Agricultural machinery operation status monitoring method and system based on Internet of Things

Abstract

The present invention discloses an agricultural machinery operation status monitoring method and system based on the Internet of Things, and the method includes: data collection, data preprocessing, building an agricultural machinery operation status monitoring model, agricultural machinery operation status monitoring model parameter search and real-time operation status monitoring. The present invention belongs to the field of fault monitoring technology, and specifically refers to an agricultural machinery operation status monitoring method and system based on the Internet of Things. The scheme extracts the features of source domain and target domain data and completes data reconstruction, calculates the difference in source domain and target domain distribution by reproducing the Hilbert space of the kernel, aligns the data distribution of the two domains by minimizing the difference, and designs a total loss function; initializes individual positions based on improved tent mapping, improves convergence factors based on Gaussian distribution change curves, calculates optimal individual positions based on proportional weighting strategies, and updates all individual positions in combination with the improved convergence factors and optimal individual positions, thereby improving the monitoring accuracy and efficiency of the agricultural machinery operation status.

Description

Technical Field

[0001] The present invention belongs to the technical field of fault monitoring, and specifically refers to an agricultural machinery operation status monitoring method and system based on the Internet of Things. Background Art

[0002] The method of monitoring the operation status of agricultural machinery is to comprehensively use the Internet and artificial intelligence technology to realize real-time monitoring of the operation status of agricultural machinery and improve the efficiency and safety of agricultural machinery. However, the existing monitoring methods for the operation status of agricultural machinery have problems such as difficulty in accurately extracting the operation characteristics of agricultural machinery, differences in the distribution of agricultural machinery operation data between different domains, and neglect of the correlation between data categories during classification, resulting in low classification accuracy; the existing monitoring methods for the operation status of agricultural machinery have poor adaptability to different types of agricultural machinery and work, poor monitoring performance in new environments and new data, and weak generalization ability. Summary of the invention

[0003] In view of the above situation, in order to overcome the defects of the prior art, the present invention provides an agricultural machinery operation status monitoring method and system based on the Internet of Things. In view of the problems in the existing agricultural machinery operation status monitoring methods that the agricultural machinery operation characteristics are difficult to accurately extract, the agricultural machinery operation data distribution between different domains is different, and the correlation between data categories is ignored during classification, resulting in low classification accuracy, this scheme designs an encoder network structure to extract the characteristics of the source domain and target domain data in a targeted manner, maps high-dimensional data to a low-dimensional space, effectively reduces the dimension of the data, and retains valid information; by designing a symmetric structure of the decoder, data reconstruction is achieved to maintain the integrity and accuracy of the data; by reproducing the Hilbert space of the kernel, the difference in the distribution of the source domain and the target domain is calculated, and the data distribution of the two domains is aligned by minimizing the difference loss, thereby improving the model's adaptability to different data domains; a smoothing label algorithm is introduced to optimize the encoding of data category labels; a total loss is designed based on reconstruction loss, difference loss and classification loss The loss function helps to understand and analyze the agricultural machinery operation status data more accurately, and improve the monitoring accuracy and efficiency of the agricultural machinery operation status; the existing agricultural machinery operation status monitoring methods have poor adaptability to different agricultural machinery types and tasks, poor monitoring performance under new environments and new data, and weak generalization ability. This scheme improves the tent mapping method by introducing random variables to generate the initial position of the individual, which helps to increase the diversity of the search; the convergence factor based on the Gaussian distribution change curve is adopted, which is conducive to balancing the ability of local search and global search in the search process; by designing a proportional weighted strategy based on fitness and the top three best individual positions, the optimal individual position is determined more effectively, which is conducive to improving the convergence and stability of the optimization algorithm; the combination of the improved convergence factor based on the Gaussian distribution change curve and the optimal individual position helps to improve the efficiency and accuracy of the search, improve the comprehensiveness and accuracy of monitoring the operation status of various agricultural machinery and work, and better guide the use, maintenance and management of agricultural machinery.

[0004] The technical solution adopted by the present invention is as follows: The method for monitoring the operation status of agricultural machinery based on the Internet of Things provided by the present invention comprises the following steps:

[0005] Step S1: data collection;

[0006] Step S2: data preprocessing;

[0007] Step S3: constructing an agricultural machinery operation status monitoring model;

[0008] Step S4: searching for parameters of the agricultural machinery operation status monitoring model;

[0009] Step S5: Real-time operation status monitoring.

[0010] Furthermore, in step S1, the data collection is to deploy the Internet of Things device on the agricultural machinery, and collect the historical operation data and real-time operation data of the agricultural machinery based on the Internet of Things device; the historical operation data and the real-time operation data both include agricultural machinery equipment information, environmental parameter data, operation trajectory data, sensor data and maintenance records, and the historical operation data also includes the operation status of the agricultural machinery; the operation status of the agricultural machinery includes normal and agricultural machinery fault types; the operation status is used as a data category label, and the collected data is transmitted to the central processing center.

[0011] Furthermore, in step S2, the data preprocessing is to perform data cleaning, data conversion, data normalization and data set construction on the collected data through the central processing center; data cleaning includes processing noise, duplicate values, missing values ​​and outliers; data conversion is to convert the data after data cleaning into vector form; data normalization is to unify the data range based on the maximum and minimum normalization method; and data set construction is to construct a source domain training set with data category labels and a target domain training set without data category labels based on the historical operation data and real-time operation data after data cleaning, data conversion and data normalization.

[0012] Furthermore, in step S3, the construction of the agricultural machinery operation status monitoring model specifically includes the following steps:

[0013] Step S31: Extract features and design an encoder to extract features from input data; the encoder consists of 5 one-dimensional convolution blocks, each of which contains a convolution layer and a maximum pooling layer, activated by the RELU function; the size of the first convolution kernel is 1×64, the size of the remaining four convolution kernels is 1×3, and the step size of the maximum pooling layer is 1×2; the formula used to extract features is as follows:

[0014] ;

[0015] ;

[0016] In the formula, Q y and Q m are the source domain features and target domain features extracted from the source domain training set and the target domain training set, respectively. enc is the network structure of the encoder. y and X m They are the source domain training set and the target domain training set respectively;

[0017] Step S32: Data reconstruction, designing a decoder for data reconstruction; the decoder consists of 5 one-dimensional deconvolution blocks, each of which contains a deconvolution layer and an upsampling layer, and the structure of the decoder is completely symmetrical with that of the encoder; wherein the size of the first deconvolution kernel is 1×64, the size of the remaining four deconvolution kernels is 1×3, and the step size of the upsampling layer is 1×2; comprising the following steps:

[0018] Step S321: Calculate the reconstructed target domain data, using the following formula:

[0019] ;

[0020] In the formula, is the reconstructed target domain data, dec is the network structure of the decoder;

[0021] Step S322: Calculate the reconstruction loss of the target domain data. The formula used is as follows:

[0022] ;

[0023] Where, L rec is the reconstruction loss of the target domain data, N m is the number of target domain data in the target domain training set, and are the i-th target domain data before reconstruction and the i-th target domain data after reconstruction, i is the target domain data index;

[0024] Step S33: Domain alignment, mapping both the source domain data and the target domain data to the Hilbert space with the reproduction kernel, aligning the distributions of the two domains by minimizing the difference between the source domain distribution and the target domain distribution, including the following steps:

[0025] Step S331: Calculate the difference between the source domain distribution and the target domain distribution. The formula used is as follows:

[0026] ;

[0027] ;

[0028] ;

[0029] In the formula, is the difference between the source domain distribution and the target domain distribution, p （c） and q （c） are the source domain distribution and the target domain distribution, l is the feature layer index, C is the total number of data category labels, c is the data category label index, and N y is the number of source domain data in the source domain training set, and are the weights of the i-th and j-th source domain data belonging to the c-th data category label in the source domain training set, and are the weights of the i-th and j-th target domain data belonging to the c-th data category label in the target domain training set, and are the number of source domain data and target domain data in the source domain training set and the target domain training set that belong to the cth category as the i-th source domain data and target domain data, respectively. k(·) is the kernel function of the Hilbert space with a reproduction kernel. and are the activation values ​​of the i-th and j-th source domain data in the l-th feature layer, respectively. and are the activation values ​​of the i-th and j-th target domain data in the l-th feature layer respectively;

[0030] Step S332: Calculate the difference loss between the source domain distribution and the target domain distribution. The formula used is as follows:

[0031] ;

[0032] Where, L LMMD is the difference loss between the source domain distribution and the target domain distribution, l max is the number of feature layers;

[0033] Step S34: Classification, input the extracted target domain features into the softmax function to calculate the probability that the target domain data belongs to each data category label, and select the data category label with the maximum probability as the output label of the target domain data on the model, including the following steps:

[0034] Step S341: Introduce a smooth label algorithm to optimize the one-hot encoding of the real data category label. The formula used is as follows:

[0035] ;

[0036] In the formula, ε is the smoothing parameter, yes The real data labels, yes The one-hot encoding of the real data label, C is the total number of data category labels, and c is the data category label index;

[0037] Step S342: Classification loss, the formula used is as follows:

[0038] ;

[0039] Where, L cls is the classification loss, and They are In the target domain training set, it belongs to class c and The probability of the class label of the class data, W and b are the weight and bias respectively;

[0040] Step S35: Design the total loss function of the model, and combine the reconstruction loss, difference loss and classification loss according to the weights to obtain the total loss function of the model. The formula used is as follows:

[0041] ;

[0042] Where L is the total loss function, β1, β2 and β3 are the first weight, second weight and third weight respectively, and L LMMD It is the difference loss between the source domain distribution and the target domain distribution.

[0043] Furthermore, in step S4, the agricultural machinery operation status monitoring model parameter search specifically includes the following steps:

[0044] Step S41: Initial individual position, establish a parameter search space for the first weight β1, the second weight β2, the third weight β3, the weight W and the bias b in the agricultural machinery operation status monitoring model, introduce random variables to improve the tent mapping, initialize F individual positions in the parameter search space based on the improved tent mapping, use the individual position as a representative of the parameters of the agricultural machinery operation status monitoring model, and use the total loss function of the agricultural machinery operation status monitoring model trained based on the individual position as the individual fitness value. The formula used for the initial individual position is as follows:

[0045] ;

[0046] ;

[0047] In the formula, A w+1 and A w are the w+1th and wth improved chaotic sequences respectively, v is the control value, lb and ub are the lower and upper limits of the parameter search space respectively, D w is the initial position of the wth individual, rand(·) is the random number generation function;

[0048] Step S42: improving the convergence factor based on the Gaussian distribution change curve, the formula used is as follows:

[0049] ;

[0050] Where a(g) is the convergence factor at the g-th iteration search, g is the search number index, G is the total number of searches, and η is the cutoff value;

[0051] Step S43: Calculate the optimal individual position, update the fitness value of the individual, sort them in order from small to large in fitness, select the top three individual positions as the first best individual position, the second best individual position and the third best individual position, and design a proportional weighted strategy based on fitness and the top three best individual positions to calculate the optimal individual position. The formula used is as follows:

[0052] ;

[0053] ;

[0054] ;

[0055] Where U o (g) and Z o (g) are the fitness ratio and position ratio of the oth optimal individual position in the gth iterative search, u1(g), u2(g) and u3(g) are the fitness of the first optimal individual position, the second optimal individual position and the third optimal individual position in the gth iterative search, D1(g), D2(g) and D3(g) are the first optimal individual position, the second optimal individual position and the third optimal individual position in the gth iterative search, respectively. best (g) is the optimal individual position in the g-th iteration search, u o (g) is the fitness of the oth optimal individual position in the gth iteration search, D o (g) is the oth optimal individual position in the g-th iterative search, U1(g) and Z1(g) are the fitness ratio and position ratio of the first optimal individual position in the g-th iterative search, U2(g) and Z2(g) are the fitness ratio and position ratio of the second optimal individual position in the g-th iterative search, U3(g) and Z3(g) are the fitness ratio and position ratio of the third optimal individual position in the g-th iterative search;

[0056] Step S44: Update individual positions using the following formula:

[0057] ;

[0058] Where D w (g+1) and D w (g) is the position of the wth individual in the g+1th and gth iteration searches, S1 and S2 are the first random value and the second random value, respectively;

[0059] Step S45: Determine the optimal parameters of the model, pre-set the fitness threshold, update the individual fitness, and when the minimum fitness is lower than the fitness threshold, use the individual position with the minimum fitness as the optimal parameters of the model, and build an agricultural machinery operation status monitoring model based on the optimal parameters of the model; otherwise, if the total number of searches is reached, go to step S41; otherwise, go to step S42 to continue the iterative search.

[0060] Furthermore, in step S5, the real-time operation status monitoring is to obtain the real-time operation status of the agricultural machinery based on the output label of the model, thereby realizing real-time monitoring of the agricultural machinery.

[0061] The agricultural machinery operation status monitoring system based on the Internet of Things provided by the present invention comprises a data acquisition module, a data preprocessing module, a module for constructing an agricultural machinery operation status monitoring model, a module for searching parameters of the agricultural machinery operation status monitoring model and a real-time operation status monitoring module;

[0062] The data acquisition module collects historical operation data and real-time operation data of the agricultural machinery based on the Internet of Things device, and sends the data to the data preprocessing module;

[0063] The data preprocessing module performs data cleaning, data conversion, data normalization and data set construction on the collected data through the central processing center, and sends the data to the agricultural machinery operation status monitoring model construction module;

[0064] The module for constructing the agricultural machinery operation status monitoring model designs an encoder to extract the features of the source domain and the target domain data, designs a decoder to complete data reconstruction, calculates the difference in the distribution of the source domain and the target domain by reproducing the Hilbert space of the kernel, aligns the data distribution of the two domains by minimizing the difference, introduces a smooth label algorithm to optimize the encoding of the data category label, designs a total loss function based on the reconstruction loss, the difference loss and the classification loss, and sends the data to the parameter search module of the agricultural machinery operation status monitoring model;

[0065] The parameter search module of the agricultural machinery operation status monitoring model introduces random variables to improve the tent mapping, thereby initializing the individual position, improving the convergence factor based on the Gaussian distribution change curve, calculating the optimal individual position based on the proportional weighting strategy, and updating all individual positions in combination with the improved convergence factor and the optimal individual position to obtain the optimal parameters of the model, and sending the data to the real-time operation status monitoring module;

[0066] The real-time operation status monitoring module obtains the real-time operation status of the agricultural machinery based on the output label of the model, thereby realizing real-time monitoring of the agricultural machinery.

[0067] The beneficial effects achieved by the present invention using the above scheme are as follows:

[0068] (1) In view of the problems in the existing agricultural machinery operation status monitoring methods, such as the difficulty in accurately extracting agricultural machinery operation features, differences in the distribution of agricultural machinery operation data between different domains, and the neglect of the correlation between data categories during classification, resulting in low classification accuracy, this scheme designs an encoder network structure, adopts multiple one-dimensional convolution blocks, combines the maximum pooling layer and the RELU activation function, and specifically extracts the features of the source domain and target domain data, maps the high-dimensional data to the low-dimensional space, effectively reduces the dimension of the data, and retains the effective information; by reproducing the Hilbert space of the kernel, the difference in the distribution of the source domain and the target domain is calculated, and the data distribution of the two domains is aligned by minimizing the difference loss, thereby improving the model's adaptability to different data domains; introduces a smooth label algorithm to optimize the encoding of data category labels; designs a total loss function based on reconstruction loss, difference loss, and classification loss, which helps to more accurately understand and analyze agricultural machinery operation status data and improve the accuracy and efficiency of agricultural machinery operation status monitoring.

[0069] (2) In view of the problems that the existing agricultural machinery operation status monitoring methods have poor adaptability to different types of agricultural machinery and work, poor monitoring performance under new environments and new data, and weak generalization ability, this scheme introduces random variables to improve the tent mapping method, generates F initial positions of individuals, increases the diversity of search, and reduces the possibility of falling into the local optimal solution; adopts a convergence factor based on the Gaussian distribution change curve, and gradually reduces the value of the convergence factor according to the number of iterative searches, which is conducive to balancing the ability of local search and global search during the search process, and improving the convergence speed and search efficiency; by designing a proportional weighting strategy based on fitness and the top three optimal individual positions, the fitness and position information of the individual positions are comprehensively considered, and the optimal individual position is determined more effectively, which is conducive to improving the convergence and stability of the optimization algorithm; combining the improved convergence factor based on the Gaussian distribution change curve and the optimal individual position, it is helpful to improve the efficiency and accuracy of the search, improve the comprehensiveness and accuracy of monitoring the operation status of various types of agricultural machinery and work, and better guide the use, maintenance and management of agricultural machinery. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] Figure 1 A schematic diagram of the flow of the method for monitoring the operating status of agricultural machinery based on the Internet of Things provided by the present invention;

[0071] Figure 2 A schematic diagram of an agricultural machinery operation status monitoring system based on the Internet of Things provided by the present invention;

[0072] Figure 3 is a schematic diagram of the process of step S3;

[0073] Figure 4 It is a schematic diagram of the process of step S4.

[0074] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention. DETAILED DESCRIPTION

[0075] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0076] In the description of the present invention, it should be understood that terms such as “upper”, “lower”, “front”, “back”, “left”, “right”, “top”, “bottom”, “inside” and “outside” indicating directions or positional relationships are based on the directions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore should not be understood as limiting the present invention.

[0077] Example 1, see Figure 1 The present invention provides an agricultural machinery operation status monitoring method based on the Internet of Things, the method comprising the following steps:

[0078] Step S1: data collection, collecting historical and real-time operation data of agricultural machinery based on IoT devices;

[0079] Step S2: Data preprocessing, data cleaning, data conversion, data normalization and data set construction are performed on the collected data through the central processing center;

[0080] Step S3: Construct an agricultural machinery operation status monitoring model, design an encoder to extract the features of source domain and target domain data, design a decoder to complete data reconstruction, calculate the difference between the source domain and target domain distribution by reproducing the kernel's Hilbert space, align the data distribution of the two domains by minimizing the difference, introduce a smooth label algorithm to optimize the encoding of data category labels, and design a total loss function based on reconstruction loss, difference loss, and classification loss;

[0081] Step S4: Parameter search of the agricultural machinery operation status monitoring model, introducing random variables to improve the tent mapping, and then initializing the individual position, improving the convergence factor based on the Gaussian distribution change curve, calculating the optimal individual position based on the proportional weighting strategy, and updating all individual positions in combination with the improved convergence factor and the optimal individual position to obtain the optimal parameters of the model;

[0082] Step S5: Real-time operation status monitoring, obtaining the real-time operation status of the agricultural machinery based on the output label of the model, and realizing real-time monitoring of the agricultural machinery.

[0083] Example 2, see Figure 1 This embodiment is based on the above embodiment. In step S1, data collection is to deploy the Internet of Things device on the agricultural machinery, and collect the historical operation data and real-time operation data of the agricultural machinery based on the Internet of Things device; the historical operation data and real-time operation data both include agricultural machinery equipment information, environmental parameter data, operation trajectory data, sensor data and maintenance records, and the historical operation data also includes the operation status of the agricultural machinery; the operation status of the agricultural machinery includes normal and agricultural machinery fault types; the operation status is used as a data category label, and the collected data is transmitted to the central processing center.

[0084] Example 3, see Figure 1 This embodiment is based on the above embodiment. In step S2, data cleaning includes processing noise, duplicate values, missing values ​​and outliers; data conversion is to convert the data after data cleaning into vector form; data normalization is to unify the data range based on the maximum and minimum normalization method; and data set construction is to construct a source domain training set with data category labels and a target domain training set without data category labels based on the historical operation data and real-time operation data after data cleaning, data conversion and data normalization.

[0085] Example 4, see Figure 1 and Figure 3 Based on the above embodiment, in step S3, building an agricultural machinery operation status monitoring model specifically includes the following steps:

[0086] Step S31: Extract features and design an encoder to extract features from input data; the encoder consists of 5 one-dimensional convolution blocks, each of which contains a convolution layer and a maximum pooling layer, activated by the RELU function; the size of the first convolution kernel is 1×64, the size of the remaining four convolution kernels is 1×3, and the step size of the maximum pooling layer is 1×2; the formula used to extract features is as follows:

[0087] ;

[0088] ;

[0089] In the formula, Q y and Q m are the source domain features and target domain features extracted from the source domain training set and the target domain training set, respectively. enc is the network structure of the encoder. y and X m They are the source domain training set and the target domain training set respectively;

[0090] Step S32: Data reconstruction, designing a decoder for data reconstruction; the decoder consists of 5 one-dimensional deconvolution blocks, each of which contains a deconvolution layer and an upsampling layer, and the structure of the decoder is completely symmetrical with that of the encoder; wherein the size of the first deconvolution kernel is 1×64, the size of the remaining four deconvolution kernels is 1×3, and the step size of the upsampling layer is 1×2; comprising the following steps:

[0091] Step S321: Calculate the reconstructed target domain data, using the following formula:

[0092] ;

[0093] In the formula, is the reconstructed target domain data, dec is the network structure of the decoder;

[0094] Step S322: Calculate the reconstruction loss of the target domain data. The formula used is as follows:

[0095] ;

[0096] Where, L rec is the reconstruction loss of the target domain data, N m is the number of target domain data in the target domain training set, and are the i-th target domain data before reconstruction and the i-th target domain data after reconstruction, i is the target domain data index;

[0097] Step S33: Domain alignment, mapping both the source domain data and the target domain data to the Hilbert space with the reproduction kernel, aligning the distributions of the two domains by minimizing the difference between the source domain distribution and the target domain distribution, including the following steps:

[0098] Step S331: Calculate the difference between the source domain distribution and the target domain distribution. The formula used is as follows:

[0099] ;

[0100] ;

[0101] ;

[0102] In the formula, is the difference between the source domain distribution and the target domain distribution, p （c） and q （c） are the source domain distribution and the target domain distribution, l is the feature layer index, C is the total number of data category labels, c is the data category label index, and N y is the number of source domain data in the source domain training set, and are the weights of the i-th and j-th source domain data belonging to the c-th data category label in the source domain training set, and are the weights of the i-th and j-th target domain data belonging to the c-th data category label in the target domain training set, and are the number of source domain data and target domain data in the source domain training set and the target domain training set that belong to the cth category as the i-th source domain data and target domain data, respectively. k(·) is the kernel function of the Hilbert space with a reproduction kernel. and are the activation values ​​of the i-th and j-th source domain data in the l-th feature layer, respectively. and are the activation values ​​of the i-th and j-th target domain data in the l-th feature layer respectively;

[0103] Step S332: Calculate the difference loss between the source domain distribution and the target domain distribution. The formula used is as follows:

[0104] ;

[0105] Where, L LMMD is the difference loss between the source domain distribution and the target domain distribution, l max is the number of feature layers;

[0106] Step S34: Classification, input the extracted target domain features into the softmax function to calculate the probability that the target domain data belongs to each data category label, and select the data category label with the maximum probability as the output label of the target domain data on the model, including the following steps:

[0107] Step S341: Introduce a smooth label algorithm to optimize the one-hot encoding of the real data category label. The formula used is as follows:

[0108] ;

[0109] In the formula, ε is the smoothing parameter, yes The real data labels, yes The one-hot encoding of the real data label, C is the total number of data category labels, and c is the data category label index;

[0110] Step S342: Classification loss, the formula used is as follows:

[0111] ;

[0112] Where, L cls is the classification loss, and They are In the target domain training set, it belongs to class c and The probability of the class label of the class data, W and b are the weight and bias respectively;

[0113] Step S35: Design the total loss function of the model, and combine the reconstruction loss, difference loss and classification loss according to the weights to obtain the total loss function of the model. The formula used is as follows:

[0114] ;

[0115] Where L is the total loss function, β1, β2 and β3 are the first weight, second weight and third weight respectively, and L LMMD It is the difference loss between the source domain distribution and the target domain distribution.

[0116] By performing the above operations, the existing agricultural machinery operation status monitoring methods have the problems that the agricultural machinery operation features are difficult to extract accurately, the distribution of agricultural machinery operation data between different domains is different, and the correlation between data categories is ignored during classification, resulting in low classification accuracy. This scheme designs an encoder network structure, adopts multiple one-dimensional convolution blocks, combines the maximum pooling layer and the RELU activation function, and specifically extracts the features of the source domain and target domain data, maps the high-dimensional data to the low-dimensional space, effectively reduces the dimension of the data, and retains the effective information; by reproducing the Hilbert space of the kernel, the difference in the distribution of the source domain and the target domain is calculated, and the data distribution of the two domains is aligned by minimizing the difference loss, thereby improving the model's adaptability to different data domains; introduces a smooth label algorithm to optimize the encoding of data category labels; designs a total loss function based on reconstruction loss, difference loss and classification loss, which helps to more accurately understand and analyze agricultural machinery operation status data and improve the monitoring accuracy and efficiency of agricultural machinery operation status.

[0117] Example 5, see Figure 1 and Figure 4 This embodiment is based on the above embodiment. In step S4, the agricultural machinery operation status monitoring model parameter search specifically includes the following steps:

[0118] Step S41: Initial individual position, establish a parameter search space for the first weight β1, the second weight β2, the third weight β3, the weight W and the bias b in the agricultural machinery operation status monitoring model, introduce random variables to improve the tent mapping, initialize F individual positions in the parameter search space based on the improved tent mapping, use the individual position as a representative of the parameters of the agricultural machinery operation status monitoring model, and use the total loss function of the agricultural machinery operation status monitoring model trained based on the individual position as the individual fitness value. The formula used for the initial individual position is as follows:

[0119] ;

[0120] ;

[0121] In the formula, A w+1 and A w are the w+1th and wth improved chaotic sequences respectively, v is the control value, lb and ub are the lower and upper limits of the parameter search space respectively, D w is the initial position of the wth individual, rand(·) is the random number generation function;

[0122] Step S42: improving the convergence factor based on the Gaussian distribution change curve, the formula used is as follows:

[0123] ;

[0124] Where a(g) is the convergence factor at the g-th iteration search, g is the search number index, G is the total number of searches, and η is the cutoff value;

[0125] Step S43: Calculate the optimal individual position, update the fitness value of the individual, sort them in order from small to large in fitness, select the top three individual positions as the first best individual position, the second best individual position and the third best individual position, and design a proportional weighted strategy based on fitness and the top three best individual positions to calculate the optimal individual position. The formula used is as follows:

[0126] ;

[0127] ;

[0128] ;

[0129] Where U o (g) and Z o (g) are the fitness ratio and position ratio of the oth optimal individual position in the gth iterative search, u1(g), u2(g) and u3(g) are the fitness of the first optimal individual position, the second optimal individual position and the third optimal individual position in the gth iterative search, D1(g), D2(g) and D3(g) are the first optimal individual position, the second optimal individual position and the third optimal individual position in the gth iterative search, respectively. best (g) is the optimal individual position in the g-th iteration search, u o (g) is the fitness of the oth optimal individual position in the gth iteration search, D o(g) is the oth optimal individual position in the g-th iterative search, U1(g) and Z1(g) are the fitness ratio and position ratio of the first optimal individual position in the g-th iterative search, U2(g) and Z2(g) are the fitness ratio and position ratio of the second optimal individual position in the g-th iterative search, U3(g) and Z3(g) are the fitness ratio and position ratio of the third optimal individual position in the g-th iterative search;

[0130] Step S44: Update individual positions using the following formula:

[0131] ;

[0132] Where D w (g+1) and D w (g) is the position of the wth individual in the g+1th and gth iteration searches, S1 and S2 are the first random value and the second random value, respectively;

[0133] Step S45: Determine the optimal parameters of the model, pre-set the fitness threshold, update the individual fitness, and when the minimum fitness is lower than the fitness threshold, use the individual position with the minimum fitness as the optimal parameters of the model, and build an agricultural machinery operation status monitoring model based on the optimal parameters of the model; otherwise, if the total number of searches is reached, go to step S41; otherwise, go to step S42 to continue the iterative search.

[0134] By performing the above operations, the existing agricultural machinery operation status monitoring methods have poor adaptability to different agricultural machinery types and tasks, poor monitoring performance under new environments and new data, and weak generalization ability. This scheme introduces random variables to improve the tent mapping method, generates F initial positions of individuals, increases the diversity of search, and reduces the possibility of falling into the local optimal solution; adopts a convergence factor based on the Gaussian distribution change curve, and gradually reduces the value of the convergence factor according to the number of iterative searches, which is conducive to balancing the ability of local search and global search during the search process, and improving the convergence speed and search efficiency; by designing a proportional weighted strategy based on fitness and the top three optimal individual positions, the fitness and position information of the individual positions are comprehensively considered, and the optimal individual position is determined more effectively, which is conducive to improving the convergence and stability of the optimization algorithm; the combination of the improved convergence factor based on the Gaussian distribution change curve and the optimal individual position is helpful to improve the efficiency and accuracy of the search, improve the comprehensiveness and accuracy of monitoring the operation status of various agricultural machinery and work, and better guide the use, maintenance and management of agricultural machinery.

[0135] Example 6, see Figure 2, this embodiment is based on the above embodiment, the agricultural machinery operation status monitoring system based on the Internet of Things provided by the present invention includes a data acquisition module, a data preprocessing module, a module for building an agricultural machinery operation status monitoring model, an agricultural machinery operation status monitoring model parameter search module and a real-time operation status monitoring module;

[0136] The data acquisition module collects historical operation data and real-time operation data of the agricultural machinery based on the Internet of Things device, and sends the data to the data preprocessing module;

[0137] The data preprocessing module performs data cleaning, data conversion, data normalization and data set construction on the collected data through the central processing center, and sends the data to the agricultural machinery operation status monitoring model construction module;

[0138] The module for constructing the agricultural machinery operation status monitoring model designs an encoder to extract the features of the source domain and the target domain data, designs a decoder to complete data reconstruction, calculates the difference in the distribution of the source domain and the target domain by reproducing the Hilbert space of the kernel, aligns the data distribution of the two domains by minimizing the difference, introduces a smooth label algorithm to optimize the encoding of the data category label, designs a total loss function based on the reconstruction loss, the difference loss and the classification loss, and sends the data to the parameter search module of the agricultural machinery operation status monitoring model;

[0139] The parameter search module of the agricultural machinery operation status monitoring model introduces random variables to improve the tent mapping, thereby initializing the individual position, improving the convergence factor based on the Gaussian distribution change curve, calculating the optimal individual position based on the proportional weighting strategy, and updating all individual positions in combination with the improved convergence factor and the optimal individual position to obtain the optimal parameters of the model, and sending the data to the real-time operation status monitoring module;

[0140] The real-time operation status monitoring module obtains the real-time operation status of the agricultural machinery based on the output label of the model, thereby realizing real-time monitoring of the agricultural machinery.

[0141] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0142] While the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that many changes, modifications, substitutions and variations can be made to the embodiments without departing from the principles and spirit of the invention.

[0143] The present invention and its embodiments are described above, and such description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if ordinary technicians in the field are inspired by it, without departing from the purpose of the invention, they can design a structure and embodiment similar to the technical solution without creativity, which should belong to the protection scope of the present invention.

Claims

1. The method for monitoring the operation status of agricultural machinery based on the Internet of Things is characterized by: The method comprises the following steps: Step S1: data collection; Step S2: data preprocessing; Step S3: Construct an agricultural machinery operation status monitoring model, design an encoder to extract the features of source domain and target domain data, design a decoder to complete data reconstruction, calculate the difference between the source domain and target domain distribution by reproducing the kernel's Hilbert space, align the data distribution of the two domains by minimizing the difference, introduce a smooth label algorithm to optimize the encoding of data category labels, and design a total loss function based on reconstruction loss, difference loss, and classification loss; Step S4: Parameter search of the agricultural machinery operation status monitoring model, introducing random variables to improve the tent mapping, and then initializing the individual position, improving the convergence factor based on the Gaussian distribution change curve, calculating the optimal individual position based on the proportional weighting strategy, and updating all individual positions in combination with the improved convergence factor and the optimal individual position to obtain the optimal parameters of the model; Step S5: real-time operation status monitoring; Step S4 includes step S42: improving the convergence factor based on the Gaussian distribution change curve, and the formula used is as follows: ； Where a(g) is the convergence factor at the g-th iteration search, g is the search number index, G is the total number of searches, and η is the cutoff value; In step S4, the agricultural machinery operation status monitoring model parameter search specifically includes the following steps: Step S41: Initial individual position, establish a parameter search space for the first weight β1, the second weight β2, the third weight β3, the weight W and the bias b in the agricultural machinery operation status monitoring model, introduce random variables to improve the tent mapping, initialize F individual positions in the parameter search space based on the improved tent mapping, use the individual position as a representative of the parameters of the agricultural machinery operation status monitoring model, and use the total loss function of the agricultural machinery operation status monitoring model trained based on the individual position as the individual fitness value. The formula used for the initial individual position is as follows: ； ； In the formula, A w+1 and A w are the w+1th and wth improved chaotic sequences respectively, v is the control value, lb and ub are the lower and upper limits of the parameter search space respectively, D w is the initial position of the wth individual, rand(·) is the random number generation function; Step S42: improving the convergence factor based on the Gaussian distribution change curve; Step S43: Calculate the optimal individual position, update the fitness value of the individual, sort them in order from small to large in fitness, select the top three individual positions as the first best individual position, the second best individual position and the third best individual position, and design a proportional weighted strategy based on fitness and the top three best individual positions to calculate the optimal individual position. The formula used is as follows: ； ； ； Where U o (g) and Z o (g) are the fitness ratio and position ratio of the oth optimal individual position in the gth iterative search, u1(g), u2(g) and u3(g) are the fitness of the first optimal individual position, the second optimal individual position and the third optimal individual position in the gth iterative search, D1(g), D2(g) and D3(g) are the first optimal individual position, the second optimal individual position and the third optimal individual position in the gth iterative search, respectively. best (g) is the optimal individual position in the g-th iteration search, u o (g) is the fitness of the oth optimal individual position in the gth iteration search, D o (g) is the oth optimal individual position in the g-th iterative search, U1(g) and Z1(g) are the fitness ratio and position ratio of the first optimal individual position in the g-th iterative search, U2(g) and Z2(g) are the fitness ratio and position ratio of the second optimal individual position in the g-th iterative search, U3(g) and Z3(g) are the fitness ratio and position ratio of the third optimal individual position in the g-th iterative search; Step S44: Update individual positions using the following formula: ； Where D w (g+1) and D w (g) is the position of the wth individual in the g+1th and gth iteration searches, S1 and S2 are the first random value and the second random value, respectively; Step S45: Determine the optimal parameters of the model, pre-set the fitness threshold, update the individual fitness, and when the minimum fitness is lower than the fitness threshold, use the individual position with the minimum fitness as the optimal parameters of the model, and build an agricultural machinery operation status monitoring model based on the optimal parameters of the model; otherwise, if the total number of searches is reached, go to step S41; otherwise, go to step S42 to continue the iterative search.

2. The method for monitoring the operation status of agricultural machinery based on the Internet of Things according to claim 1, characterized in that: In step S3, the construction of the agricultural machinery operation status monitoring model specifically includes the following steps: Step S31: Extract features and design an encoder to extract features from input data; the encoder consists of 5 one-dimensional convolution blocks, each of which contains a convolution layer and a maximum pooling layer, activated by the RELU function; the size of the first convolution kernel is 1×64, the size of the remaining four convolution kernels is 1×3, and the step size of the maximum pooling layer is 1×2; the formula used to extract features is as follows: ； ； In the formula, Q y and Q m are the source domain features and target domain features extracted from the source domain training set and the target domain training set, respectively. enc is the network structure of the encoder. y and X m They are the source domain training set and the target domain training set respectively; Step S32: Data reconstruction, designing a decoder for data reconstruction; the decoder consists of 5 one-dimensional deconvolution blocks, each of which contains a deconvolution layer and an upsampling layer, and the structure of the decoder is completely symmetrical with that of the encoder; wherein the size of the first deconvolution kernel is 1×64, the size of the remaining four deconvolution kernels is 1×3, and the step size of the upsampling layer is 1×2; comprising the following steps: Step S321: Calculate the reconstructed target domain data, using the following formula: ； In the formula, is the reconstructed target domain data, dec is the network structure of the decoder; Step S322: Calculate the reconstruction loss of the target domain data. The formula used is as follows: ； Where, L rec is the reconstruction loss of the target domain data, N m is the number of target domain data in the target domain training set, and are the i-th target domain data before reconstruction and the i-th target domain data after reconstruction, i is the target domain data index; Step S33: Domain alignment, mapping both the source domain data and the target domain data to the Hilbert space with a reproduction kernel, and aligning the distributions of the two domains by minimizing the difference between the source domain distribution and the target domain distribution; Step S34: Classification, input the extracted target domain features into the softmax function to calculate the probability that the target domain data belongs to each data category label, and select the data category label with the maximum probability as the output label of the target domain data on the model, including the following steps: Step S341: Introduce a smooth label algorithm to optimize the one-hot encoding of the real data category label. The formula used is as follows: ； In the formula, ε is the smoothing parameter, yes The real data labels, yes The one-hot encoding of the real data label, C is the total number of data category labels, and c is the data category label index; Step S342: Classification loss, the formula used is as follows: ； Where, L cls is the classification loss, and They are In the target domain training set, it belongs to class c and The probability of the class label of the class data, W and b are the weight and bias respectively; Step S35: Design the total loss function of the model, and combine the reconstruction loss, difference loss and classification loss according to the weights to obtain the total loss function of the model. The formula used is as follows: ； Where L is the total loss function, β1, β2 and β3 are the first weight, second weight and third weight respectively, and L LMMD It is the difference loss between the source domain distribution and the target domain distribution.

3. The method for monitoring the operation status of agricultural machinery based on the Internet of Things according to claim 2 is characterized in that: In step S33, the domain alignment specifically includes the following steps: Step S331: Calculate the difference between the source domain distribution and the target domain distribution. The formula used is as follows: ； ； ； In the formula, is the difference between the source domain distribution and the target domain distribution, p （c） and q （c） are the source domain distribution and the target domain distribution, l is the feature layer index, C is the total number of data category labels, c is the data category label index, and N y is the number of source domain data in the source domain training set, and are the weights of the i-th and j-th source domain data belonging to the c-th data category label in the source domain training set, and are the weights of the i-th and j-th target domain data belonging to the c-th data category label in the target domain training set, and are the number of source domain data and target domain data in the source domain training set and the target domain training set that belong to the cth category as the i-th source domain data and target domain data, respectively. k(·) is the kernel function of the Hilbert space with a reproduction kernel. and are the activation values ​​of the i-th and j-th source domain data in the l-th feature layer, respectively. and are the activation values ​​of the i-th and j-th target domain data in the l-th feature layer respectively; Step S332: Calculate the difference loss between the source domain distribution and the target domain distribution. The formula used is as follows: ； Where, L LMMD is the difference loss between the source domain distribution and the target domain distribution, l max is the number of feature layers.

4. The method for monitoring the operation status of agricultural machinery based on the Internet of Things according to claim 1, characterized in that: In step S1, the data collection is to deploy the Internet of Things device on the agricultural machinery, and collect the historical operation data and real-time operation data of the agricultural machinery based on the Internet of Things device; the historical operation data and real-time operation data both include agricultural machinery equipment information, environmental parameter data, operation trajectory data, sensor data and maintenance records, and the historical operation data also includes the operation status of the agricultural machinery; the operation status of the agricultural machinery includes normal and agricultural machinery fault types; the operation status is used as a data category label, and the collected data is transmitted to the central processing center.

5. The method for monitoring the operation status of agricultural machinery based on the Internet of Things according to claim 1, characterized in that: In step S2, the data preprocessing is to clean, convert, normalize and construct a data set of the collected data by the central processing center; The data set is constructed by constructing a source domain training set with data category labels and a target domain training set without data category labels based on the historical operation data and real-time operation data after data cleaning, data conversion, and data normalization.

6. The method for monitoring the operation status of agricultural machinery based on the Internet of Things according to claim 1, characterized in that: In step S5, the real-time operation status monitoring is to obtain the real-time operation status of the agricultural machinery based on the output label of the model, so as to realize the real-time monitoring of the agricultural machinery.

7. An agricultural machinery operation status monitoring system based on the Internet of Things, used to implement the agricultural machinery operation status monitoring method based on the Internet of Things as described in any one of claims 1 to 6, characterized in that: It includes a data acquisition module, a data preprocessing module, a module for building an agricultural machinery operation status monitoring model, a module for searching parameters of the agricultural machinery operation status monitoring model, and a real-time operation status monitoring module; The data acquisition module collects historical operation data and real-time operation data of the agricultural machinery based on the Internet of Things device, and sends the data to the data preprocessing module; The data preprocessing module performs data cleaning, data conversion, data normalization and data set construction on the collected data through the central processing center, and sends the data to the agricultural machinery operation status monitoring model construction module; The module for constructing the agricultural machinery operation status monitoring model designs an encoder to extract the features of the source domain and the target domain data, designs a decoder to complete data reconstruction, calculates the difference in the distribution of the source domain and the target domain by reproducing the Hilbert space of the kernel, aligns the data distribution of the two domains by minimizing the difference, introduces a smooth label algorithm to optimize the encoding of the data category label, designs a total loss function based on the reconstruction loss, the difference loss and the classification loss, and sends the data to the parameter search module of the agricultural machinery operation status monitoring model; The parameter search module of the agricultural machinery operation status monitoring model introduces random variables to improve the tent mapping, thereby initializing the individual position, improving the convergence factor based on the Gaussian distribution change curve, calculating the optimal individual position based on the proportional weighting strategy, and updating all individual positions in combination with the improved convergence factor and the optimal individual position to obtain the optimal parameters of the model, and sending the data to the real-time operation status monitoring module; The real-time operation status monitoring module obtains the real-time operation status of the agricultural machinery based on the output label of the model, thereby realizing real-time monitoring of the agricultural machinery; The agricultural machinery operation status monitoring model parameter search module specifically includes the following steps: Step S41: Initial individual position, establish a parameter search space for the first weight β1, the second weight β2, the third weight β3, the weight W and the bias b in the agricultural machinery operation status monitoring model, introduce random variables to improve the tent mapping, initialize F individual positions in the parameter search space based on the improved tent mapping, use the individual position as a representative of the parameters of the agricultural machinery operation status monitoring model, and use the total loss function of the agricultural machinery operation status monitoring model trained based on the individual position as the individual fitness value. The formula used for the initial individual position is as follows: ； ； In the formula, A w+1 and A w are the w+1th and wth improved chaotic sequences respectively, v is the control value, lb and ub are the lower and upper limits of the parameter search space respectively, D w is the initial position of the wth individual, rand(·) is the random number generation function; Step S42: improving the convergence factor based on the Gaussian distribution change curve, the formula used is as follows: ； Where a(g) is the convergence factor at the g-th iteration search, g is the search number index, G is the total number of searches, and η is the cutoff value; Step S43: Calculate the optimal individual position, update the fitness value of the individual, sort them in order from small to large in fitness, select the top three individual positions as the first best individual position, the second best individual position and the third best individual position, and design a proportional weighted strategy based on fitness and the top three best individual positions to calculate the optimal individual position. The formula used is as follows: ； ； ； Where U o (g) and Z o (g) are the fitness ratio and position ratio of the oth optimal individual position in the gth iterative search, u1(g), u2(g) and u3(g) are the fitness of the first optimal individual position, the second optimal individual position and the third optimal individual position in the gth iterative search, D1(g), D2(g) and D3(g) are the first optimal individual position, the second optimal individual position and the third optimal individual position in the gth iterative search, respectively. best (g) is the optimal individual position in the g-th iteration search, u o (g) is the fitness of the oth optimal individual position in the gth iteration search, D o (g) is the oth optimal individual position in the g-th iterative search, U1(g) and Z1(g) are the fitness ratio and position ratio of the first optimal individual position in the g-th iterative search, U2(g) and Z2(g) are the fitness ratio and position ratio of the second optimal individual position in the g-th iterative search, U3(g) and Z3(g) are the fitness ratio and position ratio of the third optimal individual position in the g-th iterative search; Step S44: Update individual positions using the following formula: ； Where D w (g+1) and D w (g) is the position of the wth individual in the g+1th and gth iteration searches, S1 and S2 are the first random value and the second random value, respectively; Step S45: Determine the optimal parameters of the model, pre-set the fitness threshold, update the fitness of the individual, and when the minimum fitness is lower than the fitness threshold, use the individual position with the minimum fitness as the optimal parameter of the model, and build the agricultural machinery operation status monitoring model based on the optimal parameter of the model; otherwise, if the total number of searches is reached, go to step S41; Otherwise, go to step S42 to continue the iterative search.

Images