JPH04291133A - Device for detecting amount of grain in thrashing device - Google Patents

Abstract

PURPOSE:To obtain an amount of leakd grain as a fourth loss of a thrashing device from a detection pattern of a piezoelectric sensor. CONSTITUTION:A pad net 32 is provided at a lower portion of a handling drum 19 which thrashes a harvested ear and a winnowing mechanism 25 is provided at the lower portion. A discharged ear can be released from a ear exhaust exit 24 at a rear portion of the winnowing mechanism 25 to the outside of a device. A piezoelectric sensor 57 for detecting an object to be detected which is leaked and dropped from the pad net 32 is provided between the pad net 32 at the lower portion of a part closer to a rear portion of the handling drum 19 and the winnowing mechanism 25. A neutral network is constructed with a detection pattern of only the grain according to the piezoelectric sensor 57 and a detection pattern of only the exhaust ear as learning data and a central control device for detecting an amount of leaked and dropped grain according to a detection waveform of the object to be detected which is leaked and dropped from the pad net is provided.

Description

[Detailed description of the invention]

0001

[Industrial Application Field] The present invention is a method for threshing harvested grain culms using a threshing device in a general-purpose combine harvester or the like for harvesting rice, wheat, soybeans, etc. This invention relates to a device that detects the amount of grain mixed into the discharged culm when the culm is discharged outside the machine.

0002

[Prior Art] As disclosed in the prior art Japanese Utility Model Application Publication No. 169940/1983, in a general-purpose combine harvester for harvesting rice, wheat, soybeans, etc., the front of a handling chamber containing a built-in screw-type handling barrel is A pre-harvesting treatment device is provided to harvest and introduce the planted grain culms from the root side. A receiving net (corn cave) is installed below the handling room, and a sorting mechanism for performing rocking sorting and wind sorting is installed below the net. Further, a waste culm processing chamber containing a built-in waste culm processing device is provided in communication with the rear of the sorting mechanism, and the entire amount of the harvested grain culms is introduced into the handling chamber and threshed by the screw-type handling barrel. The threshed grains and straw waste that pass through the receiving net and fall are subjected to a rocking sorting action and a wind sorting action in the sorting mechanism. While waste such as straw waste is carried by the wind of the sorting fan in the sorting mechanism and sent to the culm processing room, this straw waste and the waste culm that falls from the end of the receiving net at the rear of the handling cylinder are The waste culm etc. are discharged to the outside of the machine by the waste culm processing device in the waste culm processing chamber. As shown in Japanese Patent Application Laid-Open No. 62-263418, although the shape of the handling teeth protruding from the outer periphery of the handling barrel is different in ordinary threshing devices such as combine harvesters, other basic structures are substantially the same. .

[0003] In this case, the culms that fall from the terminal end of the receiving net are often in the form of clumps, and grains called sari grains are mixed into these culms. If such culms are discharged outside the machine as they are, useful grains will be lost. This loss is usually called No. 4 loss. In order to reduce such No. 4 loss, the waste culms that fall from the end of the receiving net are received at the sieve line part of the swinging sorting mechanism and shaken, and the grains are subjected to wind sorting action by the sorting wind to separate the grains. I'm trying to lead it to the turn back gutter. Therefore, a piezoelectric sensor is disposed behind the terminal end of the receiving net, and this piezoelectric sensor detects both the falling grains and the culm. However, simply integrating the detection values of the piezoelectric sensor is insufficient.
There was a problem in that the amount of grains due to grain loss could not be accurately detected.

Therefore, in Japanese Patent Application Laid-Open No. 62-263418, the size of the vortex flow rate of the air generated when the grains falling at the sieve line are subjected to the sorting air is measured by using a superstructure installed in the discharge culm processing chamber. It is proposed to indirectly detect the amount of the second reduction target material by detecting it with a sonic sensor.

[0005]

[Problem to be Solved by the Invention] However, even in the case of this indirect detection, since the culm is riding on the sorting wind that comes to the culm processing chamber, the amount of culm that flows with the sorting wind can be exceeded. We were unable to solve the problem of not being able to accurately detect the amount of grains because the sonic sensor was used to detect them. The present invention focuses on the dissimilarity between the waveform pattern detected by the piezoelectric sensor for grains and the waveform pattern detected by the culm, and processes the detection results of the piezoelectric sensor using a new data processing method called a neural network. It is an object of the present invention to solve the above-mentioned problem with a control means that can be used.

[0006]

[Means for Solving the Problems] In order to achieve this object, the present invention provides a receiving net below the handling barrel for threshing the harvested grain culms, a sorting mechanism below the receiving net, and a sorting mechanism. In a threshing device configured to discharge waste culm outside the machine from a rear waste culm outlet, there is a space between the receiving net below the rear portion of the handling barrel and the sorting mechanism, where water leaks from the receiving net. A piezoelectric sensor is provided to detect objects to be detected, and a neural network is constructed using the waveform detected by the piezoelectric sensor for only grains and the waveform detected only for culms as learning data, and a neural network is constructed to detect objects leaking from the receiving net. A control means is provided for detecting the amount of grain leakage from the detected waveform of the detected object.

[0007]

[Example] Next, an example applied to a general-purpose combine harvester will be described. In the figure, reference numeral 1 indicates a traveling crawler 2
4 is a threshing section provided on one side of the machine 1, 5 is a grain tank from which grain is taken out via a grain lifting cylinder, 8 is an engine provided at the rear of the grain tank 5, Reference numeral 9 indicates a reaping pre-processing unit mounted on the front part of the machine base 3 so as to be able to be raised and lowered via a hydraulic cylinder 10, 11 indicates a driver cabin located in front of the grain tank 5, and 12 indicates the grains in the grain tank 5. This is a carry-out auger that takes out the machine.

The reaping pre-processing section 9 has a rectangular cylindrical feeder house 13 that opens at the front of the threshing section 4 and can be raised and lowered by a hydraulic cylinder 10, and an oblong bucket-shaped platform attached to the front end of the feeder house 13. 14, a grain culm scraping auger 15 which is installed horizontally within the platform 14 and rotates in the direction of the arrow in FIG. and a clipper-shaped cutting blade 17 installed horizontally at the front end of the lower surface of the platform 14 for reaping the roots of the planted grain culms. The grains are collected at approximately the center of the left and right sides of the grains, and sent to the threshing section 4 by the supply chain conveyor 18 in the feeder house 13.

Reference numeral 19 denotes a longitudinal axis of rotation 21 in the handling chamber 20 of the threshing section 4 along the traveling direction of the combine harvester.
The handling cylinder 19 has screw blades 22, which are screw-shaped handling teeth, wound around the outer periphery of the body, and a stem culm scattering plate 23 is attached to the rear end of the outer periphery of the handling cylinder 19. erected.

Reference numeral 32 indicates a handling chamber 2 located below the handling cylinder 19.
The receiving net 25 is stretched leaving the dust exhaust port 24 at the rear end of the 0, and the reference numeral 25 indicates a sorting mechanism disposed below the handling chamber 20. The sorting mechanism 25 includes a swing plate 28 that swings back and forth through swing links 26 and 27, and includes a feed pan 29 located at the lower front part of the handling cylinder 19, and a feed pan 29 provided at the rear of the feed pan 29. A first chaff sieve 30, a sorting net 31 stretched below the first chaff sieve 30, and a first chaff sieve 30 connected to the rear of the first chaff sieve 30.
A sieve (sieve) line 33, a second chaff sieve 34 provided at the rear lower part of the first sieve line 33, a second sieve line 35,
A first grain plate 36 and a second grain plate 37 are provided on a swinging plate 28.

Reference numeral 38 is a dust removal fan that supplies sorting air to the upper surface side of the feed pan 29, and reference numeral 39 is the first grain grain plate 3.
A first sorting fan 40 supplies sorting air toward the first grain plate 36 and sends sorting air toward a waste culm processing chamber 42 (to be described later); First conveyor with gutter 41, code 43 is second chaff sieve 3
4, and the culm processing chamber 4, which will be described later.
This is a second sorting fan that sends sorting air toward No. 2.
The sorting fan 43 is arranged at a position closer to the discharge culm processing chamber 42 than the first sorting fan 39.

Reference numeral 44 designates a second conveyor with a second gutter 45 for delivering the second processed product, which is a reduced product such as grains from the second flow grain plate 37, to the second reduction cylinder 46; Cylinder 46
The upper end (discharge port) is configured to open at a position close to the front end of the handling cylinder 19 in the handling chamber 20 or open onto the feed pan 29 to return the second processed material. Dust exhaust port 2
4 and the rear end of the sorting mechanism 25, the inclined culm shedding plate 47 at the lower end of the culm processing chamber 42 faces the rear end side of the swing plate 28, and the culm processing chamber 42 includes a culm processing device Rotating diffusion tooth 4
8 is attached to the horizontal rotation shaft 49, and the fixed teeth 50 are arranged to face the side of the rotation diffusion teeth 48 so as to overlap the rotation locus of the plurality of rotation diffusion teeth 48 which are rotationally driven in the direction of the arrow. 42 so as to be able to protrude and retract upward from the bottom plate 54.

A plurality of culm guide rods 51 whose base ends are attached to the dust exhaust port 24 are hung down on the sides of the rotational diffusion teeth 48 so as to overlap the rotation locus, and each culm guide rod 51 has a base end attached to the dust exhaust port 24. It is disposed at the front position of the culm processing chamber 24 so as to extend from above to below. The exhaust culm outlet 52 provided facing backward at the bottom of the exhaust culm processing chamber 42 is formed in a horizontally elongated manner so that the exhaust culm scraped by the rotating diffusion teeth 48 is discharged to the outside of the machine. The first sorting fan 39 and the second sorting fan 43 are installed on the upper rear end side of the culm processing chamber 42.
An air exhaust passage 53 for discharging the wind from the machine to the rear of the machine base 1 is provided so as to be open.

Reference numeral 55 denotes a cutting blade provided at a rear portion of the outer periphery of the handling cylinder 19, and a fixed blade 56 protruding into the handling chamber 20 so as to be adjacent to the rotation locus of the cutting blade 55. , at a rear part of the outer periphery of the handling cylinder 19, cut the culm sent in the form of a lump into short pieces, or loosen the culm, and receive the grains (grains mixed in) in the culm from below from the net 32. to leak.

With the above configuration, the grain culms that have been raked up by the reel 16 in the reaping pre-processing section 9 and whose root part has been cut by the cutting blade 17 are collected by the raking auger 15 in the platform 14.
The raw materials are scraped together and supplied to the front end of the handling cylinder 19 via the supply chain conveyor 18 in the feeder house 13. Handling room 20
Inside, the grains are sequentially shed from the grain culm by the rotation of the handling cylinder 19.
The grains and straw waste that have fallen through the receiving net 32 are sorted by a sorting mechanism 25.
Under the oscillating sorting action and the wind sorting action of the first sorting fan 39, only the grains are collected in the first gutter 41 and transferred to the first conveyor 4.
The grains are collected in the grain tank 5 from 0 through the grain barrel.

The processed material containing a lot of straw gathers at 34 locations of the second chaff sieve, which is the rear part of the sorting mechanism, and is moved rearward by the oscillating sorting action, while being subjected to the wind sorting action by the second sorting fan 43. The straw waste and dust are blown toward the exhaust culm processing chamber 42, and together with this exhaust air, the straw waste and dust can be smoothly discharged out of the machine through the exhaust passage 53 on the upper rear end side of the exhaust culm treatment chamber 42. .

On the other hand, grain culms discharged from the dust exhaust port 24 at the rear end of the handling cylinder 19 are received by the front surface of the culm guide rod 51, fall in the direction of the rotating diffusion teeth 48, and are scraped out by the culm processing device 48. Since the air is discharged outside the machine from the exhaust culm outlet 52, the exhaust culm does not clog the exhaust passage 53. In other words, light items such as straw and dust are discharged out of the machine along with the exhaust air through the exhaust passage 53 on the upper side of the exhaust culm processing chamber 42, and large waste culms are discharged out of the machine. From the top to the bottom of the culm treatment chamber,
Since the discharge part is separated into two parts corresponding to the nature of the material to be discharged, the sorting air will not flow back toward the sorting mechanism inside the culm processing chamber. Therefore, the amount of straw and dust falling onto the second flow grain board 37 is reduced, and as a result, the number of straw and dust mixed with the second processing material that collects from the second gutter 45 to the second conveyor 44 is reduced. The second reduction cylinder 4
6 to the front part of the handling cylinder 19, the handling cylinder 1
The amount of waste to be discharged from the dust outlet 24 at the rear end of the filter 9 is reduced, and the loss of No. 4 grains can be reduced. In addition, the No. 2 treated material with less straw waste and dust is placed in the No. 2 reduction cylinder 4.
6 to the front feed pan 29 and first chaff sieve 30 side of the swing plate 28, the amount of small particles mixed into the first gutter 41 is drastically reduced, and furthermore, the third loss (exhaust culm processing chamber 42) is reduced. It is also possible to reduce the loss caused by grains discharged outside the machine from the lower culm outlet 52.

Reference numeral 57 is provided at the rear side of the handling barrel 19 from where the cutting blade 55 and fixed blade 56 are installed, and below the receiving net 32 (above the chaff sheave 34). Multiple piezoelectric sensors. In the embodiment, the detection surface of each of the three piezoelectric sensors 57 is arranged so that grains leaking through the receiving net 32 are likely to collide with each other as the handling cylinder 19 rotates (see FIG. 4).

FIG. 4 shows a block diagram of the control means of the present invention, in which reference numeral 58 is an amplifier for amplifying the analog input signal of each piezoelectric sensor, and 59 is an amplifier for performing neural network processing, etc., which will be described later. 60 is a read-only memory (ROM) that stores in advance control programs and initial values necessary for calculation processing by the central controller 59;
Reference numeral 61 is a read/write memory (RAM) for temporarily storing various data used for arithmetic processing;
Reference numerals 62 each indicate an input/output (I/O) interface unit. In addition, an actuator 63 represented by an electromagnetic solenoid driven in accordance with a command signal from the central control device 59 is used to change the speed of the threshing section such as the rotation speed of the handling drum 19 or to change the traveling speed of the traveling body 1. , change the threshing process speed.

Next, a method of processing signal data detected by the piezoelectric sensor 57 will be explained. Through experiments, only an appropriate amount of grain was allowed to leak from the receiving net 32 at a rear portion of the rotating handling drum 19. An example of a detected waveform pattern when the leaking grain collides with the inspection surface of the piezoelectric sensor 57 is shown in FIG. 5, with time t plotted on the horizontal axis and detected voltage plotted on the vertical axis. Similarly, FIG. 6 is an example of a detected waveform pattern when only the culm collides with the piezoelectric sensor 57.

[0021] The m experimental pattern data of only the grains, K1, K2, K3, ..., Ki, ..., Km, and the m experimental pattern data of only the culms, H1, H2, H3,
, Hj, and Hm are prepared. In each of the experimental pattern data, for example, in the case of only grains in FIG. 5, sampling is performed for T time (for example, 150 milliseconds) from the time when the detected signal value becomes v0 or more, and the sampling time is n Vki1, Vki2, Vki3, ..., Vkii,
..., Vkin voltage signals are sampled. Regarding the culm, Vhj1, Vhj2, Vhj3, ‥‥, V
There are hji, . . . , Vhjn voltage signals.

FIG. 7 is a main flowchart for detecting the amount of grain per unit time as the No. 4 loss. Following start and initialization, in step 701, m pieces of experimental data of only grains are read. , each experimental data K
Sampling value Vki every T/n time interval for i
is stored (step 702). Similarly step 703
Then, in step 704, m pieces of experimental data concerning only the culm are read, and sampling values Vhj for each T/n time interval are stored for each experimental data Hj. Next, in step 705, the sampled values of the experimental data of only the grains are input and trained by the neural network. Similarly, in step 706, the sampling values of the experimental data of only the culms are input and the neural network is trained. Next, in step 707, the learned offset values, coupling coefficients, etc. of the neural network are stored in the RAM. Detection signals from the piezoelectric sensor 57 during actual threshing processing are input to these learned neural networks, and the frequency of appearance of grain patterns is measured (step 708). Step 70
In step 9, the amount of grains per unit time as No. 4 loss is calculated from the ratio between the area of the detection surface of the piezoelectric sensor 57 and the actual leakage area.

Next, learning using a backpropagation neural network will be explained. A subroutine flowchart of the neural network is shown in FIG. The backpropagation learning method in this embodiment is a supervised learning method, and is executed on a neural network having a hierarchical three-layer structure consisting of input layer I→intermediate layer→output layer.

The number of input layer units P in the neural network structure in this example is n=1 for experimental data only for grains and experimental data for only culms, respectively.
00 units, the number of intermediate layer units Q is 25, and the number of output layer units S is 2. FIG. 9 illustrates the structure of the neural network. The learning by the neural network is performed on m pieces of experimental data for only grains and only for culms, and sufficiently learns modified examples of patterns for grains and modified examples of patterns for culms. shall be taken as a thing.

A specific learning method is executed as follows. First, n sample values Vki1, Vki2, Vki3, . . . , Vki in the i-th experimental data of only grains.
Regarding i,..., Vkin (note that among the symbols in the text below, those enclosed in { } refer to multiple units) [1] Coupling coefficients {WPQ }, {WQS } and offset value {θP } that determine the state of the neural network
, {θQ } are each initialized with small random values. [2] Sample values Vki1, Vki2, Vki3, ...
, Vkii, . . . , Vkin are the initial learning data. [3] The value of the learning data is set as the output of the input layer unit P {Po
ut } and the coupling coefficient from the input layer to the hidden layer {WP
Q } and the offset value θP of the intermediate layer unit Q, the input value Qin to the intermediate layer unit is determined. This input value Qin and f(X)=1/(1+exp(-X)
), the output value Qout of the intermediate layer unit Q is determined by a sigmoid function f(X). Note that the output Pout of the input layer unit P is Pout = 1/(1+exp (-(X-θ)/
T) ...Formula (5) Here, X=W1*Vki1
+W2*Vki2+‥‥+Wi*Vkii+‥‥+Wn
*VkinW1, W2, . . . Wi, . . . , Wn are coupling coefficients and are weighted. θ is an offset value that changes due to learning, and T is
The temperature parameter of the energy function by the Boltzmann machine is set to T=1 in the example. [4] The output value {Qout} of the intermediate layer unit, the coupling coefficient {WQS} from the intermediate layer to the output layer S, and the output layer Q
using the offset value θQ of output layer unit S1
, S2 are determined. The output Qout of the intermediate layer unit Q is also expressed as Pou in equation (5) above.
It is derived from a similar calculation formula with t as X. [5] The teacher signal in learning only grains is set to S1out = 1 for the output layer unit S1, and set to S2out = 0 for the output layer unit S2. For each output layer unit, an error δs1 for the coupling coefficient connected to the output layer unit S1 and the offset of the intermediate layer unit Q is determined from the difference between the teacher signal and the output value of the output layer. The error δs2 is similarly determined for the output layer unit S2. [6] Error δs1 and coupling coefficient from the intermediate layer to the output layer {W
QS } and the output value Qout of the intermediate layer, an error σq with respect to the coupling coefficient connected to the intermediate layer unit Q and the offset value θP of the intermediate layer unit is determined. [7] Error σs in output layer unit S1 obtained in [5]
By adding the product of 1, the output Qout of the intermediate layer unit Q, and a constant α, the coupling coefficient WQS connecting from the intermediate layer unit Q to the output layer unit S1 is corrected. Also,
The offset value θQ of the output layer unit S1 is corrected by adding the product of the error σs1 and the constant β. Similarly, the output layer unit S2 is also determined. [8] Correct the coupling coefficient WPQ connecting the input layer unit P to the intermediate layer unit Q by adding the product of the error σq in the intermediate layer unit Q, the output value Pout of the input layer unit P, and the constant α . Furthermore, the offset value θP of the intermediate layer unit Q is corrected by adding the product of the error σq and the constant β.

[9] Input the next learning data. [10] Return to [3] until the learning data is completed. [11] Update the number of learning repetitions. [12] If the number of learning repetitions is less than the limit number, [2
Return to ].

In the final input layer unit P, the output value Pout is compared with the teacher signal, and the error σs1 and the error σq are back-propagated to correct the weights and perform learning. Learning ends when the difference between the output value Pout of the input layer unit and the output Sout of the output layer unit becomes less than or equal to the allowable error. The same learning as above is also performed for experimental data of only culms. In this case, the teacher signal for the culm in the flow [5] inverts the teacher signal for the grain. In other words, S1out =0 for the output layer unit S1, and S2out =1 for the output layer unit S2. As mentioned above, since it is possible to reliably provide a teacher signal when learning the experimental data on the grain and when learning the experimental data on the culm, the coupling coefficients and the like that ensure the convergence of the learning results are It is possible to reliably determine the offset value in advance.

[0027]

Effects of the Invention As described above, according to the present invention, the pattern when the target grain collides with the piezoelectric sensor,
The difference from the pattern when the culm collides with the piezoelectric sensor can be learned in advance using a neural network. Therefore, even if the target object to be detected (grain) is mixed with a foreign object (culm) and detected by the piezoelectric sensor at the same time, the detection pattern is the learning pattern of the grain that has completed the learning. If the object is similar to the above, it can be determined that it is a grain, so that in actual detection, only the object to be detected (grain) can be easily selected and detected. As a result, it is possible to detect the amount of grain leakage as No. 4 loss.

[Brief explanation of the drawing]

FIG. 1 is a side view of a combine harvester.

FIG. 2 is a plan view of the combine harvester.

FIG. 3 is a sectional view of the threshing device.

4 is a block diagram of the arrangement of piezoelectric sensors and a control device shown in a cross section taken along the line IV-IV in FIG. 3; FIG.

FIG. 5 is a diagram showing a detection pattern of only grains by a piezoelectric sensor.

FIG. 6 is a diagram showing a detection pattern by a piezoelectric sensor for only the culm.

FIG. 7 is a main flowchart of the present invention.

FIG. 8 is a flowchart of a backpropagation method.

FIG. 9 is a diagram showing the structure of a neural network.

[Explanation of symbols]

4 Threshing department 9　　　　Pre-reaping treatment section 13 Feeder House 18 Supply chain conveyor 19 Handling cylinder 20　　　　Handling room 24 Dust exhaust port 25 Sorting mechanism 28 Swing plate 30 1st Chaff Sheave 32 Reception net 34 Second chaff sieve 36 First-class grain board 37 Second flow grain board 42 Culm processing room 46 Second reduction tube 53 Ventilation duct 55 Cutting blade 56 Fixed blade 57 Piezoelectric sensor 59 Central control unit

Claims (1)

[Claims] Claim 1: A receiving net is provided below the handling barrel for threshing the harvested grain culm, and a sorting mechanism is provided below the receiving net, while the waste culm is discharged outside the machine from the waste culm outlet behind the sorting mechanism. In the threshing device, a piezoelectric sensor is provided between the receiving net below the rear portion of the handling barrel and the sorting mechanism for detecting objects leaking from the receiving net,
A neural network is constructed using the waveform detected by the piezoelectric sensor of only grains and the waveform detected only of the culm as learning data, and the amount of grain leakage is detected from the waveform detected of the object leaking from the receiving net. 1. A grain amount detection device in a threshing device, characterized in that a control means for controlling the grain amount is provided.