Abstract

Calibration of a capacitance-type moisture meter for oil palm kernels. A calibration chart for a capacitance-type moisture meter (Motomco model 919) was prepared to be used with oil palm kernels. It was determined that operating the meter with a dial reading of 53, a sample of 250 g of kernels without endocarp sized between a #30 sieve (to exclude those that were too big to fit in the test cell), and a #24 sieve (to remove small pieces), allowed the proper functioning of the meter in a moisture range from 4 to 15% w.b. Though kernels from three sites were used, only one calibration chart was needed, since precision was satisfactory, ranging from 0.20% for kernels mostly of the Tenera-type palms to 0.28% for those of the Dura-type. A correction of 0.1 in the moisture reading is required for each Celsius degree the sample temperature deviates from 25°C.

Introduction

In order to achieve an adequate control during drying and a proper monitoring of stocks during the storage of cereals, pulses and other agricultural commodities, the industry relies today on the availability of devices capable of giving quick estimations of their moisture content. The accuracy of the available meters is such that they are accepted in most countries for trading purpose.

For many years, the decision-making process in the grain industry has relied on the information provided by these moisture meters. No other principle in which these devices are based upon combines accuracy, precision, and low cost in the way capacitance does. The Motomco Moisture Meter is just one example of an apparatus that uses the dielectric property of materials. It can, in a matter of minutes, offer an estimate of the moisture content of a sample easing the monitoring of drying operations and others alike.

To use the Motomco or any other apparatus of this kind, a conversion chart is required to correlate meter readings with moisture content values. However, a table for each product is needed, since different products have different dielectric properties. These differences are found between cultivars of a given vegetal species, as indicated by Alizaga (1981) and Soto (1990).

manufacturer of this equipment, a chart for oil palm kernels is not yet available. The Oil Palm Industry in Costa Rica, and possibly in other countries as well, could definitely use a device with such characteristics calibrated for the oil palm kernels.

The objective of this research was to prepare a conversion chart for the Motomco meter to be used with oil palm kernels.

Materials and Methods

The study was conducted in the Grain and Seed Research Center of the University of Costa Rica (CIGRAS). The experimental material was collected from three oil palm mills owned by the Palma Tica company, located in Coto, Naranjo and Palo Seco. The first plant is close to the south end of the west coast of Costa Rica, and the other two are in the central pacific region, also close to the coastal border.

Preliminary Tests

Since this was the first attempt to use the Motomco 919 with oil palm kernels, the operating conditions (sample weight, composition, device internal calibration) had to be found. To achieve this, batches of wet and clean kernels (without fragments of endocarp "shell" and other impurities) were prepared. Portions of 5 kg were dried slowly in an air-conditioned room at approximately 25°C and 50% relative humidity. Samples were drawn every 12 hours until constant readings were obtained with the Motomco for 200 g and 250 g samples.

Once all the adjustments that allowed the meter to function normally were found, an evaluation of the effect of the physical composition of the samples upon meter performance was made. It was determined that kernels that passed through a sieve of 1.2 cm in diameter (#30) but were retained by another sieve of the same type with perforations of 0.95 cm in diameter (#24), fit properly into the test cell. Kernels retained by the #30 sieve tended to hinder the even filling of the cell.

Meter readings were obtained for samples with the following composition:

• the original condition with impurities (fragments of shells) and with small fragments of kernels
• with impurities but without small fragments (sieved)
• without impurities (cleaned) and without fragments (sieved)
• only whole kernels (no fragments at all)
• only fragments (no impurities or whole kernels)

The moisture content of 500 g samples of wet, clean and sieved kernels were adjusted by drying, in such a manner that the moisture of a sample obtained varied approximately by 1% from the previous one. The objective was to use the whole range of readings the Motomco could offer. A total of 233 samples from the three mills were used.

Samples were then stored in plastic bags at 25°C during 10 days to allow a uniform distribution of moisture within the kernels. After obtaining three readings per sample in the Motomco, the moisture content of a 60 g sample was then determined using an oven (120 ±1 °C, during 12 hours). Three replicates were conducted for each sample. The moisture content was expressed in a wet weight basis (w.b.). This procedure is normally used at the laboratories of Palma Tica's mills.

With the Motomco average reading (capacitance value) and the oven moisture content for each sample, linear and quadratic regression equations were estimated for each of the three lots of kernels. The criterion for selection of the best equation included the determination coefficient (R²), the F (Fisher test) value, and the distribution of errors. Regression equations were then calculated for the 233 experimental data points and the same criteria was used to select one which best fitted all data. A single conversion chart for oil palm kernels was prepared with this equation.

Correction for temperature

To determine the effect of sample temperature on moisture content estimated with the conversion chart, seven groups of 10 samples with different moisture contents each, were placed at one of seven temperatures (7, 15, 20, 25, 30, 35, and 40°C). After thermal equilibrium was reached, individual moisture contents were estimated with the conversion chart prepared. In order to avoid moisture changes in the samples, they were kept within airtight plastic bags except while taking readings with the Motomco meter.

A coefficient that indicated the average change in moisture for each degree Celsius that the sample temperature deviated from the 25°C (reference temperature at which all previous tests were performed) was computed for each sample at each temperature. Then, an average coefficient was calculated from absolute values of 60 coefficients obtained.

Results and Discussions

Sample size and moisture range

A sample of 250 g, and an internal calibration of the Motomco meter with the dial set at 53, permitted accurate readings for samples whose moisture ranged from 4 to 15%.

This moisture range was considered appropriate for practical purposes, since kernels without endocarp enter the dryer with a moisture content near 20% and normally leave them with 4% or more, unless overdrying occurs. The estimation of kernel moisture values above 15% have little practical importance.

Under normal conditions, a conversion chart covering such moisture range (4 to 15%) would allow an adequate control of the drying process. Furthermore, the kernel moisture considered appropriate for safe storage (6.0 to 7.5% according to Jiménez et al. 1994), is within the same range and thus the same conversion chart could be used in monitoring kernel condition throughout the storage period.

When moisture tests were done on sample kernels in their original condition, that is, without cleaning or sieving, lower Motomco readings were obtained (Table 1). This was partially attributed to the presence of fragments of kernels and shells interfering with the settling of the material in the meter test cell.

Kernel fragments gave lower readings than the original sample where they came from. Small fragments may dry faster than whole kernels or large fragments, due to their large surface area relative to their weight.

The preliminary tests led to the conclusion that samples of 250 g of cleaned and sieved kernels could be used with the Motomco 919, without problems in the cell of the meter. In addition, since the meter proved to be sensitive to small changes in the physical composition and moisture of samples, the cleaning and sieving of these seemed the proper steps toward minimi- zing such sources of variation. Hurburgh (1987), in studies with soybeans and three capacitance-type moisture meters, including the Motomco 919, found that 89% of the total variance in the data could be attributed to variations in the dielectric properties of the samples, while only 3.3% of this variation was due to the precision of the meters used.

Regression model

A quadratic equation fitted the data (Fig 1. Motomco readings vrs oven moisture). For all three mills, both linear and quadratic equations had an R² of 0.99; however, the linear equations showed larger coefficients of variation. These high R² values indicated that the Motomco meter responded positively to small changes in moisture and that its readings were highly correlated with oven data.

From a practical view point, these results show that the Motomco 919 meter can detect small changes in moisture content of samples of oil palm kernels. Another important characteristic of the apparatus is its accuracy as shown by its small standard deviation of the differences between data obtained by means of the quadratic equation and oven data. The standard deviation of these differences for the three mills were 0.23%, 0.28%, and 0.20% (w.b.) for Coto, Naranjo, and Palo Seco respectively. These values are even smaller than those obtained by Hurburgh et al. (1987) for corn (0.50% w.b.) and by Zeledón and Alizaga (1992) for rice (0.42% w.b.).

General Equation

Due to the great similarity between the equations for the three mills ( Fig. 1a , Fig. 1b and Fig.1c ), a general equation was evaluated. Dispersion of data was as small as in the previous cases and also had a defined tendency ( Fig. 2 ). Both the linear and the quadratic equations had again values of R² equal to 0.99; thus, either could estimate the moisture content of kernels with acceptable accuracy. Nevertheless, the quadratic equation showed a smaller coefficient of variation.

The linear regression equation showed a higher F value than the second degree equation, meaning that the quadratic component did not contribute significantly to the model. Despite this, the conversion chart was prepared with the quadratic equation because the standard deviation was smaller for the quadratic equation than for the linear equation (0.24% and 0.30% respectively). Of the estimates obtained with the quadratic equation 60% were within 0.20% (w.b.) of oven values, 30% within 0.21% and 0.40%, and only 10% were greater than 0.40%. Furthermore, there was a systematic pattern of error distribution for the first order equation, while those for the second order equation showed a random distribution ( Fig.3a and  Fig. 3b ).

Conversion chart

With the general quadratic equation, a single conversion chart was prepared ( Table 2 ). There were differences due to the physical and the genetic composition of the materials evaluated. For example, kernels from the Naranjo mill normally contained a high proportion of small fragments (20% or more) which caused the Motomco 919 to underestimate moisture values. As for the genetic composition, the samples from Naranjo were mostly of the "Dura" type while those from Coto and Palo Seco were of the "Tenera" type.

Hurburgh et al. (1987) found that the main sources of variation in the estimation of mois- ture in corn samples with three capacitance-type meters were the year of harvest and the variabi- lity in dielectric properties of the samples. That was not the case in this experiment, since the conversion chart performed well with all materials tested regardless of their origin or genetic composition. It can be concluded that the dielectric properties of the materials tested in this experiment were similar, despite the forementioned different origins.

Correction for sample temperature

The correction for temperature obtained in 60 trials averaged one tenth of a moisture percentage point per degree Celsius (0.1% (w.b.)/°C). Therefore, the adjustment to a particular moisture value obtained with the Motomco should be done by adding or subtracting 0.1 for each degree Celsius that the sample temperature deviates from 25°C.

The correction factor obtained in this work coincides with the corrections recommended in the Manual of Operation of the Motomco 919 meter, for many other agricultural commodities (Motomco 1977).

References

ALIZAGA, R. 1981. Medición del contenido de humedad en granos básicos con el determinador Motomco 919. Tesis Ing. Agr., San José, Universidad de Costa Rica, 42p.

HURBURGH, C.R. 1987. Moisture Meter Performance II. Soybeans. Transactions of the ASAE 30(2):582-584.

HURBURGH, C.R.; PYNTER, L.N.; SCHMITT, S.G. 1987. Moisture Meter Performance I. Corn Over Five Crop Years. Transactions of the ASAE 30(2):579-581.

JIMENEZ, R.; ZELEDON, M.E.; ALIZAGA, R.E. 1994. Relación de equilibrio entre el contenido de humedad de almendra de palma aceitera ( Elaeis guineensis Jacq.) producida en Costa Rica y la humedad relativa del aire. ASD Oil Palm Papers. [en preparación].

MOTOMCO. 1977. Motomco Model 919 Moisture Meter. Operating Instructions. 14p.

SOTO, H. 1990. Prueba adicional sobre la medición del contenido de humedad en granos básicos con el determinador Motomco 919. Tesis Ing. Agr., San José, Universidad de Costa Rica, 61p.

SPRINGER, E. 1991. Available conversion charts for the Motomco 919. New Jersey, Motomco, Inc.

ZELEDON, M.E.; ALIZAGA, R.E. 1992. Procedimiento alternativo para la estimación de la humedad del arroz con el Motomco 919. Agronomía Costarricense 16(2):249-255.