CN113177629A - Method and system for diagnosing blockage fault in operation of combined harvester - Google Patents
Method and system for diagnosing blockage fault in operation of combined harvester Download PDFInfo
- Publication number
- CN113177629A CN113177629A CN202110321753.2A CN202110321753A CN113177629A CN 113177629 A CN113177629 A CN 113177629A CN 202110321753 A CN202110321753 A CN 202110321753A CN 113177629 A CN113177629 A CN 113177629A
- Authority
- CN
- China
- Prior art keywords
- fault
- neural network
- data
- combine harvester
- lstm neural
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 27
- 238000013528 artificial neural network Methods 0.000 claims abstract description 52
- 238000003745 diagnosis Methods 0.000 claims abstract description 50
- 238000009825 accumulation Methods 0.000 claims abstract description 6
- 238000012549 training Methods 0.000 claims description 28
- 239000011159 matrix material Substances 0.000 claims description 12
- 238000007781 pre-processing Methods 0.000 claims description 11
- 230000000903 blocking effect Effects 0.000 claims description 10
- 239000012535 impurity Substances 0.000 claims description 8
- 238000012795 verification Methods 0.000 claims description 8
- 238000009826 distribution Methods 0.000 claims description 7
- 238000012360 testing method Methods 0.000 claims description 7
- 238000004364 calculation method Methods 0.000 claims description 5
- 238000004891 communication Methods 0.000 claims description 5
- 238000004458 analytical method Methods 0.000 claims description 4
- 238000003860 storage Methods 0.000 claims description 4
- 230000009467 reduction Effects 0.000 claims description 3
- 238000003306 harvesting Methods 0.000 abstract description 4
- 206010063385 Intellectualisation Diseases 0.000 abstract description 3
- 230000006870 function Effects 0.000 description 10
- 238000012986 modification Methods 0.000 description 4
- 230000004048 modification Effects 0.000 description 4
- 241001124569 Lycaenidae Species 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 238000011161 development Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000007726 management method Methods 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 230000002093 peripheral effect Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000013523 data management Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000012217 deletion Methods 0.000 description 1
- 230000037430 deletion Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000002035 prolonged effect Effects 0.000 description 1
- 239000004576 sand Substances 0.000 description 1
- 238000010187 selection method Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/047—Probabilistic or stochastic networks
-
- A—HUMAN NECESSITIES
- A01—AGRICULTURE; FORESTRY; ANIMAL HUSBANDRY; HUNTING; TRAPPING; FISHING
- A01D—HARVESTING; MOWING
- A01D41/00—Combines, i.e. harvesters or mowers combined with threshing devices
- A01D41/12—Details of combines
- A01D41/127—Control or measuring arrangements specially adapted for combines
- A01D41/1278—Control or measuring arrangements specially adapted for combines for automatic steering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/048—Activation functions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Health & Medical Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Probability & Statistics with Applications (AREA)
- Environmental Sciences (AREA)
- Testing And Monitoring For Control Systems (AREA)
- Combines (AREA)
Abstract
The invention provides a method and a system for diagnosing a blockage fault in the operation of a combine harvester, and belongs to the technical field of agricultural machinery intellectualization. The network server trains the LSTM neural network by utilizing the accumulated original data of the work information database of the combine harvester to obtain a fault diagnosis model, when the blockage fault diagnosis system runs, the data collected by the sensor is continuously accumulated, the LSTM neural network parameters before accumulation are further trained and optimized to obtain an updated fault diagnosis model, and the real-time data of the sensor is input into the updated fault diagnosis model after being preprocessed to obtain a system fault inference result. The method can continuously improve the capacity of deducing the blockage fault of the combined harvester, and further improve the intelligent degree and the harvesting operation efficiency of the combined harvester.
Description
Technical Field
The invention relates to the technical field of agricultural machinery intellectualization, in particular to a method and a system for diagnosing a blockage fault in the operation of a combine harvester based on an LSTM neural network.
Background
China is a big agricultural country, develops fine agriculture, and the realization of agricultural intellectualization is a necessary way for agricultural development. China manufacturing 2025 has agricultural machinery as one of ten major fields, and has proposed to vigorously develop advanced equipment such as harvesters and the like, and has demanded to improve data acquisition, high-end analysis and accurate operation capabilities of agricultural machinery. The combine harvester has the disadvantages of complex structure, severe working environment, large fluctuation of working load and easy occurrence of blockage faults. A set of fault diagnosis system with higher intelligent degree is researched and developed, so that the working efficiency and quality of the harvester can be improved, and the service life of the combined harvester can be prolonged.
The mode of manually extracting the rotating speed characteristics is mostly adopted for diagnosing the operation faults of the combine harvester at home and abroad, and the fault characteristics are extracted from the operation information data of the combine harvester by partially combining machine learning and deep learning technologies. For example, Craessaerts, gerert et al, belgium, proposed the application of self-organizing maps (SOM) and multiple feed-forward back-propagation network methods to the fault identification of combine harvesters. Compared with the traditional method, the method reduces the addition of excessive human experience, and can establish a more direct connection between the working state and the fault of the combine harvester to a certain extent, but the diagnosis of the blockage fault of the combine harvester by utilizing the working information time sequence data of the combine harvester is less.
Disclosure of Invention
Aiming at the defects in the prior art, the invention provides the method and the system for diagnosing the blockage fault of the combined harvester based on the LSTM neural network, which can monitor the working state of the combined harvester in real time, can diagnose the blockage fault of the combined harvester during working in time by utilizing the LSTM network model, improve the intelligent degree of the combined harvester, reduce the occurrence probability of the blockage fault of the combined harvester and improve the harvesting working efficiency.
The present invention achieves the above-described object by the following technical means.
A method for diagnosing the jam fault in the operation of combine harvester features that the original data accumulated in the database of operation information of combine harvester is used to train LSTM neural network to obtain a fault diagnosing model, the data collected by sensor is accumulated continuously while the jam fault diagnosing system is running, the parameters of LSTM neural network before accumulation are further trained and optimized to obtain an updated fault diagnosing model, and the real-time data of sensor is preprocessed and then input to the updated fault diagnosing model to obtain the system fault deducing result.
Further, the training of the LSTM neural network by using the accumulated raw data of the work information database of the combine harvester to obtain a fault diagnosis model specifically comprises:
step (1), adding a sliding time window to original data in a job information database so as to obtain a time sequence
Step (2), finishing the standardized preprocessing of the sensor acquisition data and the data label making
Selecting time sequence data with proper sliding time window length according to the step (1), and making a data set of a neural network by using One-hot codes of the failure types of the combine harvester as tags of the time sequence
Step (3), dividing the data set into a training set, a verification set and a test set
Step (4), constructing an LSTM neural network structure
The input layer dimension of the LSTM neural network is WLX m, wherein WLThe length of the sliding time window is m, and the number of the sensors is m;
the LSTM neural network defines two LSTM network layers, and the LSTM network layers have the capacity of memorizing and deleting time series information.
Further, the LSTM network layer is implemented by a forgetting gate, an input gate, and an output gate, and specifically calculated as follows:
ft＝σ(Wf·[ht-1，xt]+bf)
it＝σ(Wi·[ht-1，xt]+bi)
Ot＝σ(W0·[ht-1，xt]+bo)
ht＝Ot·tanh(Ct)
wherein: the function σ () is a sigmoid function, Wf、Wi、WC、W0As a weight matrix, bf、bi、bc、boTo be offset, ht-1Output for the last moment, htFor the current time output, xtFor input at the present moment, Ct-1Is the cell state at the previous time, CtIs the cell state at the current time, ftTo forget the output value of the gate, itTo input the output value of the gate,
as a new candidate value vector, OtIs the output value of the output gate.
Further, the cross entropy is calculated by the following formula:
where n is the number of fault types, yiIs a true distribution, aiTo predict the distribution.
Further, the method also comprises the following steps:
step (5), calculating the cross entropy of a certain time parameter in the LSTM neural network training process
Step (6), judging whether the termination condition is reached, if not, updating the network parameters by using an RMSprop algorithm
And (5) repeating the step (5) and the step (6), realizing gradient reduction of the LSTM neural network parameters until the cross entropy obtained by calculation in the training set meets a preset target, verifying the correctness of the LSTM neural network structure and the parameters by using the verification set, obtaining the actual fault inference capability by using the test set after the LSTM neural network parameters are determined, and finishing the training of the LSTM neural network to obtain a fault diagnosis model if the accuracy of the fault diagnosis result is close to the actual fault rate of the data set.
A blockage fault diagnosis system in the operation of a combine harvester comprises a sensor, an airborne embedded system and a network server, wherein the sensor and the sensor are both communicated with the airborne embedded system, and the sensor comprises a multi-path rotating speed sensor, a grain impurity-containing breakage rate sensor, a grain entrainment loss rate sensor and a Beidou positioning module;
the network server completes storage and analysis of the operation data of the combine harvester, constructs an operation information database of the combine harvester, and completes training of an LSTM neural network through the original data of the database, thereby constructing a blocking fault diagnosis model; and (4) preprocessing the real-time data acquired by the sensor after operation, inputting the trained blocking fault diagnosis model, and deducing the blocking fault.
In the above technical solution, when the onboard embedded system performs fault diagnosis, the preprocessing operation includes forming a time sequence matrix from a plurality of time point data acquired by a plurality of sensors by using a sliding window, and normalizing the time sequence matrix.
In the technical scheme, the vehicle-mounted embedded system further comprises a touch display in signal connection with the vehicle-mounted embedded system.
The invention has the beneficial effects that:
(1) the airborne embedded system can acquire the operation information of the combine harvester in real time, including the rotating speed of the multi-path rotating part, the impurity content of grains, the crushing rate, the grain entrainment loss rate, the advancing speed and the like. And based on the data acquired in real time, the airborne embedded system infers the fault type of the combine harvester by using the updated fault diagnosis model, and if the combine harvester breaks down, the airborne embedded system prompts an operator on a touch display screen and displays the inferred fault type.
(2) The work blockage fault diagnosis system of the combine harvester continuously optimizes the LSTM neural network structure and parameters by collecting and accumulating work information data of the combine harvester, and continuously improves the capacity of deducing the blockage fault of the combine harvester, thereby improving the intelligent degree and the harvesting work efficiency of the combine harvester.
(3) According to the method, the LSTM neural network is adopted, time dimension information in the operation parameters of the combine harvester is utilized, the fault characteristics are identified in an end-to-end mode, compared with the traditional manual characteristic extraction, redundant data processing operation is reduced, and the effect of a manual characteristic selection method on algorithm fault inference is reduced.
(4) The method and the system for diagnosing the blockage fault of the combine harvester can effectively monitor the operation parameters of the combine harvester, diagnose various operation blockage fault types, and continuously optimize and update the algorithm model. The system has the advantages of perfect scheme, high feasibility and convenience for further development, and the used diagnosis model has the characteristics of small volume, high speed and strong mobility and has the capability of continuously optimizing the model by using big data.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art are briefly introduced below; it is obvious that the drawings in the following description are some embodiments of the invention, and that for a person skilled in the art, other drawings can be derived from them without inventive effort.
FIG. 1 is a block diagram of a jam fault diagnostic system for operation of a combine harvester according to the present invention;
FIG. 2 is a flow chart of the programming of the onboard embedded system of the present invention;
FIG. 3 is a flow chart of the LSTM neural network-based fault diagnosis training of the present invention;
fig. 4 is a flow chart of the jam fault diagnosis in the operation of the combine harvester of the invention.
Detailed Description
In order to make the purpose, technical effect and technical solution of the embodiments of the present invention clearer, the present invention is described in detail and completely with reference to the accompanying drawings and the detailed description; it is to be understood that the described embodiments are only some of the embodiments of the present invention. Other embodiments, which can be derived by one of ordinary skill in the art from the disclosed embodiments without inventive faculty, are intended to be within the scope of the invention.
As shown in fig. 1, a schematic block diagram of a jam fault diagnosis system in combine harvester operation includes a sensor, an onboard embedded system and a network server; the sensors comprise a multi-path rotating speed sensor, a grain impurity-containing breakage rate sensor, a grain entrainment loss rate sensor and a Beidou positioning module; the rotating speed sensor is a Hall sensor, is directly connected with an interrupt pin of the airborne embedded system, and completes rotating speed measurement in a pulse interrupt counting mode; the kernel impurity-containing breakage rate sensor and the kernel entrainment loss rate sensor are communicated with the airborne embedded system through the CAN bus, so that kernel breakage, impurity-containing and loss rate monitoring data are collected; the Beidou positioning module is communicated with the airborne embedded system through a USART serial port, and analyzes GPS protocol data to obtain geographic position information of the combine harvester.
In fig. 1, an onboard embedded system collects work information of a combine harvester, is connected with a network server through a mobile network (3G/4G network), and uploads real-time work information to the network server. The network server completes the storage and analysis of the operation data of the combine harvester, constructs a perfect operation information database of the combine harvester, completes the training of the LSTM neural network through the database, and the trained LSTM neural network is used as a blocking fault diagnosis model. After the real-time data collected by the sensor is preprocessed by the airborne embedded system, the processed data is input into a fault diagnosis model which is trained and optimized, an inference result of a blocking fault is obtained after the processed data is output by the fault diagnosis model, and the diagnosis result and operation information are displayed on an airborne touch display. In addition, the network server provides a work data management interface of the combine harvester through internet service, and data of the work information database is presented to a user in a visual and user-friendly mode.
As shown in fig. 2, a flow chart of programming of an onboard embedded system of a combine harvester is provided, which mainly includes peripheral initialization, data acquisition, fault diagnosis, network communication and GUI display functions. The peripheral initialization is to perform initialization setting of the sensor and the airborne touch display, establish network communication between the airborne embedded system and the network server, start each working thread, and send the working thread to an operating system (located inside the airborne embedded system) to complete thread scheduling. The data acquisition thread acquires the operation data of the combine harvester through external interruption, serial port transceiving, CAN bus and other modes, analyzes and processes the data into an actual operation data format, and indicates that the acquisition of one-time operation data is completed after an acquisition completion signal is sent. The graphical interface (GUI) thread is responsible for completing the realization of a human-computer interaction interface, responding to the operations of user parameter input modification and the like, displaying operation data on a screen and drawing a chart, and updating the screen to display the operation data if a display updating signal (a data acquisition thread) is received. In the fault diagnosis thread, preprocessing operation is carried out on real-time data of the sensors (including that a time sequence matrix is formed by a plurality of time point data collected by a plurality of sensors by using a sliding window and the time sequence matrix is normalized), then the data after the preprocessing operation is used as the input of a fault diagnosis model, the output result of the diagnosis model is converted into the corresponding inference result of the blocking fault, and if the blocking fault is diagnosed, an operation fault alarm signal is sent. In the network communication thread, the formatting of a data packet is mainly completed, and an http protocol interface is used for sending the packaged data to a network server and receiving the data from the network server.
The network server is developed by using techniques such as golang and mysql, supports cross-platform operation, and is in data communication with the airborne embedded system in http and websocket modes. The job information database uses maridb and provides HeidiSQL as a database graphical access tool. The network server provides a plurality of api interfaces based on an http protocol, data contents are json format character strings, and data reporting, operation information database query, equipment management (creation, modification and deletion) for monitoring the combine harvester, and user management (creation, login, password modification and logout) can be completed. The data of a plurality of combine harvesters are respectively stored in the combine harvester operation information database, and comprise reporting time, longitude and latitude data, analyzed geographical position data, multi-path rotating speed, advancing speed, grain breakage rate, impurity rate and loss rate.
As shown in fig. 3, a training flowchart of the fault diagnosis model based on the LSTM neural network is shown, which specifically includes the following steps:
step (1), adding a sliding time window to original data in a job information database so as to obtain a time sequence
Assuming that the length of the raw data of the plurality of sensors is SLThe length of the sliding window is WLStep length of sliding window is WSThe number of sensors is m, and the size of each sample is W after the original data is extracted through a sliding time windowLXm, number of time slices of size of divisible sliding time window WNComprises the following steps:
raw data by sliding window length WLCutting and stepping WSSliding the window can obtain the required time sequence.
Step (2), completing standardized preprocessing of data collected by the sensors and data label manufacturing, wherein the standardized preprocessing mainly comprises the steps of forming a time sequence matrix by using a sliding window for a plurality of time point data collected by various sensors, and normalizing the time sequence matrix to eliminate dimension; the normalized mathematical expression is:
in the formula, L is the number of samples,
in order to be the normalized sample set,
is the g-th sample of sample set XThis, xgFor the g sample before normalization, xmeanDenotes the mean value of X, XstdRepresents the standard deviation of X, and:
the label making uses One-hot coding mode, that is, N state registers are used to code N states, each state has its independent register bit, and only One bit is valid at any time. Coding the fault types of feeding auger blockage, conveying groove blockage, threshing roller blockage, secondary impurity blockage and grain conveying auger blockage of the combined harvester. And (2) selecting time sequence data with a proper sliding time window length according to the step (1), and making a data set of the neural network by using the One-hot code of the fault type as a label of the time sequence.
Step (3), dividing the data set into a training set, a verification set and a test set
In order to ensure the generalization of the training result, the data volume of each fault type is equivalent, and the three types of data sets are from the same distribution as much as possible; the proportion of the training set to the test set is usually determined according to the actual training data amount, and usually the proportion can be 7: 3, and the verification set is divided in the training set, and can be cross-verification or fixed proportion.
Step (4), constructing an LSTM neural network structure
The length W of the sliding time window selected according to the step (1)LStep length WsAnd the number m of sensors, determining the input layer dimension of the LSTM neural network as WLX m; the neural network defines two LSTM network layers, the dimension of a final output layer is n multiplied by 1, and a vector is in a One-hot coding format. Wherein, the LSTM network layer has the capability of memorizing and deleting time series information, and the time series information is mainly combined by three gates (a forgetting gate, an input gate and an output gate)The method is realized by the following specific calculation:
ft＝σ(Wf·[ht-1，xt]+bf)
it＝σ(Wi·[ht-1，xt]+bi)
Ot＝σ(W0·[ht-1，xt]+bo)
ht＝Ot。tanh(Ct)
wherein, the function of the sigma () is a sigmoid function, and the expression is
Wf、Wi、WC、W0Is a weight matrix; bf、bi、bc、boIs an offset; h ist-1Output for the last moment, htOutputting for the current moment; x is the number oftInputting for the current moment; ct-1Is the cell state at the previous time, CtThe cell state is the current time; f. oftThe output value of the forgetting gate is between 0 and 1; i.e. itIs the output value of the input gate, and the value is between 0 and 1;
calculating a new candidate value vector by using a tanh function; o istThe output value of the output gate is between 0 and 1.
Step (5), calculating the cross entropy of a certain time parameter in the LSTM neural network training process
The final output of the LSTM neural network uses a Softmax function, where the cross entropy is specifically calculated as follows:
where n is the number of fault types, yiIs a true distribution, aiFor predicting the distribution, i.e. the output result of the Softmax function, the Softmax function is specifically defined as follows:
wherein z isiIs the output value of the ith node, zjIs the output value of the jth node.
And (6) judging whether a termination condition is reached, if not, updating the network parameters by using an RMSprop algorithm: the algorithm uses a differential squared weighted average for the gradient of weights iii and offsets b, assuming that during the t-th iteration, the network parameters are calculated as follows:
sdw＝βsdw+(1-β)dW2
sdb＝βsdb+(1-β)db2
wherein s isdwAnd sdbRespectively, the gradient momentum accumulated by the loss function in the previous t-1 iteration, beta is an index of the gradient accumulation, d is the network learning rate, and epsilon is an empirical value, which can be generally selected to be 10-8。
And finally, repeating the step (5) and the step (6) to realize gradient reduction of the LSTM neural network parameters until the cross entropy obtained by calculation in the training set meets the preset target. Meanwhile, the correctness of the LSTM neural network structure and parameters is verified by using a verification set, after the LSTM neural network parameters (weight matrix and bias) are determined, the actual fault inference capability is obtained by using a test set, and if the accuracy rate of the fault diagnosis result is close to the actual fault rate of the data set, the training of the LSTM neural network is completed, and a fault diagnosis model is obtained.
As shown in fig. 4, the fault diagnosis model is obtained by training an LSTM neural network with the raw data accumulated in the work information database of the combine harvester, and then the data collected by the sensor continues to be accumulated during the operation of the blockage fault diagnosis system; and under the condition of keeping the structure and parameters of the LSTM neural network before accumulation, further training and optimizing the parameters of the LSTM neural network before accumulation by using the newly added data, thereby improving the fault inference capability of the LSTM neural network. And the fault diagnosis model optimized and finished in each training is transmitted to the airborne embedded system through the network, and the fault diagnosis network model stored in the airborne embedded system of the combined harvester is updated and replaced. Meanwhile, the real-time data of the sensor is input into the updated fault diagnosis model after being preprocessed, and the system fault inference result can be obtained after the real-time data of the sensor is output through the fault diagnosis model.
The method can acquire the working state of the combined harvester in real time, can diagnose the blockage fault of the combined harvester in time by using the fault diagnosis model based on the LSTM neural network, reduces the probability of the blockage fault of the combined harvester, improves the intelligent degree of the harvester, and improves the harvesting working efficiency. The system has data acquisition and storage capacity, and continuously optimizes the fault diagnosis model based on the LSTM neural network by using the stored work data of the combine harvester, thereby further improving the accuracy of the blockage fault diagnosis.
The present invention is not limited to the above-described embodiments, and any obvious improvements, substitutions or modifications can be made by those skilled in the art without departing from the spirit of the present invention.
Claims (8)
1. A method for diagnosing blockage faults in the operation of a combined harvester is characterized by comprising the following steps:
training an LSTM neural network by utilizing the accumulated original data of the work information database of the combine harvester to obtain a fault diagnosis model, continuously accumulating the data acquired by the sensor when the blockage fault diagnosis system runs, further training and optimizing LSTM neural network parameters before accumulation to obtain an updated fault diagnosis model, preprocessing the real-time data of the sensor, and inputting the updated fault diagnosis model to obtain a system fault inference result.
2. The method for diagnosing the blockage fault in the operation of the combine harvester as recited in claim 1, wherein the method for training an LSTM neural network by using the accumulated raw data of the operation information database of the combine harvester to obtain a fault diagnosis model comprises the following steps:
step (1), adding a sliding time window to original data in a job information database so as to obtain a time sequence
Step (2), finishing the standardized preprocessing of the sensor acquisition data and the data label making
Selecting time sequence data with proper sliding time window length according to the step (1), and making a data set of a neural network by using One-hot codes of the failure types of the combine harvester as tags of the time sequence
Step (3), dividing the data set into a training set, a verification set and a test set
Step (4), constructing an LSTM neural network structure
The input layer dimension of the LSTM neural network is WLX m, wherein WLThe length of the sliding time window is m, and the number of the sensors is m;
the LSTM neural network defines two LSTM network layers, and the LSTM network layers have the capacity of memorizing and deleting time series information.
3. The method for diagnosing the blockage fault in the operation of the combine harvester according to claim 2, wherein the LSTM network layer is realized by a forgetting gate, an input gate and an output gate, and the specific calculation is as follows:
ft＝σ(Wf·[ht-1，xt]+bf)
it＝σ(Wi·[ht-1，xt]+bi)
Ot＝σ(W0·[ht-1，xt]+bo)
ht＝Ot·tanh(Ct)
wherein: the function σ () is a sigmoid function, Wf、Wi、WC、W0As a weight matrix, bf、bi、bc、boTo be offset, ht-1Output for the last moment, htFor the current time output, xtFor input at the present moment, Ct-1Is the cell state at the previous time, CtIs the cell state at the current time, ftTo forget the output value of the gate, itTo input the output value of the gate,
as a new candidate value vector, OtIs the output value of the output gate.
4. The method for diagnosing a jam fault in the operation of a combine harvester according to claim 2, characterized in that the cross entropy is calculated by the following formula:
where n is the number of fault types, yiIs a true distribution, aiTo predict the distribution.
5. The method for diagnosing a jam fault in the operation of a combine harvester according to claim 2, further comprising:
step (5), calculating the cross entropy of a certain time parameter in the LSTM neural network training process
Step (6), judging whether the termination condition is reached, if not, updating the network parameters by using an RMSprop algorithm
And (5) repeating the step (5) and the step (6), realizing gradient reduction of the LSTM neural network parameters until the cross entropy obtained by calculation in the training set meets a preset target, verifying the correctness of the LSTM neural network structure and the parameters by using the verification set, obtaining the actual fault inference capability by using the test set after the LSTM neural network parameters are determined, and finishing the training of the LSTM neural network to obtain a fault diagnosis model if the accuracy of the fault diagnosis result is close to the actual fault rate of the data set.
6. The diagnosis system for realizing the method for diagnosing the blockage faults in the operation of the combine harvester as claimed in any one of claims 1 to 5 is characterized by comprising a sensor, an onboard embedded system and a network server, wherein the sensor and the sensor are communicated with the onboard embedded system, and the sensor comprises a multi-path rotating speed sensor, a grain impurity-containing breakage rate sensor, a grain entrainment loss rate sensor and a Beidou positioning module;
the network server completes storage and analysis of the operation data of the combine harvester, constructs an operation information database of the combine harvester, and completes training of an LSTM neural network through the original data of the database, thereby constructing a blocking fault diagnosis model; and (4) preprocessing the real-time data acquired by the sensor after operation, inputting the trained blocking fault diagnosis model, and deducing the blocking fault.
7. The diagnostic system of claim 6, wherein the preprocessing operation comprises using a sliding window to construct a time-series matrix from the plurality of time points collected by the plurality of sensors and normalizing the time-series matrix when the onboard embedded system performs fault diagnosis.
8. The diagnostic system of claim 6, further comprising a touch display in signal communication with the onboard embedded system.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110321753.2A CN113177629A (en) | 2021-03-25 | 2021-03-25 | Method and system for diagnosing blockage fault in operation of combined harvester |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110321753.2A CN113177629A (en) | 2021-03-25 | 2021-03-25 | Method and system for diagnosing blockage fault in operation of combined harvester |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113177629A true CN113177629A (en) | 2021-07-27 |
Family
ID=76922331
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110321753.2A Pending CN113177629A (en) | 2021-03-25 | 2021-03-25 | Method and system for diagnosing blockage fault in operation of combined harvester |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113177629A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114295377A (en) * | 2021-12-13 | 2022-04-08 | 南京工业大学 | CNN-LSTM bearing fault diagnosis method based on genetic algorithm |
| CN115119608A (en) * | 2022-06-30 | 2022-09-30 | 扬州市金谷机械有限公司 | Intelligent control method for integrated harvesting of corn ears and corn stalks |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20200022948A (en) * | 2018-08-24 | 2020-03-04 | 포항공과대학교 산학협력단 | Method for diagnosing and predicting robot arm's failure |
| CN111259717A (en) * | 2019-10-21 | 2020-06-09 | 杭州安脉盛智能技术有限公司 | Method and system for judging abnormal state of rotating equipment |
| CN112001418A (en) * | 2020-07-14 | 2020-11-27 | 华北电力大学（保定） | Method for diagnosing blockage fault of cooling water channel of stator coil of generator |
-
2021
- 2021-03-25 CN CN202110321753.2A patent/CN113177629A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20200022948A (en) * | 2018-08-24 | 2020-03-04 | 포항공과대학교 산학협력단 | Method for diagnosing and predicting robot arm's failure |
| CN111259717A (en) * | 2019-10-21 | 2020-06-09 | 杭州安脉盛智能技术有限公司 | Method and system for judging abnormal state of rotating equipment |
| CN112001418A (en) * | 2020-07-14 | 2020-11-27 | 华北电力大学（保定） | Method for diagnosing blockage fault of cooling water channel of stator coil of generator |
Non-Patent Citations (2)
| Title |
|---|
| 刘景: "基于深度学习的联合收割机故障预警技术研究", 《全国硕士论文》, 15 March 2021 (2021-03-15), pages 25 * |
| 王维锋等: "基于双层长短时记忆网络的齿轮故障诊断方法", 《装甲兵工程学院学报》, vol. 32, no. 2, 30 April 2018 (2018-04-30), pages 1 - 2 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114295377A (en) * | 2021-12-13 | 2022-04-08 | 南京工业大学 | CNN-LSTM bearing fault diagnosis method based on genetic algorithm |
| CN115119608A (en) * | 2022-06-30 | 2022-09-30 | 扬州市金谷机械有限公司 | Intelligent control method for integrated harvesting of corn ears and corn stalks |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109272123B (en) | Sucker-rod pump working condition early warning method based on convolution-circulation neural network | |
| CN111027487B (en) | Behavior recognition system, method, medium and equipment based on multi-convolution kernel residual error network | |
| CN112887145B (en) | Distributed network slice fault detection method | |
| CN111079977B (en) | Heterogeneous federal learning mine electromagnetic radiation trend tracking method based on SVD algorithm | |
| CN108126987B (en) | Remote online monitoring and intelligent diagnosis method for high-speed wire rolling mill | |
| CN113177629A (en) | Method and system for diagnosing blockage fault in operation of combined harvester | |
| CN114118224B (en) | Neural network-based full-system telemetry parameter anomaly detection system | |
| CN111339092B (en) | Multi-scale air quality forecasting method based on deep learning | |
| WO2022174679A1 (en) | Method and apparatus for predicting voltage inconsistency fault of battery cells, and server | |
| KR20180054413A (en) | Methd for optimizing crop production efficency and apparatus for the same | |
| CN108875118B (en) | Method and device for evaluating accuracy of prediction model of silicon content of blast furnace molten iron | |
| CN110243595B (en) | Long-range gear box fault monitoring system based on LabVIEW | |
| CN116631186B (en) | Expressway traffic accident risk assessment method and system based on dangerous driving event data | |
| CN115220133B (en) | Rainfall prediction method, device and equipment for multiple meteorological elements and storage medium | |
| CN108090515A (en) | A kind of environmental rating appraisal procedure based on data fusion | |
| CN114444582A (en) | Mechanical equipment fault diagnosis method based on convolutional neural network and Bayesian network | |
| CN114002597A (en) | Motor fault diagnosis method and system based on GRU network stator current analysis | |
| CN116645232A (en) | Intelligent management system for agricultural cultivation | |
| CN114152442B (en) | Rolling bearing cross-working condition fault detection method based on migration convolutional neural network | |
| CN117909860A (en) | Cotton picker and device and method for diagnosing and early warning blockage faults of cotton picking head of cotton picker | |
| CN115659249B (en) | Abnormality detection method for intelligent station measurement control system | |
| CN112907971A (en) | Urban road network short-term traffic flow prediction method based on genetic algorithm optimization space-time residual error model | |
| CN116028849B (en) | Emulsion pump fault diagnosis method based on depth self-coding network | |
| CN113537459A (en) | Method for predicting humiture of drug storage room | |
| CN113469228A (en) | Power load abnormal value identification method based on data flow space-time characteristics |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |