6.5   Dynamic Testing of Force-Impetus Yield Monitors Under Rough Terrain Conditions

Principal Investigators

Thomas F. Burks, Post-Doctoral Scholar, Biosystems and Agricultural Engineering
Scott A. Shearer, Associate Professor, Biosystems and Agricultural Engineering
Larry G. Wells, Professor, Biosystems and Agricultural Engineering
Sam G. McNeill, Assistant Ext. Professor, Biosystems and Agricultural Engineering
John Fulton, Engineering Associate, Biosystems and Agricultural Engineering
Steve Higgins, Research Associate, Biosystems and Agricultural Engineering

Cooperators

Mike Ellis, Shelby County, Kentucky.

Introduction

            Researchers have conducted experiments in the laboratory to evaluate the accuracy of yield monitors while operating under stationary conditions (Kormann et al., 1998). These test have evaluated the influence of static variation in grain inflow and static slope conditions. The traditional yield monitor used in the United States is a force impetus system.  In a force based mass flow sensor grain impacts a curved plate mounted on a force transducer at the discharge point in the clean grain elevator. As a result, dynamic motion in the clean grain elevator could have a significant effect on instantaneous grain yield prediction.  The accuracy of any yield mapping technique is highly dependent upon the instantaneous yield monitor accuracy. It is therefore important to first quantify the type of dynamics which the combine elevator experiences at harvest and then use this criteria to conduct laboratory tests to determine what effects these dynamics have on the yield monitor accuracy.

            Test facilities capable of simulating the effects of dynamic motion on the accuracy of yield monitors have not been available to date. The recent development of the Yield Monitor Test Facility at the University of Kentucky offers a platform by which this valuable research could be conducted. Through the addition of dynamic simulation components and modifications to the control system, the yield monitor test facility would be capable of testing dynamic acceleration and undulating slope conditions. The results of these tests will potentially show any errors experienced when transversing rough terrain, (i.e., harvesting no-till crops, abrupt changes in slope associated with much of Kentucky cropland).

The objectives of this project are:

1)                  To conduct field experiments to establish ranges of accelerations and rotational motions at the top of the clean grain elevator;

2)                  To complete modification of the combine Yield Monitor Test Facility and control system to accommodate dynamic simulation of undulating terrain; and

3)                  To evaluate the accuracy of commercially available yield monitors under dynamic terrain simulation, using results from the field studies to define experiment parameters.

Background

            Significant research has been conducted during the past decade in an attempt to develop new combine yield monitor instrumentation, to quantify the errors which exist in current yield monitor technology, and to develop techniques for filtering yield monitor data to increase overall accuracy of the data. 

            A study was conducted by Shearer et al. (1997) in which four combines (two John Deere 9500's, a Gleaner R70 and a Case IH 1680) were used to harvest wheat within the same field. Each combine was equipped with yield monitors, the John Deere’s used GreenStar® and the other combines used Ag Leader 2000. The data from each combine was imported into a popular mapping package to compare the yield estimates. They found significant differences in measured yield existed between adjacent harvest swaths. These differences were attributed to potential machine/operator variability. In response to these differences, they developed data filtering techniques to help minimize the influence of calibration and machine/operator differences. They concluded that machine calibration techniques were critical in multiple-combine harvest systems.

            In a study conducted by Kettle and Peterson (1998), it was observed that the accuracy of combine yield monitors were significantly effected by hillside conditions and variation in harvest rates. Tests were conducted on two John Deere combines equipped with Greenstar® yield monitors in the hilly Palouse region of northern Idaho. They observed that the yield monitor’s response to variation in inflow was more exponential rather than linear and suggested that calibration should been done at several operating points rather than the traditional two point method. The observed yield monitor error was 20.3% when operating the combine at one-third the calibration throughput and 5.7% at half the calibration throughput. They also found that the error in yield monitor estimates were as high as 18.2% when harvesting uphill and 60.7% when harvesting downhill, on 6 to 9% slopes. Finally, they found a weak correlation between the yield monitor estimates and hand samples (R² =0.203) indicating that the yield monitor is not as sensitive to local variation in crop yield.

            A test facility has been developed by Kormann et al. (1998) to test multiple clean grain elevator yield monitors simultaneously. A system was built using a Massey Ferguson 38/40 elevator with three reference yield monitoring systems installed, the RDS Ceres II volumetric meter, MF Flow-Control, and the Ag Leader 2000 mass flow sensors. A second elevator was placed in series with the first elevator and was equipped with the CLASS Quantimeter II. It should be noted that this system did not use a fountain auger on either of the elevators. The system was designed so that the elevators could be tilted up to 15 degrees forward or aft with the potential for a simultaneous tilt of 15 degrees to port or starboard. In addition, a provision was made to vary mass flow rate from 1 ton/hr to 35 ton/hr. They conducted test varying the inflow rate from 10 to 35 ton/hr and found mean calibration errors less than 3%. Only at the lowest level did they experience errors of up to 7%. The volumetric sensors tended to perform better under low flow conditions than did the mass flow sensors. A second set of test were conducted by varying the tilt of the elevator system. They found mean errors up to 6% with standard deviations of up to 6%. The radiometric MF Flowcontrol device performed the best of all sensors with mean errors less than 2%, while the other sensors performed from most to least accurate as follows; CLASS Quantimeter II (<3%), AG Leader 2000 (<4%) and RDS Ceres II (<6%).

            Recent developments in yield monitor technology has resulted in improvements of yield monitor accuracy under static hill-side conditions, but to date, research has not determined the influence of rough terrain on yield monitor accuracy. All reported laboratory investigations have been conducted under static conditions.

Procedures

            The University of Kentucky Yield Monitor Test Facility will be used in this investigation to provide a closed loop grain supply to a John Deere 9600 clean grain elevator. In this system, a 17.5 cubic meter (500 bushel) bulk storage tank is used to supply grain to the elevator and another similar tank is used to receive grain from the elevator. The grain flow rates are controlled by a grain metering device capable of varying flow rates from 2 to 158 cubic meters per hour (50 to 4500 bushels per hour). The receiving tank is mounted on a weigh scale system which provides instantaneous mass flow and total mass flow. The clean grain elevator is equipped with both the standard John Deere GreenStar® and AG leader PF 3000 yield monitor systems, which includes the impact sensors, moisture meters, user interfaces, mapping processors, and  PC-MCIA cards for data storage. The UK Yield Monitor Test Facility has been developed to comply with the design requirements of ASAE Standard X578, (Proposed 1999).           The clean grain elevator from a John Deere 9600 combine was installed in a fixture which was designed for simulating hillside conditions. The fixture consists of a pitch and yaw pivoting support gimble , hydraulic cylinder mounting brackets and supports; hydraulic motor assembly to power elevator; elevator support tower; elevator infeed hopper; discharge grain collection hopper; and flexible spouting. A steel support frame is mounted horizontally on the pitch shaft using a sleeve bearing at each end of the shaft. This shaft allows the elevator to rotate for simulating forward to aft motion. Another shaft is mounted on top of the support frame (orthogonal to the pitch shaft) using sleeve bearings to simulate port to starboard motion. The elevator support tower sets on top of the yaw shaft and is attached to the sleeve bearings. The support tower serves to hold the clean grain elevator, the infeed hopper, the hydraulic motor, and the discharge grain collection hopper.

            The hillside simulation fixture is capable of pitch and yaw of 12 degrees  (21% slope) in either direction, and can accommodate simultaneous pitch and yaw. In its original configuration, the axes are rotated and held in position by turnbuckles.  However, for this study the system will be enhanced with the addition of a hydraulic power supply and two hydraulic cylinders equipped with flow control valves and rod position feedback sensors. These components will actuate the gimble to facilitate dynamic testing of rough terrain and hillside conditions for the clean grain elevator. In addition, the system software and electrical controls will be upgraded to provide a robust control system for the dynamic simulator.  The hydraulic flow control values are mounted on the cylinders and are controlled by an analog signal provided by the Yield Monitor Test Facility controller.  The cylinder’s position sensor provides feedback to the controller to allow closed-loop control of the cylinder actuation. This control feature provides a system capable of simulating the accelerations experienced in the field.

            Field experiments will be conducted to measure dynamic accelerations on the clean grain elevator during normal and adverse harvesting conditions.  A set of three accelerometers will be mounted on the combine clean grain elevator at the yield sensor location to measure the accelerations in the three orthogonal directions.  Data will be logged at harvest using a portable digital data acquisition system, and will be referenced to DGPS data to provide a geographic position record for all acceleration data.  This will provide a means for correlating unusual acceleration events to terrain disturbances.  In addition, yield monitor data will be collected. Applicable guidelines established in ASAE  Standard X579 (Proposed 1999) will be followed in designing the field experiments.

Yield monitor data, field elevation data and significant acceleration events will be mapped using ArcView®.  Pertinent themes will be overlaid to identify relationships between unusual acceleration events, actual field conditions and the measured grain yield.  Accelerometer data will be recorded for normal, undulating, rough and abrupt terrain conditions.  In turn, this data will be used to develop simulation profiles for the laboratory test.

            Laboratory investigations will be conducted using the modified UK Yield Monitor Test Facility. Three different terrain conditions will be simulated.  The first test series will simulate cyclic accelerations which might be experienced in a rolling terrain. The second test series will simulate rough terrain.  In these test, random accelerations of a moderate magnitude will be applied to the elevator.  The final test series will simulate abrupt disturbances (sinkholes or gullies). Extreme accelerations will be applied to the elevator. It should be noted that the test facility will be capable of two dimensional accelerations, all accelerations will be applied in the horizontal plane. Applicable guidelines established in ASAE  Standard X578 (Proposed 1999) for sample selection, test procedure, and test reporting will be followed in designing the laboratory experiments.

Each test will be conducted at three inflow levels (low, medium and high) and will be replicated three times.  At each inflow level, 300 yield monitor readings will be collected at one-second intervals for a total of five minutes.  The primary test variable to be evaluated in these test will be the instantaneous yield estimates. The results of these test will determine what  influence dynamic accelerations of the elevator housing have on instantaneous yield estimates.     

Expected Benefits

            As identified in Shearer et al. (1997), yield monitor data is occasionally susceptible to un-explained high and low yield predictions, which may be a result of dynamic conditions at the clean grain elevator. It is practical to filter these data points, but it would be preferable to identify them and thus develop a strategy for eliminating them. These series of test will attempt to determine if rough undulating terrain and the resulting accelerations of the combine elevator are responsible for these anomalies. Since a large percentage of the farmland in Kentucky is rolling hills, these studies may have a direct impact on the accuracy of grain yield prediction when using combine yield monitors. Dynamic combine yield monitor testing will be a unique development and could offer significant results regarding the accuracy of instantaneous yield prediction from existing yield monitors, as well as assisting designers in the development of new sensing technologies. 

Deliverables

            The results from these studies will be presented in refereed journal articles as warranted by the findings. It is expected that two journal articles will be developed through this research, one for the field acceleration test and a second for the laboratory testing of yield monitor response to dynamic accelerations. In addition, the results will be disseminated to the cooperating manufacturers for product improvement. Similarly, the findings will be documented in an extension publication to alert grain producers of observed sources of yield monitor error, and of possible calibration or operation practices to minimize errors.