ASABE Publications - For Review Only SUGAR CANE YIELD MONITORING SYSTEM. Journal: American Society of Agricultural and Biological Engineers

Transcription

1 ASABE Publications - SUGAR CANE YIELD MONITORING SYSTEM Journal: American Society of Agricultural and Biological Engineers Manuscript ID: PM Journal Name: Applied Engineering in Agriculture Manuscript Type: Full-length article Date Submitted by the Author: 22-Jan-2009 Complete List of Authors: Mailander, Michael; Louisiana State University AgCenter, Biological & Agricultural Engineering Benjamin, Caryn; Louisiana Dept of Health & Hospitals, Engineering Price, Randy; Kansas State University, Biological and Agricultural Engineering Hall, Steven; Louisiana State University AgCenter, Biological and Agricultural Engineering Keywords: Abstract: yield monitor, sugarcane, harvester, precision agriculture, weighing, sensors, GPS A yield monitoring system was developed for a billet-type harvester consisting of a weigh plate, a data acquisition system, and a differential global positioning system. The weigh plate was mounted on a 1997 CAMECO CH2500 combine in the upper portion of the elevator and supported by load cells mounted in an adjustable protective box. The data acquisition system was set up to record consecutive measurements based on a slat conveyer speed of 183 cm (72 inches) per second. Experiments were run with different levels of maturity, variety, row/section length, and flow rate, and the measured weight of material was compared to a weigh wagon. Results indicate that the system was able to predict sugar cane yield with a slope of and an R-squared of The average error for all readings was 11 percent. This work uniquely showed statistically that the scale readings varied by yield and variety, but that maturity, section length, and flow rate did not have a significant effect on the readings. abstract.doc

2 Page 1 of 15 ASABE Publications -

3 ASABE Publications - Page 2 of Title: SUGAR CANE YIELD MONITORING SYSTEM Authors: Mike Mailander, Caryn Benjamin, Randy Price, and Steven Hall Keywords: yield monitor, sugar cane, harvester, precision agriculture, weighing, sensor, GPS ABSTRACT A yield monitoring system was developed for a billet-type harvester consisting of a weigh plate, a data acquisition system, and a differential global positioning system. The weigh plate 7 was mounted on a 1997 CAMECO CH2500 combine in the upper portion of the elevator and supported by load cells mounted in an adjustable protective box. The data acquisition system was set up to record consecutive measurements based on a slat conveyer speed of 183 cm (72 inches) per second. Experiments were run with different levels of maturity, variety, row/section length, and flow rate, and the measured weight of material was compared to a weigh wagon. Results indicate that the system was able to predict sugar cane yield with a slope of and an R- squared of The average error for all readings was 11 percent. This work uniquely showed statistically that the scale readings varied by yield and variety, but that maturity, section length, and flow rate did not have a significant effect on the readings INTRODUCTION Sugar cane is an economically important crop in southern Louisiana, which produces 37 percent of the United States sugar (Haley et al, 2008). In sugar cane farming practices, the soil in the field/block is assumed homogenous. This is generally not true within a field and large variations can occur (Karlen et al., 1990). For this reason, precision farming has become an important tool to allow farmers to effectively manage fields, optimize farm profits, and minimize

4 Page 3 of 15 ASABE Publications effects on the environment (Colvin et al., 1991). For this reason the adoption of precision farming practices is desirable in sugar cane production. One important application of precision farming is yield mapping. Yield maps provide sitespecific yields that can aid in the management of fertilization and pesticide rates. Yield maps consist of two variables, the site-specific crop yield and the position in the field. There are currently no commercially available yield monitors for measuring sugar cane harvested by a 29 harvester. The proposed system measures and stores data required to make yield maps (Benjamin, 2002). Another important advantage of a yield monitor is that the farmer can reduce the problem of overloading the tractor-trailers with cane. In Louisiana there is a weight limit (45,360 kg [100,000 lbs ] GVW) on tractor-trailers used to haul sugar cane over the road. OBJECTIVES The objectives of this study were to: 1. Design a scale that measures the yield of sugar cane as it is harvested by a billet sugar cane harvester. 2. Install the scale on a harvester and test the accuracy by comparing its measured yield to a weigh wagon. 3. Test the effects of sugar cane variety, maturity, flow-rate, and section/row length on the readings. LITERATURE REVIEW A comprehensive review of the application of precision agriculture techniques in sugar cane concluded that a high priority be given to the development of yield monitoring (Davis and Schmidt, 2007). Even though there are no commercially available sugar cane yield monitors, there have been studies and even some patents on these devices. Australia s National Centre for

5 ASABE Publications - Page 4 of Engineering in Agriculture has been working on a prototype sugar cane yield monitor for a combine harvester (Cox et al., 1998). This group has used direct and indirect techniques for measuring the cane yield on the harvester. The direct techniques involve mass and volume measurements. Their indirect techniques involved measurements of power consumption. After field tests, the final technique selected was the direct mass measurement. The resulting patent was titled, Mass Flow Rate Sensor for Sugar Cane Harvesters, Australian Patent No (Cox et al., 1999) Specific details of the sensor design were published and indicated that the mass flow sensor consisted of a weigh platform supported by load cells mounted in the upper section of the elevator of the harvester. Numerous variations of the platform and load cell configurations were tested. Additionally, they used an accelerometer and an inclinometer for error correction. Supporting test results were not given in the patent, but an interesting component of the system was an apparatus to deter and lessen the air suction effect caused by nearness to the secondary extractor fan. A similar yield sensor was developed by Pagnano et al. (2001) and consisted of a weighing frame, load cells, conveyor speed sensor, and a data acquisition apparatus. The weight sensor was mounted in the upper section of the harvester s elevator. Accelerometers were used to determine the frequencies produced by the harvester and a Butterworth low-pass filter was used to filter out certain frequencies. An instrumented trailer was used to check the accuracy of the system and showed percent errors ranging from 8.74 to percent. Wendte et al., 2001, described a sugar cane yield monitor in a U.S. Patent that uses an elevator pressure sensor and a deflection plate to measure the quantity of the harvested sugar cane. They also used a low pass filter to smooth the peaks in the elevator pressure signal. To

6 Page 5 of 15 ASABE Publications account for the effects of dynamic forces, the deflection plate was pre-loaded to always read positive even in the worst field conditions. Results were not given for this unit and no formal research paper seems to exist. Another group in Brazil (Magalhães and Cerri, 2007, Cerri et al., 2008) tested a load cell based weight plate system and reported a correlation of 0.66 and a mean error of 8.7%, as compared to wagons weighed with a scale. Their system also included a speed sensor and 75 accelerometer to calculate incline Non-contacting monitors have been reported (Juniper Systems, Inc. 2008, Price, et.al 2007). These systems use either ultra-sonic or optical methods of estimating the volume of material via profile. The HarvestMaster system from Juniper is no longer commercially available, but showed 0.6% error until the calibration failed (Price, et. al, 2007). The optical sensor is still under development. The accuracies reported were varied, and may be accounted for by changes in the sensor or materials being sensed DESIGN AND DEVELOPMENT To develop a yield monitor for a sugar cane harvester, the harvesting process needed to be examined (fig. 1). The sugar cane is cut at the ground by the harvester s base blades and topped by the topper as it enters the harvester. Once inside, the cane is cleaned and chopped into 8-12 inch (20-30 cm) pieces called billets. Seventy five percent of the trash is blown out through the primary extractor fan as the billets are dumped onto the elevator (Sciortino, 1999). The billets are then conveyed up the elevator by a chain driven slat system. Slats are approximately 61 cm (24 inches) apart and move at a maximum speed of cm/s (88 in/s ) (which can vary at the operator s control and the hydraulic capacity of the machine).

7 ASABE Publications - Page 6 of The elevators floor is 80 cm (31.5 inches) wide and the elevator is divided into two lifting sections. The first section operates at a 55 degree angle (with the horizontal) when the elevator is at its highest position during harvesting and the second section makes a 25 degree angle (fig. 1). The first section is usually constructed of serrated metal to allow dirt to fall through it and the second section is solid. Before the billets are dumped into a wagon the secondary extractor fan removes approximately percent of leftover trash (Sciortino, 1999). 98 Fundamental techniques considered for measuring sugar cane yield are direct mass and volume measurements, and indirect methods. Direct mass measurement, or weighing the cane as it travels along the elevator, were determined by examination of the machine, researching other crop yield monitors, and speaking with manufacturers to be the best way to measure cane yield. Since the floor of the elevator is stationary, it seemed plausible to place a weighing device there. A scale was then developed and placed co-planar with the floor in the upper section of the harvester s elevator (fig.1). The upper section was chosen since it had the least inclination and would create less offset in weighing Scale and Weigh Plate Design The design of the weight scale consisted of a plate supported by a single load cell mounted in the middle of the conveyer which was mounted to a protective box. The protective box allowed the system to be mounting to the frame and provided fine adjustments of the scale to match the floor. After the 1999 field tests, the single load cell approach proved unstable without extra supports because of vibration and noise interference (Benjamin and Mailander, 2000). The 2000 scale design used two load cells and the plate width was cut in half to reduce noise and vibration effects. Figure 2 shows the 2000 scale after several weeks running and illustrates

8 Page 7 of 15 ASABE Publications possible problems in Louisiana that occur from silting-in and caking of mud and dirt in the gap between the floor and the plate. The weigh plate s dimensions played a critical role in the system s ability to measure the sugar cane billets accurately. The billets moved up the elevator via slats which were spaced 24 inches (61 cm) apart. To ensure that all cane was measured, the platform was given a length of the distance between two slats, and the width of the elevator floor (61 cm x 76 cm [24 in x in]). Also, the weigh plate s thickness (0.5 cm [ in]) was designed to match the floor and not have large deflections (maximum deflection was cm [0.125 inches]). The weigh plate dimensions were changed after the 1999 field test to reduce the amount of error created by the torsion and vibration effects of the plate. This length was changed from 61 cm (24 inches) to 30.5 cm (12 inches) in length, and the sampling frequency was doubled Load Cell Design and Selection Since the slats on the elevator traveled back along the bottom of the elevator, the amount of clearance was limited, with little room to place a scale in the floor. For this reason, a low profile load cell was chosen. The load cell s maximum capacity was based on the maximum cane weight 131 per slat of the conveyer. Using the maximum density of sugar cane as 353 kg/m 3 (22 lb/ft 3 ) and the maximum volume of cane as m 3 (3.33 ft 3 ), the maximum weight of the per slat was 33 kg (73 pounds). Since the load cell location was at an angle of 25 degrees, the scale had to be re-calibrated for this angle. The final capacity chosen for the load cell was 50 kg (110 pounds), which also accounted for the weight of the plate. A single-point low-profile load cell (Model 60060, Sensortronics) was selected, which was typically used for a 91 cm x 91 cm (36 in x 36 in) plate and was internally compensated for offset loading.

9 ASABE Publications - Page 8 of Protective Box Design A protective box was designed to protect the load cell from being damaged by the weather, billets, trash (such as leaves), and dirt from the returning slats. The box was also designed to rigidly support the load cell to the elevator and allow adjustments in the plate height and orientation. The tolerance between the elevator floor and the weigh plate had to be as small 144 as possible to prevent trash and dirt from becoming wedged in-between the floor and the plate and reduce billet strikes on the plate when they move from the elevator to the weight plate. A small tolerance of cm [ in]) was chosen for this clearance MATERIALS AND METHODS Laboratory Testing A scale was set up in the research laboratory to run various tests. The laboratory scale was an exact replication of the scale that was mounted on the harvester and used to test the load cells linearity by applying weights at increments of 4.5 kg (10 pounds). The load cells linearity was found to have a slope of 4.5 and an R-squared of (fig. 3). A laboratory test was also done to determine the appropriate size and material of the weigh plate. Since the 1999 field tests showed that vibration problems existed with using a large steel plate and only one support load cell (Benjamin and Mailander, 2000); four different plates were tested to determine which one would have the smallest vibration amplitude. Vibrations were induced by setting an air pump (model # 2545B-01, Welch Vacuum Pumps) on top of each weigh plate when mounted on a load cell and sampling at 10 hertz. The small steel plate with

10 Page 9 of 15 ASABE Publications dimensions of 30.5 x 76 x cm (12 x 30 x inches ) made of Exten 50 exhibited the best dampening qualities and was selected for the 2000 field tests. A power spectral density test was done to determine the dominate frequencies and consisted of taking data 10 hertz data while the harvester went down a previously harvested row. These frequencies included 20, 75, 100, 130, 350, 370, 400, 440, and 480 hertz. Some damping was expected with crop present Field Testing Before each field test, the elevator was raised to the highest position (harvesting angle) and weights were placed on the scale to establish calibration. The harvester was then run through the crop, depositing the billets into a small capacity weigh wagon (3-Ton Weigh Wagon, CAMECO). The weigh wagon was instrumented with three load cells, two on each axle and one on the tongue (Bischoff et al., 2001). Field tests were conducted in the fall of 1999 and the fall of The 1999 field tests were considered preliminary because of considerable loss of data due to the vibration of the plate and harvester (Benjamin and Mailander, 2000). The 1999 field tests were performed at the Sugar Research Station in St. Gabriel, LA on one variety of sugar cane (LCP ). Twenty-four 50- foot (16 m) sections were marked off. Three harvester speed levels were tested to produce different flow rates. The levels were approximately 2.9 km/h (1.8 mph), 4.5 km/h (2.8 mph), and greater than 5.1 km/h (3.2 mph). Each level was tested on eight sections. The harvester was at the desired velocity before it entered the uncut cane and then stopped at the end of the 16 m (50- ft). At the end of each section, the weigh wagon weight was recorded. The GPS and data-logger ran continuously during the tests. A slat of cane passed over the scale every 0.27 seconds, which

11 ASABE Publications - Page 10 of was based on the elevator speed of cm/s (88 in/s) and the weigh plate s length of 61cm (24 in). To ensure that every load was measured only once, the sample rate was set to 3.7 hertz. The 2000 field tests were performed at the St. Gabriel Research Station on two varieties - LCP and CP (Benjamin et. al., 2001). Four tests were performed, numbered one through four (table 1). Tests 1, 2, and 3 consisted of rows that were divided into four different lengths: 15.2, 22.8, 30.5, and 46.7 m (50, 75, 100, and 150 ft). Test 4 consisted of whole rows 189 that were 114 m (375 ft) long. 190 Table 1. Field Test Design Test Rows Variety Maturity 1 10 LCP months 2 12 LCP months 3 6 CP months 4 6 CP months Two harvester speed levels - slow (0-4.5 km/h [0 2.8 mph]) and fast (4.7 km/h - 7.2km/h [2.9 mph mph]) - were tested to produce different flow rates. Test 1 was done on October 27, 2000, and the other three tests on December 5, The process of each sub-test (section) started with the harvester entering the cane to achieve the appropriate speed and stopped at the end of the section. At the end of each section the weight of the weigh wagon was recorded manually. The GPS logged readings were taken continuously during the tests. A slat of cane passed over the scale every 0.14 seconds (7.14 hertz) and was based on the elevator speed of cm/s (88 in/s) and the weigh plate s length of 30.5 cm (12 in ). Tests 1 and 2 were designed to test for effects of maturity, 2 and 3 for varietal effects, 1 through 3 for resolution and spatial variability, and 4 for overall accuracy.

12 Page 11 of 15 ASABE Publications Analysis Phase The following analysis was performed to determine the scale s accuracy and establish what factors affected the scale s ability to measure the yield. Definitions: Spot yield - the instantaneous weight measurement of cane taken by the scale. 208 Scale yield - the summation of spot yields for a given section Wagon yield- the weight of the cane measured by the weigh wagon for a given section. The spot yields were filtered with a twenty point moving average and were summed for each section to get a total yield. The weigh wagon yields were considered the actual yield for that section of harvested sugar cane. To establish how well the scale predicted the weigh wagon yield, a regression was done along with calculation of residuals and 95 percentile prediction intervals. Statistical tests were conducted comparing the scale yield and wagon yield for the different treatments. A general linear model was constructed to test the possible effects of variety, maturity, flow-rate, and section length on the scale readings. RESULTS AND DISCUSSION The results of the scale yield compared to the weigh wagon yield are shown in figures 4 and 5. Figure 4 shows that the yield monitor predicted the wagon yield with a slope of 0.9 and an R-squared of This figure is constructed from the regression and 95 percent prediction data of the weigh wagon yield versus the scale yield across all row/sections lengths. Figure 5 is constructed from the regression and 95 percent prediction interval data of the wagon yield versus the scale yield on just the sections, not the whole rows. This figure indicated that the yield

13 ASABE Publications - Page 12 of monitor predicted the wagon yield with a slope of and an R- squared of The purpose of plotting both sets of data were to show that the short sections were indicative of the spot yield accuracy while the whole rows showed the expected accuracy for loading a wagon. A residual analysis showed no trends, reinforcing the validity of the statistical model. Table 2 shows the effect of the different parameters on the scale readings. Table 2. General Linear Model Results Parameter F-Value Probability Weight in the Wagon Variety Maturity (age) Speed (flow-rate) Length of row harvested Wagon yield had the highest F-value (which is to be expected since it is the value against which the scale readings were compared) while variety also had a partial effect on the readings. It is thought that this variable was significant because the LCP variety was a higher yielding cane and had more trash content than the other variety (which was measured by the weight plate, but not necessarily by the wagon if the secondary extractor fan removed it). Maturity, speed, and section lengths had very low F-values indicating an insignificant effect on the scale reading. Percent errors ranged from 0 to 33 percent but only 14 of the 118 tests were above 20 percent error. The average error was 11 percent; on average the yield monitor predicted the wagon yield with 89 percent accuracy. Sources of error were due to the large amount of vibration (induced by the moving parts of the harvester), tilt angle of the elevator changing as the machine moved through the field, and

14 Page 13 of 15 ASABE Publications fact that the weight plate was measuring material that might have been removed by the secondary extractor fan and not measured by the weigh wagon. The secondary extractor fan (fig. 1) removes anywhere from 2 to 10 percent of material. Figures 6 and 7 are the results of the Global Positioning System (GPS) for the 1999 and 2000 field tests. Both figures show the path of the harvester as it harvested the sugar cane during the field tests. The 1999 field tests had 122 m (400 ft) rows and the 2000 field tests had m (375 ft) rows. CONCLUSIONS A method for measuring yield in real-time was developed for a sugar cane billet harvester. This system consisted of a weight plate, electronics, and a GPS to record points with location. Results of the system indicated that it was able to predict yield with a slope of and an R-squared of The system was basic in nature and did not contain tilt sensors to correct the calibration equation or an accelerometer to offset the mass and momentum changes of the weight plate as the machine moved through the field. Still, the scale s overall average error was 11%. The results show that scales readings varied by variety, but that maturity of the cane, section length, and flow rate did not have a significant effect on the readings REFERENCES Benjamin, C.E Sugar Cane Yield Monitoring System. MS thesis. Baton Rouge, LA.: Louisiana State University, Dept of Biological and Agricultural Engineering. Benjamin, C.E., R.R. Price and M. P. Mailander, Sugar Cane Yield Monitoring System. ASAE Paper No ASAE, St. Joseph MI. Benjamin, C.E., and M. P. Mailander, Sugar Cane Yield Monitoring System. ASAE Paper No ASAE, St. Joseph MI. Bischoff, K.P., K.A. Gravois, H.P.Schexnayder, and G.L. Hawkins The Effect of Harvest Method and Plot Size on Sugarcane Yield. J. of the American Soc. of Sugar Cane Technology, vol. 21, pp. 3.

15 ASABE Publications - Page 14 of Colvin, T. S., K. L. Karlen, and N. Tischer Yield Variability Within Fields In Central Iowa. Automated Agriculture for the 21st Century, Proc. of the 1991 Symp. pp ASAE, St. Joseph, MI. Cox, G. J., H.D. Harris, and D.R. Cox Application of Precision Agriculture to Sugar Cane. Proc. of the Fourth International Conf. on Precision Agric. ASA-CSSA-SSSA, St. Paul, MN. Cox, G. J., H.D. Harris, D.R. Cox, D.M. Bakker, R.A. Pax, and S.R. Zillman. November 4, Mass Flow Rate For Sugar Cane Harvesters. Australian Patent No Cerri, D.G.P., G.R. Gray, F.S.G. Magalhaes Technological Improvement on Sugar Cane Yield Monitor. Proc. of the 9 th Intl. Conf. on Precision Agric. CD_ROM. ASA-CSSA-SSSA, Denver, CO. Davis, R.J., and E. J. Schmidt Review, Analysis and Discussion of Precision Agriculture Technologies. Final Report SRDC Project NCA 009. National Centre for Engineering in Agriculture, University of Southern Queensland, Toowoomba Queensland, Australia. Haley, Stephen, Jose Toasa, and Andy Jerardo. September Sugar and Sweeteners Outlook. USDA, Economic Research Service. SSS-253. Juniper Systems Inc Available at Accessed August 1, Karlen, D.L., E.J. Sadler, and W.J. Busscher Crop Yield Variation Associated With Coastal Plain Soil Map Units. Soil Science Soc. American J., vol. 54, pp Magalhães, F.S.G. and D.G.P. Cerri Yield Monitoring of Sugar Cane. Biosystems Eng. Volume 96, Issue 1, Pages 1-6. Pagnano, N.B., and P.S. G.Magalhaes Sugarcane Yield Measurement. Faculdade de Engenharia Agricola Unicamp Campinas SP, Brasil Proc. of the 3 rd European Conf. on Precision Agric., pp Montpellier, France. Price, R.R., John Larsen, and Alex Peters Development of an Optical Yield Monitor for Sugar Cane Harvesting. ASAE Paper No ASAE, St. Joseph, MI. (Sciortino, Dominick Jr., personal communication with the sugar cane combine production engineer, CAMECO, A John Deere Company, ) Wendte, K.W., A. Skotnikov, and K.K. Thomas. August 14, Sugar Cane Yield Monitor. United States Patent No

16 Page 15 of 15 ASABE Publications Figure 1. Sugar Cane Harvesting Process Figure 2. Weigh Plate Figure 3. Load Cell Calibration Figure 4. Wagon Yield vs. Scale Yield

17 ASABE Publications - Page 16 of Figure 5. Wagon Yield vs. Scale Yield (no whole rows) Figure GPS Field Track Figure GPS Field Track