact between the surface of the mesocarp andthe open-ended coaxial sensor as illustrated in Fig. 1.

Fig. 2 shows the flow diagram to calculate the complexpermittivity from the measured reflection coefficients usinga HP8720B vector network analyzer (VNA) in the frequencyrange between 2–4 GHz. The VNA was calibrated by imple-menting a standard full one-port calibration technique at theconnector end of the coaxial cable. The calibration plane wasthen extended to the sensor aperture by defining air as the openstandard, with a copper plate as the reflect standard and wateras the load standard. All microwave measurements of fruitsamples were done at 26 C. The samples were then dried ina forced-air oven for four days at 105 C for moisture contentdetermination on a wet basis.

The improved quasistatic model [7] was employed to cal-culate the values of complex permittivity of the sample fromthe measured reflection coefficient using a root finding method.The permittivity values were then fitted to a dielectric mixturemodel (10) by the least-squares method to predict the amountof moisture content in the oil palm fruit. Further, a regression

ABBAS et al.: COMPLEX PERMITTIVITY AND MOISTURE MEASUREMENTS 1285

Fig. 5. Variation in " with moisture content at 26 C.

analysis was carried out to establish the relationship betweenthe predicted and measured values of the amount of moisturecontent. Finally, the optimized values of moisture content wereused to determine the complex permittivity values using themixture model. The reliability of the permittivity results wasfurther checked with the commercial HP85070B coaxial probe[12].

IV. RESULTS AND DISCUSSION

The values of moisture content of fruits were initially es-timated from the combined mixture model and improved ad-mittance quasistatic model of an open-ended coaxial line. Thevalues were then compared with the measured values of mois-ture content obtained from oven drying method as shown inFig. 3(a). The relationship between the estimated and measuredvalues are best represented by the equation

(16)

where and are the estimated and measured percentage valuesof moisture content, respectively. Equation (16) has been usedto determine the amount of moisture content in the oil palm

fruits for 70 different fruit samples shown in Fig. 3(b) and werefound to be accurate within when compared to oven dryingmethod.

The values of moisture content found from (16) using4.1-mm OD coaxial line were inserted into the mixture modelto determine the complex permittivity values at various fre-quencies as shown in Figs. 4 and 5. Also shown in the figuresare the variations in the real and imaginary parts of permittivityfrom the mixture model and the commercial HP85070B probekit with moisture content obtained from oven drying method.As expected, the 4.1-mm OD coaxial line results are in goodagreement with the mixture model as the permittivity valueswere calculated based on the mixture model which takesinto account of the contribution of the oil, water, and fibercomponents but using (16) to determine the moisture content.In contrast, the HP85070B probe assumes that the sampleis homogeneous when calculating the values of permittivityfrom measured reflection coefficients. The assumption of alinear, isotropic, homogeneous medium of quasistatic admit-tance model when using an open-ended coaxial sensor fordetermination of permittivity of nonhomogeneous material isnot accurate. For anisotropic materials such as oil palm fruits,

1286 IEEE SENSORS JOURNAL, VOL. 5, NO. 6, DECEMBER 2005

Fig. 6. Variation in �" =" and �" =" with moisture content at 2.6 GHz.

wave polarization has to be taken into account [13]. The valuesof the magnitude and phase of the reflection coefficient forparallel and perpendicular polarization are not similar, which inturn yield different permittivity values in different direction. Inaddition, the effects of higher order modes and air gap betweenthe sample and sensor are not considered in the admittancemodel. Thus, the inaccuracy of the admittance model can bereduced by applying optimization to a known dielectric model.

The effect of the relative error in the determination of mois-ture content on the calculated permittivity can be simulatedusing (16). It was found that the relative errors in permittivity

and are almost linearly related to the relativeerror in moisture content as shown, for example at2.6 GHz, in Fig. 6.

The use of open-ended coaxial sensor for the determination ofcomplex permittivity of nonhomogeneous material can be fur-ther improved by using more accurate dielectric models. Thecomputational procedure to calculate the permittivity given inFig. 2 can be extended easily to any nonhomogeneous mate-rial by choosing suitable dielectric model. Improved accuracyof the mixture model demands calculation of the correct shapesand fields at particle levels in the fruit sample. The use of multi-scattering theory combined with correlation functions is knownto be accurate in predicting mixture properties in the more gen-eral case of the nonrandom, nondilute mixture of oil palm fruit.Unfortunately, the multiscattering theory is too computationallyextensive and demands many assumptions regarding the statis-tical nature and geometry of the inclusions.

ABBAS et al.: COMPLEX PERMITTIVITY AND MOISTURE MEASUREMENTS 1287

V. CONCLUSION

An open-ended coaxial technique suitable for measuring thedielectric properties of oil palm fruits has been developed andtested. The dielectric properties are calculated using a combinedquasistatic admittance model and dielectric mixture model. Thetechnique allows a fast method to determine the amount of mois-ture content in oil palm fruits. Moisture measurements using thesensor do not offer the accuracy of the time-consuming standardoven drying method but is suitable for a fast, first quality check offruit ripeness. A portable microwave measurement system shallbe constructed in the near future to replace the VNA for in situmeasurements.

REFERENCES

[1] I. M. Siregar, “Assessment of ripeness and crop control in oil palm,”in Proc. Malaysian Int. Agricultural Oil Palm Conf., Kuala Lumpur,Malaysia, 1976, pp. 711–723.

[2] A. Southworth, “Oil palm harvesting—A practical approach to the op-timization of oil quantity and quality,” in Proc. Malaysian Int. Agricul-tural Oil Palm Conf., Kuala Lumpur, Malaysia, 1976, pp. 721–740.

[3] A. Ariffin, S. R. Mat, M. Banjari, and O. W. E. Wan, “Morphologicalchanges of the cellular component of the developing palm fruit (Tenera:Elaeis guineensis),” in PORIM Bull., vol. 21, 1990, pp. 30–34.

[4] K. Khalid and Z. Abbas, “A microstrip sensor for determination of har-vesting time for oil palm fruits (Tenera: Elaeis Guineesis),” J. Microw.Power Electromagn. Energy, vol. 27, no. 1, pp. 3–10, 1992.

[5] N. Marcuvitz, Waveguide Handbook. Boston, MA: Boston Tech. Pub.,1964, pp. 213–216.

[6] D. Misra, M. Chabbra, B. R. Epstein, M. Mirotznik, and K. R. Foster,“Noninvasive electrical characterization of materials at microwave fre-quencies using an open-ended coaxial line: Test of an improved calibra-tion technique,” IEEE Trans. Microw. Theory Tech., vol. MTT-38, no. 1,pp. 8–14, Jan. 1990.

[7] D. V. Blackham and R. D. Pollard, “An improved technique for per-mittivity measurements using a coaxial sensor,” IEEE Trans. Instrum.Meas., vol. IM-46, no. 4, pp. 1093–1099, Aug. 1997.

[8] A. Kraszewski, S. Kulinski, and M. Matuzewski, “Dielectric propertiesand a model of biphase water suspension at 9.45 GHz,” J. Appl. Phys.,vol. 47, no. 4, pp. 1275–1277, 1976.

[9] A. Nyshadham, C. L. Sibbald, and S. S. Stuchly, “Permittivity measure-ments using open-ended sensors and reference liquid calibration—Anuncertainty analysis,” IEEE Trans. Microw. Theory Tech., vol. MMT-40,no. 3, pp. 305–314, Mar. 1992.

[10] C. W. S. Hartley, The Oil Palm. London, U.K.: Longman, 1977, pp.222–223.

[11] K. B. Khalid and Z. Abbas, “Development of a microstrip sensor foroil palm fruits,” in Microwave Aquametry: EM Wave Interaction WithWater- Containing Materials. New York: IEEE, 1996, pp. 239–248.

[12] HP85070B Dielectric Probe Kit Manual, Hewlett Packard, 1993.[13] A. Priou, “Dielectric properties of heterogeneous materials,” in Progress

in Electromagnetic Research. New York: Elsevier, 1992, vol. 6, pp.1–40.

Zulkifly Abbas (M’97) was born in Alur Setar,Malaysia, in 1962. He received the B.Sc. degreewith honors in physics from the University ofMalaysia, Kuala Lumpur, in 1986, the M.Sc. degreein microwave instrumentation from the UniversitiPutra Malaysia (UPM), Serdang, in 1994, and thePh.D. degree in electronic and electrical engineeringfrom the University of Leeds, Leeds, U.K., in 2000.

He is currently a Senior Lecturer with the Depart-ment of Physics, UPM, where he has been a facultymember since 1987. He is also the Head of the Labo-

ratory of Mathematical Sciences and Applications at the Institute for Mathemat-ical Research, UPM, where he is the Leader of the Wave Propagation ResearchGroup. His main personnel research interest is in the theory, simulation, andinstrumentation of electromagnetic wave propagation at microwave frequenciesfocusing on the development of microwave sensors for agricultural applications.

You Kok Yeow was born in Kuantan, Malaysia, in1977. He received the B.Sc. degree in physics fromthe National University of Malaysia in 2001 andthe M.Sc. degree in microwaves from the UniversitiPutra Malaysia, Serdang, in 2003, where he iscurrently pursuing the Ph.D. degree.

Abdul Halim Shaari was born in Kedah, Malaysia,in 1948. He received the B.Sc. degree in physicsfrom the National University of Malaysia in 1976,the M.Sc. degree in physics from the Universityof Southampton, Southampton, U.K., in 1978, andthe Ph.D. degree in physics for work in the areaof cryogenics and magnetic materials from theUniversity of Hull, Hull, U.K., in 1982.

Currently, he is a Professor of physics andChairman for the Centre of Advanced Materials(CAM), Faculty of Science, Universiti Putra

Malaysia, Serdang. He leads research work in various ceramics materials,such as cuprate, titanate, and the manganite systems, looking into both thefundamentals and applied properties. His research activity is focused on theapplication of magnetic materials for sensor devices. He is also a team memberof the Engineering Institute for Micro and Nanoelectronics, National Universityof Malaysia, which does research on the development of sensors and sensormaterials for MEMS applications.

Kaida Khalid was born in Terengganu, Malaysia,in 1952. He received the B.Sc. degree in physicsfrom the National University of Malaysia in 1976,the M.Sc. degree in solid-state physics from the Uni-versity of London, London, U.K., in 1978, and thePh.D. degree in electronic and electrical engineeringfrom the University of Birmingham, Birmingham,U.K., in 1986.

He is currently a Professor of microwave physicsand techniques with the Physics Department, Univer-siti Putra Malaysia, Serdang. His research interests

include microwave moisture sensors, microwave dielectric properties of mate-rials, and high-power microwave applications.

Most of his work is primarily engaged with the development of moisture sen-sors for hevea rubber latex and oil palm fruit.

Jumiah Hassan was born in Muar, Johor, Malaysia,in 1958. She received the B.S. degree in physicsfrom Northern Illinois University, DeKalb, IL, in1980, the M.S. degree in physics (research work insurface physics focusing on surface fracture) fromWashington State University, Pullman, in 1982, andthe Ph.D. degree for research work in dielectriccharacterization of materials in the microwave andlow-frequency region with dielectric modeling fromthe Universiti Putra Malaysia (UPM), Serdang, in1999.

She joined the Department of Physics, UPM, as a Lecturer in 1982. She is cur-rently involved in research work characterizing biological materials like latex,soil, fruit juices, milk, coconut milk, wood, etc., and ceramics (advanced mate-rials).

Elias Saion was born in Tangkak, Malaysia, in 1952.He received the B.Sc. degree in physics from theNational University of Malaysia in 1975, the M.Sc.degee in medical physics from the University ofSurrey, Surrey, U.K., in 1978, and the Ph.D. degreein radiation biophysics from the University of St.Andrews, St. Andrews, U.K., in 1998.

He is an Associate Profesor with the UniversitiPutra Malaysia, Serdang, and his research interestsinclude optical and electrical properties of irradiatedpolymer composites for dosimeters and electro-

chemical devices.He is a member of the Institute of Physics and the U.K. and the Malaysian

Solid State Science and Technology.