Method and system for determining the amount of grain loss during the operation of a combine harvester

Abstract

The present disclosure describes a system and method for a virtual sensor that determines an amount of grain loss during a combine harvester operation without requiring any direct measurement of grain loss. An electronic controller is configured to determine values of a set of operating conditions including a plurality of sensor values and actuator settings. The electronic controller then applies an artificial neural network configured to receive the values of the set of operating conditions as input and produce a value indicative of an estimated amount of grain loss as output.

Description

Technical Field

[0001] This invention relates to systems and methods for operating combine harvesters. Specifically, it relates to monitoring grain loss during the operation of a combine harvester. Background Technology

[0002] Machines (such as combine harvesters) are used to collect crops such as wheat, oats, rye, barley, and corn from farmland. In some embodiments, the machine is also configured to separate grains from other materials (e.g., straw). However, grain loss can occur during the separation process. For example, some grains may not fall through the sieve in the cleaning screen but are carried to the rear and discharged with the straw, etc. This can happen, for example, when the sieve becomes clogged due to high moisture content, or when collecting high-moisture crops / vegetables. Grain loss can also occur, for example, when the cleaning screen is overloaded, when the sieve angle is inappropriate, when the sieve opening is too small, or when the fan airflow speed is too high. Grain loss can also occur at other stages of the combine harvester. For example, separation loss can occur when grains are discharged by the thresher to the chaff advancer instead of passing through the concave plate to reach the cleaning screen. In the case of harvesting corn, separation loss can include kernels still attached to the entire cob or small pieces of the cob, and separation loss may occur, for example, when the gap between the concave plate and the threshing drum is too small for the size of the cob being harvested. Summary of the Invention

[0003] In one embodiment, the invention provides a system configured to apply virtual sensors to determine grain loss during combine harvester operation. An electronic controller is configured to apply actuator settings to each of a plurality of actuators to control the operation of the combine harvester and to receive output signals from each of a plurality of sensors, wherein the output signal from each sensor indicates different operating states of the combine harvester. An electronic processor determines values ​​for a set of operating parameters and then applies an artificial neural network configured to receive the values ​​of the set of operating parameters as input and produce values ​​as output indicating an estimate of grain loss. The set of operating parameters received as input by the artificial neural network does not include any direct measurement of grain loss.

[0004] In another embodiment, the present invention provides a method for determining the amount of grain loss during combine harvester operation. An electronic processor determines values ​​for a set of operating parameters that do not include any direct measurements of grain loss. The electronic processor then applies an artificial neural network configured to receive the values ​​of the set of operating parameters as input and produce values ​​as output indicating an estimate of the amount of grain loss.

[0005] Other aspects of the invention will become apparent from consideration of the specific embodiments and the accompanying drawings. Attached Figure Description

[0006] Figure 1 This is a front view of a combine harvester according to an embodiment.

[0007] Figure 2 It is used in Figure 1 A schematic diagram of the system in a combine harvester that separates grains from other materials.

[0008] Figure 3 It is used for Figure 1 A block diagram of the control system of a combine harvester.

[0009] Figure 4 It is used to determine Figure 1 A schematic diagram of an artificial neural network for measuring grain loss in a combine harvester.

[0010] Figure 5 It is used to determine separately Figure 1 A schematic diagram of an artificial neural network representing grain loss in a combine harvester due to separation loss and cleaning screen loss.

[0011] Figure 6 It is used to determine the redundancy mechanism of artificial neural networks. Figure 1 A flowchart illustrating methods for addressing grain loss in combine harvesters.

[0012] Figure 7 It is used to determine whether artificial neural networks are used selectively as a redundancy mechanism or as a single mechanism. Figure 1 A flowchart illustrating methods for addressing grain loss in combine harvesters. Detailed Implementation

[0013] Before explaining any embodiments of the invention in detail, it should be understood that the invention is not limited in its application to the details of the construction and arrangement of the components set forth in the following description or shown in the following drawings. The invention can have other embodiments and can be practiced or implemented in various ways.

[0014] Figure 1 An example of a combine harvester 100 configured to collect (i.e., “harvest”) crops, including, for example, wheat, oats, rye, barley, and corn, is shown. The combine harvester 100 is also configured to separate grains from other materials, such as straw. The combine harvester 100 includes a header 104 and a feeding mechanism 106. The header 104 is configured to collect and cut the crops, and the feeding mechanism 106 is configured to convey the cut crops from the header 104 to a separation system housed in the body 108 of the combine harvester 100.

[0015] Figure 2 An example of the separation system 112 in a combine harvester 100 is shown. The threshing drum 116 includes rasp bars 118 that separate grains and husks from the straw. Most of the straw is conveyed from the threshing drum 116 to a straw collector 126, which is configured to move the straw toward the rear of the combine harvester 100, where it is discharged. Grains, husks, and some straw separated from the remaining crop by the threshing drum 116 pass through openings in a concave plate 114 and reach a cleaning screen 132. The cleaning screen 132 includes an upper screen 120 (or husk screen) and a lower screen 122. The screens 120 and 122 include adjustable fingers that allow material smaller than a certain size (e.g., grains) to fall through. Grains falling through sieves 120, 122 enter the collection area, where they are then conveyed to a grain bin by screw conveyor 124. Materials other than grains (e.g., straw) that are too large to fall through the openings in sieves 120, 122 into the collection area are instead carried away from the rear of sieves 120, 122 by airflows A1, A2 from fan 134 and / or by the regular, repetitive vibrations of sieves 120, 122.

[0016] However, in Figure 2 During the separation process shown, grains may be lost. For example, some grains may not fall through the sieves 120 and 122 in the cleaning sieve, but may be carried to the rear and discharged with the straw, etc. This may occur, for example, when sieves 120 and 122 become clogged due to high humidity, or when collecting high-moisture crops / vegetables. Grain loss may also occur when the cleaning sieve is overloaded, when the angle of sieves 120 and 122 is inappropriate, when the openings of sieves 120 and 122 are too small, or when the airflow speed of fan 134 is too high. Grain loss occurring at cleaning sieve 132 is referred to herein as "cleaning sieve grain loss".

[0017] Grain loss can also occur at other stages of the combine harvester 100. For example, “separation loss” occurs when grains are discharged by the thresher 116 to the thresher 126 instead of passing through the concave plates 114 to the cleaning screen 132. In the case of harvesting corn, separation loss can include grains still attached to the entire cob or small pieces of the cob, and separation loss may occur, for example, when the spacing between the concave plates 114 and the thresher drum 116 is too small for the size of the cob being harvested. Separation loss may also occur, for example, when: (1) the rotor runs too slowly and therefore cannot separate the grains from the waste; (2) the harvester header 104 collects an excessive amount of waste; and (3) the spacing of the concave plates 114 is too wide. Grain loss that occurs after the material leaves the feeding mechanism and before the material enters the cleaning screen is referred to herein as “separation loss”.

[0018] One method for measuring grain loss is to use physical sensors. For example, a physical grain loss sensor can be configured to detect grains impacting the surface of a measuring device by detecting contact (e.g., a capacitive or piezoelectric sensor) or by detecting the sound of grains striking the surface. However, various examples further described below provide machine learning or artificial intelligence mechanisms configured to operate as “virtual sensors” for estimating the amount of grain loss.

[0019] Figure 3 An example of a control system for a combine harvester 100 is shown, which is configured to apply an artificial neural network mechanism to estimate grain loss. The system includes a controller 301, which includes an electronic processor 303 and a non-transitory computer-readable storage 305. The storage 305 stores data and instructions that, when executed by the electronic processor 303, implement the functions of the controller 301 (e.g., including the functions described herein).

[0020] The controller 301 is communicatively connected to a plurality of actuators and configured to provide control signals to the actuators to regulate the operation of the combine harvester 100. For example, the controller 301 is communicatively connected to one or more actuators of the feeder 307, which control the speed at which crop material is fed into the separating mechanism 112 via the feeding mechanism 106. The controller 301 is also configured to provide control signals to the fan 309 to regulate the operating speed of the fan 309, to the husk sieve actuator 311 to regulate the opening size of the sieves 120, 122, and to the threshing actuator 313 to regulate the rotational speed of the threshing drum 116 and / or the distance between the threshing drum 116 and the concave plate 114.

[0021] exist Figure 3In this example, controller 301 is also communicatively connected to multiple sensors that provide feedback and / or other information regarding the operating status of combine harvester 100. For example, controller 301 is connected to a hygrometer 315 and an ambient light sensor 317, the hygrometer 315 being configured to provide a signal indicating the measured humidity, and the ambient light sensor 317 being configured to measure the amount of ambient light. Controller 301 may also be connected to tilt sensors 319 and 321, which are configured to provide information to controller 301 regarding the lateral and longitudinal tilt of the field surface on which combine harvester 100 operates. Controller 301 is also communicatively connected to a physical grain loss sensor 323, which is configured to provide an output signal indicating the measured amount of grain loss.

[0022] exist Figure 3 In the example, controller 301 is also communicatively coupled to toggle switch 325 and / or other user input controls, as well as wireless transceiver 327 for communicating with a remote computer system. In various different embodiments, controller 301 can be configured to communicate with, in addition to, Figure 3 The example shows sensors and actuators other than or replacing them. Figure 3 The example illustrates communication between the sensor and actuator and other sensors or actuators. Furthermore, in various implementations, the controller 301 can be configured to communicate with the sensors and actuators via a wired communication interface, a wireless communication interface, or a combination of wired and wireless interfaces.

[0023] Figure 4 An example of an artificial neural network (ANN) designed to estimate / predict grain loss based on machine settings, configuration, and runtime input is shown. Figure 4 In a specific example, the artificial neural network receives the total feed rate (e.g., based on the current operating settings of the feeder 307), humidity (based on the output of the hygrometer 315), longitudinal tilt (based on the output of the longitudinal tilt sensor 319), lateral tilt (based on the output of the lateral tilt sensor 321), fan speed (based on the current operating settings of the fan 309), husk sieve opening size (based on the current operating settings of one or more husk sieve actuators 311), and ambient light measurements (based on the output of the ambient light sensor 317) as inputs. Figure 4 The artificial neural network is trained to generate an estimate of grain loss (e.g., grain loss rate) based on these inputs. Through continuous retraining of the artificial neural network, the system will be able to better identify the inputs causing grain loss. Therefore, in various different implementations, the artificial neural network can be configured to receive more inputs, fewer inputs, or more inputs than... Figure 4 The examples listed are other than or replaceable inputs. Figure 4 Other inputs listed in the example.

[0024] Figure 4 The example is configured to produce a single output indicating "grain loss". Therefore, in some implementations, a single artificial neural network can be implemented to produce an estimate of the total grain loss. However, in other implementations, one or more artificial neural networks can be trained to estimate specific types of grain loss. For example, controller 301 can be configured to use an artificial neural network trained to estimate only separation loss. In another example, controller 301 can be configured to use multiple different artificial neural networks, one trained to estimate separation loss and another trained to estimate cleaning screen loss. In other implementations, controller 301 can also be configured to apply artificial neural networks trained to produce multiple outputs, each indicating a different type of grain loss. For example, Figure 5 An artificial neural network is shown, which is configured to receive and Figure 4 The artificial neural network takes the same input but is trained to produce two outputs: an estimate of the grain separation loss and an estimate of the grain loss at the cleaning screen. Other implementations can be configured to use an artificial neural network trained to estimate different types of grain loss, such as pre-harvest loss, header loss, and leakage loss.

[0025] exist Figure 4 and Figure 5 The artificial neural network shown in the example receives various sensor values ​​and actuator values ​​as input. However, the artificial neural network does not receive the measured amount of grain loss from the grain loss sensor 323 as input. Therefore, the controller 301 can be configured to use the artificial neural network mechanism as a redundancy mechanism (to detect problems or inconsistencies in the hardware sensors) or as a single grain loss sensor (to provide an alternative mechanism for measuring grain loss). In some embodiments, the combine harvester 100 can be configured to exclude any physical grain loss sensor 323 and instead use the artificial neural network mechanism as the sole method for estimating / measuring grain loss.

[0026] Figure 6 An example of a method for verifying the correct operation of a physical grain loss sensor 323 (e.g., a physical grain loss sensor configured to detect separation loss) using an artificial neural network as a redundancy mechanism is shown. The controller 301 receives the output signal from the grain loss sensor 323 and determines the amount of separation loss based on the sensor output (step 601). The controller 301 also applies an artificial neural network (e.g., Figure 4The controller 301 uses an artificial neural network to determine an estimate of the separation loss (step 603). The controller 301 compares the two determined grain loss values, and if the difference is within a defined tolerance threshold (step 605), the controller 301 determines that the physical grain loss sensor 323 is operating normally and limits the separation loss based on the sensor output (step 607). However, if the difference is greater than the tolerance threshold, the controller 301 determines that there is an error in the output from the physical grain loss sensor 323 (step 609) and limits the separation loss based on the output of the artificial neural network (i.e., the "virtual sensor") (step 611).

[0027] In some implementations, controller 301 is configured to update and retrain the artificial neural network based on a set of inputs and a defined output determined by physical grain loss sensor 323. For example, in Figure 6 In the method, controller 301 is configured to retrain the artificial neural network in response to determining that the physical grain loss sensor 323 is operating normally (step 613). This can be achieved, for example, by using "supervised learning," in which the output of the physical grain loss sensor 323 is provided as a defined "output" corresponding to a current set of "inputs." In some embodiments, controller 301 may be configured to perform the retraining operation locally, while in other embodiments, the set of inputs and the defined "output" are transmitted to a remote computer system (e.g., via wireless transceiver 327), which retrains the artificial neural network and transmits the updated artificial neural network back to controller 301 for future use. In some embodiments, the artificial neural network is retrained based only on data from a single combine harvester 100, while in other embodiments, the artificial neural network is configured to be retrained using data from multiple different combine harvesters. For example, the artificial neural network may be retrained based on data from multiple combine harvesters in a convoy.

[0028] exist Figure 6 In the example, in response to determining that the physical grain loss sensor 323 is operating normally, the separation loss is defined based on the output of the physical grain loss sensor 323 (step 607). In some embodiments, the controller 301 may be configured to achieve this by simply defining the separation loss as equal to the output of the grain loss sensor 323. However, in other embodiments, the controller 301 may be configured to determine the amount of grain loss based on both the output of the physical grain loss sensor 323 and the output of a “virtual sensor” based on an artificial neural network. For example, the controller 301 may be configured to define the separation loss as the average of the output from the physical grain loss sensor 323 and the output from the virtual sensor.

[0029] Figure 6 The example requires both a "virtual sensor" and a physical grain loss sensor 323 as redundant mechanisms for measuring grain loss. However, in other embodiments, the physical grain loss sensor 323 can be completely omitted and replaced by a virtual sensor. For example, the controller 301 can be configured to limit total grain loss, separation loss, and / or cleaning screen loss based on the output of one or more artificial neural networks. Therefore, the cost of the combine harvester 100 can be reduced by replacing the physical grain loss sensor 323 with a virtual sensor.

[0030] Similarly, in some implementations, the system can be configured to allow the operator to manually select whether to use a virtual sensor as a redundant mechanism for physical grain loss sensors, or to use only a virtual sensor as a “single” mechanism for determining grain loss. Figure 7 An example of such a method is shown. Controller 301 monitors the state of a user-input control (e.g., toggle switch 325) (step 701) to determine whether the operator has selected a “redundant mode” or a “single mode” (step 703). If the operator selects a “single mode” (i.e., by moving the toggle switch to the first position), controller 301 uses a “virtual sensor” to apply an artificial neural network to determine grain loss (step 705). Conversely, if the operator selects a “redundant mode” (i.e., by moving the toggle switch to the second position), controller 301 applies a similar method... Figure 6 The method shown uses the output of the "virtual sensor" as redundancy for the output of the physical grain loss sensor (step 707).

[0031] Therefore, the present invention particularly provides systems and methods for virtual grain loss sensors using artificial neural networks configured to estimate grain loss based on machine operating settings and / or other sensor outputs. Various features and advantages of the invention are set forth in the following claims.

Claims

1. A method for determining grain loss during combine harvester operation, the method comprising: The electronic processor determines the values ​​of a set of operating parameters, which do not include any direct measurements of grain loss. as well as The electronic processor applies an artificial neural network, which is trained to receive the set of operating parameters as input and produce an estimate of grain loss as output. The electronic processor monitors the state of the toggle switch to determine whether the toggle switch is in the first position or the second position; In response to determining that the toggle switch is in the first position, the amount of grain loss is limited based on the output of the physical grain loss sensor, while the output of the artificial neural network is used to confirm the correct operation of the physical grain loss sensor; as well as In response to determining that the toggle switch is in the second position, the amount of grain loss is limited based on the output of the artificial neural network. Determining the values ​​of the set of operating parameters includes: At least one sensor value is determined based on the output of the electronic processor, which indicates the state measured by the sensor. At least one actuator setting is determined, wherein the electronic processor is configured to control the combine harvester by applying the at least one actuator setting to the actuators of the combine harvester.

2. The method according to claim 1, wherein, Determining the values ​​of the set of operating parameters includes determining the values ​​of the following items: The total feeding rate of the combine harvester humidity, longitudinal tilt Lateral tilt The speed of the fan, which is configured to blow airflow through at least one sieve of the cleaning screen of the combine harvester. The size of the rice husk screen opening of the cleaning screen of the combine harvester, and The amount of ambient light.

3. The method according to claim 1, further comprising: The electronic processor receives an output from the physical grain loss sensor indicating the amount of grain loss measured by the physical grain loss sensor; as well as The correct operation of the physical grain loss sensor is confirmed at least in part by comparing the output from the physical grain loss sensor with the output of the artificial neural network.

4. The method according to claim 1, further comprising: The electronic processor receives an output from the physical grain loss sensor indicating the amount of grain loss measured by the physical grain loss sensor; as well as The artificial neural network is retrained using supervised machine learning by using the set of operating parameters as input and the output of the physical grain loss sensor as the defined output.

5. The method of claim 1, further comprising adjusting at least one actuator setting of the combine harvester by the electronic processor based at least in part on the determined amount of grain loss.

6. The method according to claim 1, wherein, The application of the artificial neural network includes a first artificial neural network trained to produce a value indicating the amount of separation loss as its output, and a second artificial neural network trained to produce a value indicating the amount of cleaning screen loss as its output.

7. The method according to claim 6, wherein, The second artificial neural network is configured to receive a second set of operating parameters as input, wherein the second set of operating parameters is different from the set of operating parameters received as input by the first artificial neural network.

8. The method according to claim 6, wherein, The set of operating parameters received as input by the second artificial neural network is the same as the set of operating parameters received as input by the first artificial neural network.

9. A system configured to apply virtual sensors to determine grain loss during combine harvester operation, the system comprising: Multiple actuators; Multiple sensors; as well as A toggle switch capable of selectively positioning in a first and a second position, wherein the electronic processor is further configured to: The toggle switch is monitored to determine whether it is in the first position or the second position. In response to determining that the toggle switch is in the first position, the amount of grain loss is limited based on the output of the physical grain loss sensor, while the output of an artificial neural network is used to confirm the correct operation of the physical grain loss sensor. In response to determining that the toggle switch is in the second position, the amount of grain loss is limited based on the output of the artificial neural network. Electronic controller, the electronic controller being configured to: The actuator settings are applied to each of the plurality of actuators to control the operation of the combine harvester. Output signals are received from each of the plurality of sensors, wherein the output signals from each sensor indicate different operating states of the combine harvester. Determine the values ​​of a set of operating parameters, which do not include any direct measurements of grain loss, and An artificial neural network is applied, which is trained to receive the set of operating parameters as input and produce a value indicating the amount of grain loss as output.

10. The system according to claim 9, wherein, The set of operating parameters includes at least one selected from the group consisting of: The total feeding rate of the combine harvester The measured humidity, longitudinal tilt Lateral tilt The speed of the fan, which is configured to blow airflow through at least one sieve of the cleaning screen of the combine harvester. The size of the rice husk screen opening of the cleaning screen of the combine harvester, and The amount of ambient light.

11. The system of claim 9, further comprising a physical grain loss sensor configured to directly measure grain loss and output a signal indicating the amount of grain loss measured by the physical grain loss sensor, wherein the electronic processor is further configured to: Receive output from the physical grain loss sensor, and The correct operation of the physical grain loss sensor is confirmed at least in part by comparing the output from the physical grain loss sensor with the output of the artificial neural network.

12. The system according to claim 9, wherein, The electronic processor is also configured to, when the toggle switch is in the first position, determine that the physical grain loss sensor is not operating correctly based on a comparison of the output of the physical grain loss sensor and the output of the artificial neural network, and limit the amount of grain loss based on the output of the artificial neural network.

13. The system of claim 9, further comprising a physical grain loss sensor configured to directly measure grain loss and output a signal indicating the amount of grain loss measured by the physical grain loss sensor, wherein the electronic processor is further configured to: Receive output from the physical grain loss sensor, and The artificial neural network is retrained using supervised machine learning by using the set of operating parameters as input and the output of the physical grain loss sensor as the defined output.

14. The system according to claim 9, wherein, The electronic processor is also configured to adjust at least one actuator setting of the combine harvester based at least in part on the determined amount of grain loss.

15. The system according to claim 9, wherein, The electronic processor is configured to apply the artificial neural network in such a way that it applies a first artificial neural network trained to produce values ​​indicating the amount of separation loss as its output, and The electronic processor is further configured to apply a second artificial neural network trained to generate values ​​indicating the amount of cleaning screen loss as output.

16. The system according to claim 15, wherein, The second artificial neural network is configured to receive a second set of operating parameters as input, wherein the second set of operating parameters is different from the set of operating parameters received as input by the first artificial neural network.

17. The system according to claim 15, wherein, The set of operating parameters received as input by the second artificial neural network is the same as the set of operating parameters received as input by the first artificial neural network.

Images