Proving a natural gamma ray soil monitor in sugar mill application : SRDC final project BS177S

Abstract

A commercially available instrument for measurement of soil in cane supplied to mills has been successfully commissioned at Tully Sugar Limited and is being used for cane payment and by mill personnel as an alarm for high soil levels in the cane supply.

A method of measurement of soil in coal based on the gamma radiation emitted by the soil has been developed by CSIRO, Minerals Division. An extension of this was to investigate whether the same technique could be used to measure soil in sugarcane consignments. As the initial experiments at the mill operated by Tully Sugar Limited indicated that this might be possible, an experimental gamma radiation detector system was installed in 1993 at Tully Mill; data collected and soil prediction equations developed through the years 1993 to 1996. Based on the prediction results obtained from the CSIRO unit, a commercial instrument, a CANESCAN 1500, was purchased and installed for the 1997 crushing season. This was supplied by Scantech Limited, formerly known as MCI Limited.

The installation arrangement of the CANESCAN differed from the CSIRO instrument with respect to both the orientation of the detector relative to the belt conveyor carrying the cane and to the shielding for background radiation attenuation. Although the size of the detector (scintillation tube) was identical to that used in the CSIRO model, these differences dictated that new calibration equation(s) be developed. Prior to 1997 the soil predictions using the CSIRO instrument had been used in the cane payment system.

Consequently, there was a need to quickly develop new equations. In 1996 it was reasoned that, because the biomass (cane) contains the gamma emitting isotope potassium 40, the relative contribution of the biomass would increase as the soil content decreased. This, therefore, should result in a change in the relationship between the measured counts per second of gamma radiation and the soil content of the cane supply. This approach led to the following ash prediction method. A preliminary ash estimation equation was applied. Depending on the result of this estimate other equations were used to obtain a final value. These equations were used throughout 1997 and 1998.

A performance objective was set based on the prediction experience prior to 1997. This was that 95 percent of predictions should have errors not greater than one percent ash, equivalent to a standard error of prediction of 0.5 percent ash. This was not met in 1997 and 1998 with only 88.1 percent of the data falling within these limits. As for the CSIRO instrument data, there was a trend to overestimate at low ash values and underestimate at high ash. Within this trend there were some extreme errors at both high and low ash levels.

The initial, incorrect assumptions made in the development of this prediction method were that the counts per second of gamma radiation would increase in a linear manner with both percent ash and the radioactivity per kg (specific activity) of the soil. Analysis of data and modelling of the effect of the gamma radiation of the biomass as a function of both percent ash and specific activity has been useful in the explanation of the errors obtained in 1997 and 1998. This analysis has shown that the major influence has been the contribution of the potassium 40 isotope in the biomass. Although no quantitative measure of the effect of extraneous matter, or leaf content has been obtained, it is apparent that it is a significant factor. Statistical, but not experimental evidence, strongly suggests that the biomass contains isotopes that contribute radiation in the thorium channel and that this is influenced by the percentage of leaf material in the cane supply. Statistical analysis also suggests that dry, dusty, harvesting conditions can result in a high specific activity clay film on the surface of the cane and leaves. This causes overestimation of the ash content.

This analysis explained why an equation using only the counts per second measured and the specific activity of the soil were inadequate in prediction for many cane samples.