An agricultural equipment status monitoring system based on the Internet of Things

Abstract

The present invention relates to the technical field of agricultural equipment status monitoring, and discloses an agricultural equipment status monitoring system based on the Internet of Things, comprising: an acquisition unit, an analysis unit, a stripping unit, a fusion unit and a decision unit. The present invention collects vibration, rotational speed, torque, soil hardness and environmental data in multiple dimensions through an Internet of Things sensor network, generates a physical compensation factor in combination with synchronous analysis of working conditions, effectively strips away the interference of soil hardness on vibration data, then generates a coupled wear index by fusing residual eigenvectors with ambient temperature and humidity, dynamically adjusts the status threshold in combination with historical fault data, and thereby accurately eliminates the interference of soil hardness changes on the vibration spectrum, avoids misjudgment of wear caused by spectrum distortion, realizes dynamic and accurate monitoring of the wear status of harvester blades, and improves the accuracy of agricultural equipment status monitoring in dry field operations.

Description

Technical Field

[0001] The present invention relates to the technical field of agricultural equipment status monitoring, and in particular to an agricultural equipment status monitoring system based on the Internet of Things. Background Art

[0002] At present, when monitoring the status of agricultural equipment, the status of key components of agricultural equipment is often monitored. These components are often in direct contact with the working surface, and their status directly affects the work quality and equipment safety. Among them, the blade of the harvester is one of the key monitoring components. Currently, the wear degree of the monitoring blade can be understood by performing spectral analysis on the vibration data of the blade, so as to achieve the effect of monitoring the status of agricultural equipment.

[0003] However, when working in large-scale dry fields, the soil hardness of each field is different, and the vibration data of the blade will be interfered with by the changes in soil hardness. The interference will cause the vibration spectrum to be distorted. Therefore, the results obtained through spectrum analysis will be somewhat different from the actual blade wear, resulting in poor agricultural equipment status monitoring results. Summary of the Invention

[0004] In view of the shortcomings of the existing technology, the present invention provides an agricultural equipment status monitoring system based on the Internet of Things to solve the above problems.

[0005] The above technical objectives of the present invention are achieved through the following technical solutions:

[0006] An agricultural equipment status monitoring system based on the Internet of Things, comprising:

[0007] The acquisition unit is used to collect vibration data, blade speed and torque data of the monitored object, as well as soil hardness data and ambient temperature and humidity data of the monitoring site in real time through the Internet of Things sensor network. It also obtains historical fault data of agricultural equipment. The monitored object is the harvester blade and the monitoring site is dry land.

[0008] An analysis unit is used to perform synchronous working condition analysis on vibration data, soil hardness data, blade speed and torque data to generate physical compensation factors;

[0009] A stripping unit, used to perform interference stripping on the vibration data according to the physical compensation factor to generate a residual eigenvector;

[0010] A fusion unit is used to fuse the residual feature vector with the ambient temperature and humidity data to generate a coupling wear index;

[0011] The decision unit is used to generate a dynamic state threshold based on historical fault data and ambient temperature and humidity data, compare the coupling wear index with the dynamic state threshold, and generate a state instruction.

[0012] Furthermore, the vibration data, soil hardness data, blade speed and torque data are synchronously analyzed to generate physical compensation factors, including:

[0013] Perform time stamp alignment on pre-processed vibration data, soil hardness data, blade speed, and torque data to generate unified synchronization timing parameters;

[0014] According to the synchronous timing parameters, the physical characteristics of the speed data, torque data and soil hardness data are extracted and integrated to obtain the composite working condition index;

[0015] The composite working condition indicators are mapped to generate physical compensation factors.

[0016] Furthermore, the composite working condition indicators are mapped to generate physical compensation factors, including:

[0017] Decouple the composite working condition indicators to obtain the dynamic working condition vector related only to the instantaneous load of the blade;

[0018] Perform time integration on the dynamic working condition vector to obtain the cumulative fatigue coefficient of the blade under the current load;

[0019] Analyze soil hardness data to derive dynamic weights;

[0020] Based on the dynamic weight, the dynamic working condition vector and the cumulative fatigue coefficient are weighted and fused to generate the physical compensation factor.

[0021] Furthermore, the vibration data is subjected to interference stripping based on the physical compensation factor to generate a residual feature vector, including:

[0022] Perform nonlinear transformation on the physical compensation factor and combine it with the torque data and soil hardness data to obtain the dynamic filter coefficient;

[0023] Filtering the vibration data based on a dynamic filter coefficient to generate a stripped vibration signal;

[0024] Wavelet packet analysis of the stripped vibration signal Kurtosis joint analysis to generate residual eigenvectors.

[0025] Furthermore, the vibration signal after stripping is subjected to wavelet packet Kurtosis joint analysis to generate residual eigenvectors, including:

[0026] Decompose the stripped vibration signal and soil hardness data to generate a feature subband set;

[0027] After calculating the kurtosis value of the characteristic subband set and performing nonlinear weighting, the pulse sensitivity coefficient is obtained;

[0028] The five frequency bands with the highest pulse sensitivity coefficients are selected for analysis to generate residual eigenvectors.

[0029] Furthermore, the residual eigenvector is fused with the ambient temperature and humidity data to generate a coupled wear index, including:

[0030] Quantify the pre-processed ambient temperature and humidity data to obtain the environmental modulation factor;

[0031] Analyze the residual eigenvector to generate the wear sensitivity vector;

[0032] The environmental modulation factor is fused with the wear sensitivity vector to obtain the preliminary coupling index;

[0033] The preliminary coupling index is calibrated based on historical failure data to generate a coupling wear index.

[0034] Furthermore, the residual eigenvector is analyzed to generate the wear sensitivity vector, including:

[0035] Calculate the five frequency bands with the highest pulse sensitivity coefficients in the residual eigenvector separately to obtain the energy proportion of each core frequency band;

[0036] Extract the peak sequence of the vibration signal in the core frequency band to obtain the peak fluctuation characteristics;

[0037] Analyze the phase difference between core frequency bands and construct a phase difference matrix;

[0038] Decompose the phase difference matrix to obtain the phase feature set;

[0039] The energy proportion, peak fluctuation characteristics and phase feature set of each core frequency band are fused to generate a wear sensitivity vector.

[0040] Furthermore, dynamic state thresholds are generated based on historical fault data and ambient temperature and humidity data, including:

[0041] Perform cluster analysis on historical fault data to classify fault mode categories under different environmental temperature and humidity conditions;

[0042] Count the ambient temperature and humidity ranges corresponding to each fault mode category and the equipment status parameter range when the fault occurs, and generate a fault characteristic parameter set;

[0043] Calculate the frequency of historical faults based on the fault characteristic parameter set to obtain the basic threshold;

[0044] The basic threshold is dynamically modified according to the ambient temperature and humidity data to generate a dynamic state threshold.

[0045] Furthermore, the coupling wear index is compared with the dynamic state threshold to generate a state instruction, including:

[0046] The coupling wear index and the dynamic state threshold are calculated to generate the dynamic degradation degree;

[0047] When the dynamic degradation When the value is 0, it indicates that the equipment status is not safe enough and a replacement instruction is issued;

[0048] When the dynamic degradation When it is 0, it indicates that the device status is in a safe range and no command is issued;

[0049] The dynamic degradation degree is corrected in real time based on the ambient temperature and humidity data and historical failure mode categories.

[0050] Furthermore, the dynamic degradation degree is corrected in real time based on the ambient temperature and humidity data and historical failure mode categories, including:

[0051] According to the real-time environmental temperature and humidity data, the nearest fault mode category in the historical fault cluster is matched to generate the mode deviation factor;

[0052] Calculate the current failure mode category to obtain the mean of the historical degradation slope;

[0053] The dynamic degradation degree is corrected in real time according to the mode deviation factor and the average of the historical degradation slope.

[0054] In summary, the present invention mainly has the following beneficial effects:

[0055] Vibration data, soil hardness, blade speed and other multi-dimensional parameters are synchronously acquired through the acquisition unit, and the working condition is synchronously analyzed by the analysis unit. The data timing is unified through the timestamp alignment technology. The speed, torque and soil hardness are combined to generate a composite working condition index. The dynamic working condition vector related only to the blade load is then decoupled through dynamic weights, and finally a physical compensation factor is generated. This factor can dynamically adjust the filter coefficient based on the soil hardness, perform nonlinear filtering on the vibration data, and combine wavelet packet kurtosis analysis to remove environmental interference, so that the residual eigenvector accurately reflects the wear state of the blade itself. Compared with the traditional solution that only relies on vibration spectrum analysis, the present invention can eliminate the spectrum distortion caused by changes in soil hardness, reduce the monitoring error of wear degree, and provide a more realistic data basis for equipment status assessment.

[0056] Cluster analysis is performed on historical fault data to classify fault modes under different temperature and humidity conditions. Dynamic status thresholds are generated in combination with real-time environmental data. For example, when the temperature and humidity in dry fields fluctuate, the nearest fault mode is matched, and the basic threshold is calculated using the fault characteristic parameter set. Dynamic corrections are made based on the current environmental parameters. This avoids the misjudgment problem of traditional fixed thresholds under different soil hardness and climatic conditions. When operating in contiguous dry fields with large differences in soil hardness, the accuracy of monitoring the status of agricultural equipment can still be improved even under environmental interference.

[0057] Through multi-dimensional feature fusion analysis, the comprehensiveness of fault prediction is improved. The fusion unit deeply couples the residual feature vector with the environmental temperature and humidity data, and generates a calibrated coupling wear index by quantifying the environmental modulation factor and extracting the wear-sensitive vector. The decision unit further combines the historical degradation slope with the real-time environmental deviation factor to dynamically correct the degradation degree, breaking through the limitations of traditional single vibration analysis. For example, in scenarios with sudden changes in soil hardness or extreme temperature and humidity, the system can identify early signs of abnormal blade wear through phase difference matrix analysis, while avoiding misjudgment caused by environmental interference, so that the monitoring results are consistent with the actual status of agricultural equipment, thereby improving monitoring accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 This is a schematic diagram of the agricultural equipment status monitoring system based on the Internet of Things of the present invention. DETAILED DESCRIPTION

[0059] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0060] refer to Figure 1 , an agricultural equipment status monitoring system based on the Internet of Things, including:

[0061] The acquisition unit is used to collect vibration data, blade speed and torque data of the monitored object, as well as soil hardness data and ambient temperature and humidity data of the monitoring site in real time through the Internet of Things sensor network. It also obtains historical fault data of agricultural equipment. The monitored object is the harvester blade and the monitoring site is dry land.

[0062] An analysis unit is used to perform synchronous working condition analysis on vibration data, soil hardness data, blade speed and torque data to generate physical compensation factors;

[0063] A stripping unit, used to perform interference stripping on the vibration data according to the physical compensation factor to generate a residual eigenvector;

[0064] A fusion unit is used to fuse the residual feature vector with the ambient temperature and humidity data to generate a coupling wear index;

[0065] The decision unit is used to generate a dynamic state threshold based on historical fault data and ambient temperature and humidity data, compare the coupling wear index with the dynamic state threshold, and generate a state instruction.

[0066] Among them, the IoT sensor network is composed of multiple sensor nodes such as vibration sensors, soil hardness sensors, and temperature and humidity sensors through wireless communication. For harvester blades, vibration sensors are used to capture blade vibration signals, soil hardness sensors are used to obtain soil parameters, and temperature and humidity sensors are used to monitor environmental changes in real time.

[0067] Among them, vibration data includes: vibration amplitude, vibration frequency, vibration direction and vibration signal waveform;

[0068] Blade speed includes: blade rotation speed and speed variation range;

[0069] Torque data includes: driving torque, torque peak, average value and torque change rate;

[0070] Soil hardness data includes: soil hardness value, soil depth and soil position, etc.;

[0071] Historical fault data includes: fault type, fault time, fault location, fault cause, fault record and fault impact, etc.

[0072] The acquisition unit collects data in multiple dimensions, covering information such as vibration, speed, torque, and soil hardness. The analysis unit generates physical compensation factors through synchronous analysis of working conditions. The stripping unit uses its interference stripping of vibration data to eliminate the influence of factors such as soil hardness on vibration data and avoid spectral distortion interference. The fusion unit fuses the processed residual feature vector with the ambient temperature and humidity data to obtain the coupling wear index. The decision unit combines historical fault data and ambient temperature and humidity to generate dynamic state thresholds for comparison. The whole process realizes accurate monitoring of the harvester blade status, overcomes the limitations of single vibration data spectrum analysis, reduces the influence of interference factors, improves the accuracy and reliability of agricultural equipment status monitoring, and ensures operation quality and equipment safety.

[0073] In one case of this embodiment, the vibration data, soil hardness data, blade speed and torque data are synchronously analyzed to generate a physical compensation factor, including:

[0074] The pre-processed vibration data, soil hardness data, blade speed, and torque data are timestamp aligned to generate unified synchronization timing parameters. Specifically, when timestamp aligning the pre-processed data, the timestamps of the vibration data, soil hardness data, blade speed, and torque data are converted to a unified time coordinate system based on the GPS clock. For data with different sampling rates, linear interpolation is used to generate an equally spaced time series based on the minimum common time interval (10ms). The data points are matched by timestamp to form synchronization timing parameters containing unified time tags.

[0075] According to the synchronous timing parameters, the physical characteristics of the speed data, torque data and soil hardness data are extracted and fused to obtain a composite working condition index, specifically including: based on the synchronous timing parameters, the main frequency amplitude in the time domain characteristics is extracted from the speed data, the peak torque is extracted from the torque data, and the gradient change rate is calculated from the soil hardness data. The three types of features, namely the main frequency amplitude, peak torque and gradient change rate, are dimensionlessly processed using the Z-score normalization method. By calculating the difference between each eigenvalue and the mean of the feature and dividing it by the standard deviation, all features are mapped to a unified dimensional space with a mean of 0 and a standard deviation of 1. During the harvesting stage, the speed data weight coefficient is set to 0.45, the torque data weight coefficient is set to 0.35, and the soil hardness data weight coefficient is set to 0.2. The standardized features are weighted and summed to finally generate a composite working condition index that can accurately represent the working condition of the harvester blade.

[0076] The composite working condition indicators are mapped to generate physical compensation factors.

[0077] By aligning timestamps based on the GPS clock and using linear interpolation to regularize data with different sampling rates to the minimum common time interval, unified synchronous timing parameters are formed to ensure the consistency of multi-source data such as vibration and soil hardness in the time dimension, avoiding analysis errors caused by time dislocation. Based on this, physical features of speed, torque, and soil hardness data can be extracted and fused, and Z-score standardization can be used to eliminate dimensional differences. Reasonable weight coefficients are set in combination with different stages to generate composite working condition indicators, which can comprehensively and accurately reflect the working conditions of harvester blades, lay a solid foundation for the subsequent generation of compensation factors, and overcome the shortcomings of inaccurate traditional monitoring data processing.

[0078] By mapping composite working condition indicators to generate physical compensation factors, the interference of factors such as soil hardness on vibration data can be specifically eliminated. From time synchronization to feature fusion, and then to the generation of physical compensation factors, a complete anti-interference analysis chain is formed. Compared with the traditional method that only relies on vibration data spectrum analysis, this solution fully considers the combined impact of multiple working condition factors, effectively avoids the vibration spectrum distortion problem caused by changes in soil hardness, and makes the monitoring results such as blade wear obtained based on vibration data more in line with the actual situation, greatly improving the accuracy of agricultural equipment status monitoring, ensuring the safe operation of equipment and stable operation quality.

[0079] In one case of this embodiment, the composite operating condition index is mapped to generate a physical compensation factor, including:

[0080] The composite working condition indicators are decoupled to obtain a dynamic working condition vector related only to the instantaneous load of the blade. Specifically, the main frequency amplitude of the speed data and the peak torque of the torque data from the synchronization timing parameters are extracted. These features are directly related to the instantaneous load of the blade. The gradient change rate of the soil hardness data is then eliminated. Only the standardized features of speed (weight 0.45) and torque (weight 0.35) are weighted summed and separated through linear combination to obtain a dynamic working condition vector that only reflects the instantaneous load of the blade.

[0081] The dynamic working condition vector is time-integrated to obtain the cumulative fatigue coefficient of the blade under the current load. Specifically, the dynamic working condition vector value in each time interval is multiplied by the time interval (0.01 seconds), and then the products of all intervals are accumulated in sequence. The accumulated result is the cumulative fatigue coefficient of the blade under the current load.

[0082] The soil hardness data is analyzed to obtain dynamic weights, including: calculating the gradient value of two adjacent points at 10ms intervals (the gradient value is the hardness change divided by the time interval), setting the gradient threshold (50N cm² ms), when the gradient exceeds the threshold, it is determined to be a mutation. The number of mutations and the average gradient value within a unit time (1 minute) are counted. Among them, the more mutations and the larger the average gradient, the more complex the soil conditions. The mutation intensity (mutation intensity is equal to the average gradient divided by the threshold) and the number of mutations are weighted and summed. The weights of mutation intensity and mutation number are 0.5 respectively. The summation result is mapped to After the interval, the mapping result is the dynamic weight;

[0083] Based on the dynamic weight, the dynamic working condition vector and the cumulative fatigue coefficient are weighted and integrated to generate the physical compensation factor, including: calculating the dynamic stress of the soil About time The partial derivative of the stress change rate term is obtained , the soil hardness gradient Take the modulus of the gradient, square it, and then multiply it by the soil mutation sensitivity coefficient And take the negative sign and apply the natural exponential function to get the exponential decay factor , multiply the exponential decay factor by the stress change rate term to obtain the integrand , for the integrand from the initial moment To the current moment Integrate and get the integral term ; Set the blade speed of each tool Vibration phase Perform coupling and obtain the coupling function , multiply the coupling function by The weight coefficient of the tool , get the weighted value , sum the weighted values ​​of all tools and apply the modified linear unit function to the summation result to obtain the tool correction result , multiply the tool correction result by the dynamic weight coefficient , and obtain the dynamic correction term ; Add the integral term and the dynamic correction term to obtain the physical compensation factor , in specific applications, it can be achieved through the following calculation formula, for example:

[0084] ;

[0085] Where, represents the physical compensation factor, Indicates time, represents the initial moment, represents the time rate of change of normalized soil dynamic stress, is the normalized soil dynamic stress, represents the exponential function, represents the soil mutation sensitivity coefficient, represents the soil hardness gradient, represents the time differential, represents the dynamic weight coefficient, represents the rectified linear unit function, Indicates the number of tools, Indicates the The weight coefficient of the tool, Indicates blade speed Vibration phase The coupling function of

[0086] The value range is as follows:

[0087] When the harvester is doing shallow tillage, if the gradient of the dry land changes slowly and the soil hardness of the dry land is low, The value range is 0.2 1. Because the soft soil has little interference on the blade vibration, there is no need for strong attenuation, so The value is too low;

[0088] When the harvester is deep plowing, if the dry land has frequent gradient changes and the soil hardness is high, The value range is 2 3. Then filter the vibration noise caused by hard soil gradient and reduce interference, so The value is too high.

[0089] The value range is as follows:

[0090] When the soil complexity is simple, such as dry land that is loose and uniform after fine tillage and retains moisture well after irrigation, The value range is 0.1 0.3, because the soil has little interference on the blade vibration and no strong compensation is required, so The value is too low;

[0091] When the soil complexity is moderate, such as dry land with slight compaction and uneven soil layers during the rotation period, The value range is 0.4 0.7, because the soil disturbance is comparable to the blade's own vibration, so The value is moderate;

[0092] When the soil is complex, such as dry land that has not been cultivated for a long time and contains a lot of gravel and is first cultivated after rain and contains hard clay, then The value range is 0.8 1. Because the soil has many mutations and the vibration and noise are serious, The value is too high.

[0093] The value range is as follows:

[0094] When the blade is the main cutting area blade, such as the blade in the middle row of the harvester with heavy load, then The value range is 0.6 1. Because the blade in the main cutting area has a strong contact strength with the soil, bears a high load, and wears relatively quickly, its vibration data is more critical for monitoring the overall wear state of the blade. In order to highlight the influence of this type of blade in compensation correction, The value is too high;

[0095] When the blade is an auxiliary cutting area blade, such as a blade at the edge of a harvester with a small load, The value range is 0 0.4, because the blades in the auxiliary cutting area have relatively little contact with the soil and bear relatively little load, vibration interference is mostly a secondary factor, and its impact on overall wear monitoring is not as great as that of the blades in the main cutting area. Reducing its weight can avoid misjudgment of secondary factors, so The value is too low.

[0096] By decoupling the composite working condition indicators, eliminating the gradient change rate of the soil hardness data, and only weighting the sum of the standardized characteristics of the speed and torque, the dynamic working condition vector that only reflects the instantaneous load of the blade is separated, thus avoiding the direct interference of soil hardness changes on the vibration data. At the same time, by analyzing the number of gradient mutations and the average gradient value of the soil hardness data, dynamic weights are generated, which can adaptively adjust the compensation strength according to the complexity of the soil conditions, effectively filtering the vibration noise caused by the hard soil gradient, making the vibration spectrum analysis results closer to the actual wear state of the blade, and improving the authenticity of the monitoring data.

[0097] By integrating the dynamic working condition vector over time to obtain the cumulative fatigue coefficient, and combining it with parameters such as the soil dynamic stress change rate and the square of the hardness gradient modulus to generate a physical compensation factor, the fatigue accumulation effect of the blade under different load and soil conditions can be comprehensively reflected. By setting differentiated weights for the blades in the main cutting area and the auxiliary cutting area, the impact of key blade vibration data on wear monitoring is highlighted, avoiding misjudgment of minor factors. The compensation parameters can be adaptively adjusted in different operating scenarios such as shallow plowing and deep plowing in dry fields, reducing the distortion effect of soil interference on spectrum analysis, and significantly improving the accuracy and reliability of agricultural equipment status monitoring.

[0098] In one case of this embodiment, interference stripping is performed on vibration data according to the physical compensation factor to generate a residual feature vector, including:

[0099] The physical compensation factor is nonlinearly transformed and combined with the torque data and soil hardness data to obtain the dynamic filter coefficient. Specifically, the physical compensation factor is nonlinearly transformed, that is, all values ​​less than 0 in the factor are set to 0, and the positive part is retained. For the torque data, its peak torque value is extracted and converted into dimensionless data through the Z-score normalization method. For the soil hardness data, the gradient change rate of adjacent time points is calculated and the same normalization processing is performed to make the two types of data in a unified dimensional space. The standardized torque (weighted 0.6) and soil hardness data (weighted 0.4) are weighted summed with the physical compensation factor after nonlinear transformation. Then, through normalization processing (mapping the weighted sum result to the range of 0 to 1), a dynamic filter coefficient that is dynamically adjusted according to the real-time working conditions is generated. The dynamic filter coefficient can adaptively match the vibration interference characteristics under different working environments.

[0100] The vibration data is filtered based on the dynamic filter coefficient to generate a stripped vibration signal, specifically including: for the vibration data in the synchronous timing parameters, for each time point, the dynamic filter coefficient is multiplied by the vibration value at that moment. When the dynamic filter coefficient is closer to 1, the more vibration value is retained, and when the dynamic filter coefficient is closer to 0, the more vibration value is attenuated. Through point-by-point weighted operation, the interference vibration caused by sudden changes in soil hardness, load fluctuations and other working conditions can be adaptively attenuated, and only the real vibration characteristics reflecting blade wear are retained. Finally, a pure vibration signal after stripping the interference is generated, which is the stripped vibration signal;

[0101] Wavelet packet analysis of the stripped vibration signal Kurtosis joint analysis to generate residual eigenvectors.

[0102] By nonlinearly transforming the physical compensation factor and integrating the torque and soil hardness data, an adaptive dynamic filter coefficient is generated. The vibration interference characteristics can be accurately matched according to the real-time working conditions. The coefficient is used to perform point-by-point weighted filtering on the vibration data, which can adaptively attenuate the interference vibration caused by sudden changes in soil hardness, load fluctuations, etc., retaining only the true vibration characteristics reflecting blade wear. Combined with the wavelet packet kurtosis joint analysis, a more sensitive residual eigenvector can be extracted from the pure vibration signal after stripping the interference, effectively avoiding the spectrum distortion problem caused by changes in soil hardness, so that the vibration spectrum analysis results more accurately reflect the actual wear status of the blade.

[0103] In one case of this embodiment, the vibration signal after stripping is subjected to wavelet packet Kurtosis joint analysis to generate residual eigenvectors, including:

[0104] The stripped vibration signal and soil hardness data are decomposed to generate a feature sub-band set, specifically including: performing 3 A five-layer wavelet packet decomposition operation is performed. Wavelet packet decomposition maps the vibration signal and soil hardness data to different frequency bands based on the signal frequency characteristics, generating multiple non-overlapping sub-band signals. Each sub-band corresponds to a specific frequency range. All sub-band signals obtained from the decomposition of the two types of data are integrated to construct a feature sub-band set containing vibration characteristics and soil hardness characteristics.

[0105] After calculating the kurtosis value of the characteristic subband set, nonlinear weighting is performed to obtain the pulse sensitivity coefficient, which specifically includes: calculating the kurtosis value of each subband signal in the characteristic subband set in turn. When the kurtosis value is greater than 5, the signal impact is extremely strong, and the subband may contain key fault information such as blade wear. At this time, a high weight of 0.8 is given to highlight its importance. When the kurtosis value is between 3-5, the subband may have some effective features, and a weight of 0.4 is given to retain the relevant features. For subbands with a kurtosis value less than 3, the signal impact is weak, mostly interference or secondary information, and only a low weight of 0.1 is given to reduce its impact on the result. According to the weight assignment method, a larger weight is given to a subband with a high kurtosis value, and a smaller weight is given to a subband with a low kurtosis value. The kurtosis value of each subband is multiplied by the corresponding weight, and all products are added up to obtain the sum. The sum is normalized so that the sum is between 0 and 1. The normalized sum is the pulse sensitivity coefficient.

[0106] The five frequency bands with the highest pulse sensitivity coefficients are selected for analysis to generate a residual eigenvector. Specifically, the following steps are performed: sorting the subbands in the characteristic subband set from large to small according to the pulse sensitivity coefficients, selecting the five frequency bands with the highest values, calculating the energy entropy and normalized kurtosis value of each selected frequency band, and arranging and combining the energy entropy and normalized kurtosis values ​​of these five frequency bands in sequence to form a residual eigenvector.

[0107] By comparing the vibration signal with the soil hardness data The five-layer wavelet packet decomposition and integration were used to construct a feature subband set, which effectively separated the interference of soil hardness on the vibration signal. The signal was mapped to different frequency bands according to its frequency characteristics, so that the characteristics of the two types of data could be clearly presented, avoiding the influence of vibration spectrum distortion caused by soil hardness changes on the analysis results. At the same time, different weights were assigned according to the kurtosis value and the pulse sensitivity coefficient was calculated, highlighting the importance of the subband containing key fault information, reducing the influence of interference and secondary information, and achieving accurate processing of complex signals.

[0108] By selecting the five frequency bands with the highest pulse sensitivity coefficients and generating residual eigenvectors by calculating their energy entropy and normalized kurtosis values, we can precisely focus on the key features related to blade wear. Compared with the traditional method that only relies on spectral analysis of vibration data, this method can effectively eliminate the interference of soil hardness and obtain characteristic information that can more truly reflect the blade wear condition, significantly improving the accuracy of agricultural equipment status monitoring. Accurate monitoring results help to detect blade wear problems in a timely manner.

[0109] In one case of this embodiment, the residual feature vector is fused with the ambient temperature and humidity data to generate a coupling wear index, including:

[0110] The pre-processed ambient temperature and humidity data are quantified to obtain the environmental modulation factor, specifically including: calculating the mean and standard deviation of the pre-processed ambient temperature and humidity data respectively, subtracting the mean from each temperature value and dividing it by the standard deviation to obtain standardized temperature data, using the same method to standardize the humidity data so that the temperature and humidity data are in the same dimension, setting the temperature weight to 0.6 and the humidity weight to 0.4 based on the characteristic that temperature has a greater impact on blade wear, multiplying the standardized temperature and humidity data by the corresponding weights and adding them together to obtain a comprehensive impact value, and finally mapping the comprehensive value to the range of 0 to 1 using the Sigmoid function. The generated value is the environmental modulation factor;

[0111] Analyze the residual eigenvector to generate the wear sensitivity vector;

[0112] The environmental modulation factor is integrated with the wear sensitivity vector to obtain a preliminary coupling index. Specifically, the following steps are performed: using the environmental modulation factor as a weight coefficient, performing a weighted operation on each dimensional eigenvalue of the wear sensitivity vector one by one (i.e., multiplying each eigenvalue by the environmental modulation factor), so that the impact of environmental factors on wear is integrated into the characteristics of each dimension. All weighted dimensional eigenvalues ​​are then accumulated and summed up. The summation result is the preliminary coupling index, which can quantify the degree of coupling between environmental factors and blade wear characteristics.

[0113] The preliminary coupling index is calibrated based on historical fault data to generate a coupling wear index, specifically including: screening records consistent with the current monitoring conditions from historical fault data, extracting vibration characteristics (kurtosis value, energy entropy), environmental parameters (temperature and humidity) and corresponding fault types (blade cracking, bearing wear), building a feature library containing the mapping relationship between feature vectors and fault modes, calculating the similarity between the current residual feature vector and the historical data in the feature library, screening out historical cases with a similarity exceeding the preset threshold (0.7), and then statistically analyzing the frequency of occurrence of the selected historical cases by fault type to generate the probability distribution of different faults in the current state. At the same time, the weights of each parameter in the preliminary coupling index are dynamically adjusted according to the fault probability distribution. Specifically, the fault probability interval is first divided. If the probability is lower than 30%, the weight is not adjusted. If the probability is above 30%, the weight is not adjusted. When the probability is greater than 60%, the weight of the corresponding major parameter will be increased by 0.1, and the weights of other parameters will be reduced in proportion. When the probability is greater than 60%, the weight of the corresponding major parameter will be increased by 0.2, and the weights of other parameters will be reduced in proportion accordingly. After adjustment, ensure that the sum of the weights of all parameters is 1. Multiply the adjusted weight with the preliminary coupling index to obtain the product result. After normalizing the product result, it is the coupling wear index. The coupling wear index can reflect both the current wear characteristics and the probabilistic influence of historical failure modes, thereby quantifying the actual wear state of the blade.

[0114] By quantifying the ambient temperature and humidity data to generate an environmental modulation factor, the interference of environmental factors such as soil hardness on the blade vibration data can be effectively eliminated. The temperature and humidity data are first standardized and dimensionalized, and weights are scientifically assigned based on the characteristic that temperature has a greater impact on blade wear. Then, through Sigmoid function mapping, the environmental factors are used in subsequent calculations in a standardized numerical form. During the fusion process, the environmental modulation factor is used as a weight coefficient to perform weighted operations on the wear-sensitive vector, accurately integrating the influence of environmental factors into each dimension of the blade wear characteristics, avoiding errors caused by vibration spectrum distortion due to changes in soil hardness, improving the accuracy of blade wear monitoring, and thus optimizing the status monitoring effect of agricultural equipment.

[0115] By screening historical cases through similarity calculation and dynamically adjusting parameter weights based on the fault probability distribution, the current wear characteristics are combined with the probabilistic impact of historical failure modes. This dynamic calibration method can not only reflect the current actual wear status of the blade, but also predict the possibility of different faults in advance based on historical experience, and better monitor the status of agricultural equipment.

[0116] In one case of this embodiment, the residual feature vector is analyzed to generate a wear sensitivity vector, including:

[0117] The five frequency bands with the highest pulse sensitivity coefficients in the residual eigenvector are calculated separately to obtain the energy proportion of each core frequency band, specifically including: converting the residual eigenvector from the time domain to the frequency domain through fast Fourier transform to obtain its power spectral density, screening out the five frequency bands with the highest pulse sensitivity coefficients and determining them as core frequency bands, for each core frequency band, accumulating the power of all frequency components in the frequency band one by one to obtain the energy of the frequency band, adding the energies of the five core frequency bands to obtain the total energy, and dividing the energy of each core frequency band by the total energy to obtain the energy proportion of each core frequency band;

[0118] Extract the peak sequence of the vibration signal in the core frequency band to obtain the peak fluctuation characteristics, specifically including: reconstructing the vibration signal in the core frequency band in the time domain to obtain discrete time series data, calculating the mean and standard deviation of the time series data of the vibration signal in the core frequency band, using the mean plus 3 times the standard deviation as the threshold, finding the points in the vibration signal that are greater than the threshold and the data on both the left and right sides are less than the threshold, these points are the peak points, arranging all the peak points in chronological order to form a peak sequence, calculating the difference between adjacent peaks in the peak sequence, obtaining the peak variation amplitude, counting the number of peaks in a unit time (1 second) to obtain the peak frequency, and arranging and combining the mean and standard deviation of the peak variation amplitude with the mean and variance of the peak frequency in order to obtain the peak fluctuation characteristics;

[0119] The phase difference between the core frequency bands is analyzed and a phase difference matrix is ​​constructed, specifically including: performing Hilbert transform on the vibration signals of the five core frequency bands respectively, converting the time domain signals into analytical signals, which contain the amplitude and phase information of the original signal. The instantaneous phase sequence of each frequency band signal is obtained by calculating the inverse tangent function of the ratio of the imaginary part to the real part of the analytical signal. For any two core frequency bands, the difference in their instantaneous phase is calculated point by point to form a phase difference time series. The phase difference time series is averaged in the time domain to eliminate random noise interference and obtain a stable phase difference value. This process is repeated to calculate the phase difference between all the core frequency bands and construct the five core frequency bands. 5, the main diagonal elements in the matrix are set to 0, indicating that the phase difference of the same frequency band itself is 0, and the remaining elements represent the phase relationship between the corresponding frequency bands. The phase difference matrix can reflect the phase coordination characteristics between the vibration signals of each core frequency band, which is convenient for understanding the inherent relationship between the equipment operation status and fault characteristics;

[0120] Decomposing the phase difference matrix to obtain a phase feature set, specifically comprising: decomposing the phase difference matrix by singular value decomposition to decompose it into a left singular matrix, a singular value matrix, and a right singular matrix, retaining the first three largest singular values ​​and their corresponding singular vectors, extracting the column vectors corresponding to the first three singular values ​​in the left singular matrix to form a feature vector set, and reducing the dimension of the feature vector set by principal component analysis to obtain a low-dimensional phase feature set;

[0121] The energy proportion, peak fluctuation characteristics and phase feature set of each core frequency band are integrated to generate a wear-sensitive vector. Specifically, the energy proportion of each core frequency band, the various features of the peak fluctuation characteristics, and the elements in the phase feature set are arranged in sequence into a long vector, and the long vector is normalized so that its values ​​are at the same order of magnitude. The normalized long vector is obtained, which is the wear-sensitive vector.

[0122] Wear-sensitive vectors are generated through multi-dimensional feature extraction, which effectively overcomes the impact of environmental interference on blade wear monitoring. In frequency domain analysis, the core frequency bands with high pulse sensitivity coefficients are screened to calculate the energy share, focusing on frequency components closely related to wear. After time domain reconstruction, peak fluctuation characteristics are extracted to capture the instantaneous change characteristics of vibration signals. The combination of the two can analyze vibration signals from both frequency and time domain perspectives, avoiding spectrum distortion interference caused by environmental factors such as changes in soil hardness, and accurately understanding the vibration signal characteristics caused by blade wear, so that the monitoring data can more truly reflect the actual wear status of the blade and improve the accuracy of agricultural equipment status monitoring.

[0123] The phase feature set is obtained through Hilbert transform and singular value decomposition, and the phase synergy characteristics between the vibration signals of each core frequency band are explored. The equipment operation status and fault characteristics are revealed from the level of intrinsic correlation of the signal. The energy proportion and peak fluctuation characteristics are integrated with the phase feature set to construct a wear-sensitive vector containing multi-source information. Compared with single spectrum analysis, it can describe the blade wear characteristics more comprehensively and detailedly.

[0124] In one case of this embodiment, generating a dynamic state threshold based on historical fault data and ambient temperature and humidity data includes:

[0125] Cluster analysis is performed on historical fault data to divide the fault mode categories under different environmental temperature and humidity conditions. Specifically, the following steps are taken: for the pre-processed historical fault data, the Z-score standardization method is used to uniformly map the environmental temperature and humidity, vibration characteristics (kurtosis value, energy entropy) and fault type data to the same dimensional space; the standardized data is reduced in dimension by principal component analysis (PCA) to extract the principal components that can reflect the main characteristics of the data; the K-means clustering algorithm is used to cluster the reduced-dimensional data; the optimal number of clusters is determined by the elbow rule: the intra-cluster sum of squares under different K values ​​is calculated, and the K value corresponding to the inflection point of the intra-cluster sum of squares curve is selected as the optimal number of categories; the environmental temperature and humidity data and the fault feature vector are merged into a multidimensional sample; the Euclidean distance is used to obtain the similarity between samples; the samples are divided into different clusters according to the distance; after clustering, the following steps are performed for each cluster: the environmental temperature and humidity data of all samples in the cluster are extracted; the minimum, maximum, mean and standard deviation of the temperature are calculated; the temperature distribution range (25 30℃ 2℃), and the humidity distribution range is obtained by the same logic (60% 70% relative humidity 5%), and count the frequency of occurrence of the corresponding fault types (blade cracking, bearing wear) of samples within the cluster. For example, if blade cracking accounts for 70% and bearing wear accounts for 30% in a cluster, then the main fault mode of this cluster is determined to be blade cracking. By traversing all clusters, a mapping table of "temperature and humidity combination interval → main fault mode" is established. For example, the first temperature and humidity interval (20-25°C, 50-60% relative humidity) corresponds to the first fault mode dominated by bearing wear, and the second temperature and humidity interval (30-35°C, 70-80% relative humidity) corresponds to the second fault mode dominated by blade cracking. And so on, thus classifying the fault mode categories under different environmental temperature and humidity conditions;

[0126] The ambient temperature and humidity range corresponding to each fault mode category and the range of equipment status parameters when the fault occurs are counted to generate a fault feature parameter set. Specifically, for each fault mode category, the ambient temperature and humidity data and equipment status parameters (vibration kurtosis value, energy entropy, blade speed, load torque) of all samples in the corresponding cluster are extracted. For the temperature and humidity data, the minimum and maximum values ​​are calculated to determine the distribution range. For the equipment status parameters, the mean, standard deviation, and fluctuation range of each data are counted. For each fault mode category (mainly blade cracking and bearing wear), the previously determined temperature and humidity range is combined with the statistical range of the equipment status parameters under this category. A table is created according to the fault mode category, with each row corresponding to a fault mode and the corresponding temperature and humidity range and equipment parameter range recorded in the columns, ultimately forming a structured fault feature parameter set.

[0127] The basic threshold is obtained by calculating the historical fault frequency based on the fault characteristic parameter set. Specifically, for each fault mode category, the number of faults that occurred within the corresponding temperature and humidity range and equipment status parameter range in the historical data is counted, and then divided by the total number of historical data samples to obtain the fault frequency per unit sample, which is the basic threshold.

[0128] Dynamically modify the basic threshold according to the ambient temperature and humidity data to generate a dynamic state threshold, specifically including: comparing the current ambient temperature and humidity data with the mapping table of temperature and humidity combination interval → main fault mode, finding the corresponding fault mode category, if the current temperature and humidity are in the center of the corresponding interval (the temperature and humidity fluctuation range is within the mean 1 3 standard deviations), set the adjustment coefficient to 1, because the area is in a stable state and the failure probability has no significant change. If the current temperature and humidity are at the edge of the interval (the temperature and humidity fluctuation range exceeds the mean 2 3 standard deviations), the adjustment coefficient is set to 1.3 to increase the threshold sensitivity. Because the environment in the edge area changes greatly, the risk of failure increases. If the current temperature and humidity are between the two, the adjustment coefficient is 1.1 to balance risk and stability. The adjustment coefficient is then multiplied by the basic threshold to obtain the dynamic state threshold.

[0129] Through cluster analysis and mapping table construction, accurate identification of failure modes under different ambient temperature and humidity conditions was achieved. After standardizing and reducing the dimensionality of historical failure data, the K-means clustering algorithm was used to classify failure modes. The optimal number of clusters was determined using the elbow rule to ensure scientific classification. Using Euclidean distance to measure sample similarity, a mapping table was established from "temperature and humidity combination intervals to major failure modes," closely linking environmental factors to failure characteristics. This process effectively eliminated the influence of environmental interference such as soil hardness on vibration data, revealing the inherent connection between ambient temperature and humidity and failure modes from historical data, facilitating more accurate monitoring of agricultural equipment status.

[0130] By counting the ambient temperature and humidity intervals and equipment status parameter ranges corresponding to the fault mode categories, the basic threshold is calculated, the frequency of faults is quantified, and the threshold is dynamically adjusted according to the interval position of the current ambient temperature and humidity data. For example, the threshold sensitivity is increased at the edge of the interval where the temperature and humidity fluctuate greatly, while the central area remains stable. This dynamic correction mechanism enables the status threshold to adapt to environmental changes in real time, avoiding misjudgment or missed judgment due to environmental factors. Compared with the fixed threshold monitoring method, it can more accurately reflect the actual operating status of agricultural equipment in complex environments.

[0131] In one case of this embodiment, the coupling wear index is compared with the dynamic state threshold to generate a state instruction, including:

[0132] Calculating the coupling wear index and the dynamic state threshold to generate a dynamic degradation degree, specifically including: subtracting the dynamic state threshold from the coupling wear index to generate the dynamic degradation degree;

[0133] When the dynamic degradation When the value is 0, it indicates that the equipment status is not safe enough and a replacement instruction is issued;

[0134] When the dynamic degradation When it is 0, it indicates that the device status is in a safe range and no command is issued;

[0135] The dynamic degradation degree is corrected in real time based on the ambient temperature and humidity data and historical failure mode categories.

[0136] By comparing the coupling wear index with the dynamic state threshold and calculating the dynamic degradation degree, accurate judgment and intelligent decision-making of the agricultural equipment state are achieved. The dynamic degradation degree is generated by subtracting the coupling wear index from the dynamic state threshold, which can intuitively determine whether the equipment state is safe, avoiding the misjudgment problem caused by environmental interference in traditional monitoring methods. At the same time, the dynamic degradation degree is corrected in real time in combination with environmental temperature and humidity data and historical failure mode categories, fully considering the impact of environmental factors on the equipment state, so that the judgment result is more in line with the actual operating conditions. When the equipment state is unsafe, a replacement instruction is issued in time, and no intervention is made when the state is safe. It can effectively avoid the waste of resources caused by excessive maintenance and improve the intelligence of agricultural equipment state monitoring.

[0137] In one case of this embodiment, the dynamic degradation degree is corrected in real time based on the ambient temperature and humidity data and the historical failure mode categories, including:

[0138] According to the real-time environmental temperature and humidity data, the nearest fault mode category in the historical fault cluster is matched to generate a mode deviation factor, which specifically includes: for each fault mode category obtained by clustering the historical fault data (the temperature and humidity interval corresponding to blade cracking and bearing wear), the temperature and humidity mean vector of the center point of each category (i.e., the cluster center) is calculated, and the covariance matrix of the temperature and humidity data under this category is calculated. The pre-processed real-time temperature and humidity data is compared with the cluster center vector of each fault mode category, and the Mahalanobis distance from the real-time temperature and humidity data to each cluster center is calculated. Specifically, the Mahalanobis distance between the real-time data and the cluster center is calculated. The difference vector between the real-time data and the cluster center is then weighted with the inverse matrix of the covariance matrix to obtain a distance value that reflects the degree of deviation from the actual data distribution, which is the Mahalanobis distance. The category with the smallest Mahalanobis distance is selected as the nearest fault mode category. The distance from the real-time data to the nearest category cluster center is converted using the Sigmoid function. The mapped value is the mode deviation factor. The closer the mode deviation factor is to 1, the greater the difference between the current temperature and humidity environment and the typical environment of the fault mode category. The closer the mode deviation factor is to 0, the higher the environmental matching degree is, and the more likely this type of fault is to be triggered.

[0139] Calculate the current fault mode category to obtain the historical degradation slope mean, specifically including: for the currently matched fault mode category, filter all fault records belonging to this category in the historical data, extract the coupling wear index and corresponding timestamp of each record, sort the coupling wear index by time, calculate the ratio of the index difference between adjacent time points to the time interval (i.e., degradation slope), and calculate the average of all degradation slope values ​​to obtain the historical degradation slope mean;

[0140] The dynamic degradation degree is corrected in real time according to the mode deviation factor and the historical degradation slope average, including: Mahalanobis distance from the historical fault cluster center Divide by the maximum Mahalanobis distance of the historical fault cluster center , get the deviation , multiply the deviation by the mode deviation correction factor Then add the unit base value 1 to get the correction factor , multiply the correction factor by the dynamic degradation degree before correction , get the environmental correction term ; The real-time degradation slope Divide by the historical degradation slope benchmark value of similar equipment , and obtain the normalized degradation slope , calculate the current load change ,when 0 o'clock The value is 1. 0 o'clock The value is 1. Multiply the normalized degradation slope by the slope attenuation coefficient Then multiply it by the load change direction function , and obtain the slope correction term ; Subtract the slope correction term from the environmental correction term to obtain the corrected dynamic degradation degree , when applied specifically, it can be achieved through the following calculation formula, for example:

[0141] ;

[0142] Where, represents the corrected dynamic degradation degree, represents the dynamic degradation degree before correction, represents the mode deviation correction coefficient, Indicates the current environment parameters Mahalanobis distance from the historical fault cluster center, Indicates the maximum Mahalanobis distance of the historical fault cluster center, represents the slope attenuation coefficient, represents the normalized degradation slope, Indicates the historical degradation slope benchmark value of similar equipment. Represents the load change direction function, when 0:00 The value is 1, when 0:00 The value is 1;

[0143] The value range is as follows:

[0144] When the soil hardness deviates significantly from the historical working conditions, The value range is 0.4-0.8. Because the blade vibration spectrum is distorted due to the sudden change of soil hardness, it is necessary to amplify the environmental deviation weight to offset the spectrum interference. The value can improve the degradation estimation, solve the problem of monitoring signal distortion caused by soil mutation, and ensure the fault sensitivity under strong interference. The value is too high;

[0145] When the soil hardness deviates moderately from the historical working conditions, The value range is 0.2-0.3. Because the change of soil hardness causes the discernible spectrum to shift, it is necessary to balance the historical data and real-time shift in the linear correction term to avoid false alarms caused by over-correction and compensate for the implicit influence of soil physical property changes on blade stress distribution. The value is moderate;

[0146] When the soil hardness is close to the historical working conditions, The value range is 0-0.1. Because the soil conditions are stable, the spectrum distortion is slight, and the spectrum characteristics under the steady state of the soil are consistent with the historical clustering, excessive correction will introduce noise, and only basic compensation is needed. The value is too low.

[0147] The value range is as follows:

[0148] When the soil hardness changes drastically and causes the spectrum to fluctuate violently, The value range is 0.1-0.3. To prevent high-frequency interference from distorting the degradation slope, the effect of noise on the dynamic degradation degree needs to be significantly attenuated. The value is too low;

[0149] When the soil hardness changes gradually and causes the spectrum to fluctuate in the medium amplitude, The value range is 0.4-0.6. Because it is necessary to balance the environmental interference and the real wear signal, only moderate attenuation can be performed to avoid missing the real degradation trend. The value is moderate;

[0150] When the soil hardness changes slowly, resulting in small fluctuations in the spectrum, The value range is 0.7-1. Because the soil interference is weak, the original degradation slope characteristics need to be retained to reflect the actual wear situation. The value is too high.

[0151] By dynamically matching ambient temperature and humidity with historical fault patterns, the problem of interference in the condition monitoring of agricultural equipment under different soil hardnesses can be accurately solved. By using Mahalanobis distance to match the nearest fault mode and generate a pattern deviation factor, the difference between the current environment and the typical fault environment can be quantified. When a sudden change in soil hardness causes vibration spectrum distortion, the environmental weight is amplified by a relatively high pattern deviation correction coefficient to offset spectrum interference, improve the accuracy of degradation estimation, and ensure the sensitivity of fault monitoring under strong interference.

[0152] By dynamically adjusting the correction term based on the slope attenuation coefficient and the load change direction function, when the soil hardness gradually changes or stabilizes, the historical data and real-time offset are balanced to avoid false alarms caused by over-correction. At the same time, the implicit influence of soil physical properties on the blade stress distribution is compensated, and real-time correction of dynamic degradation degree can be achieved. It can adapt to different soil working conditions, reduce the interference of soil hardness changes on vibration data, make the spectrum analysis results closer to the actual wear of the blade, and effectively improve the accuracy of agricultural equipment status monitoring.

[0153] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. An agricultural equipment status monitoring system based on the Internet of Things, characterized in that: include: The acquisition unit is used to collect vibration data, blade speed and torque data of the monitored object, as well as soil hardness data and ambient temperature and humidity data of the monitoring site in real time through the Internet of Things sensor network. It also obtains historical fault data of agricultural equipment. The monitored object is the harvester blade and the monitoring site is dry land. The analysis unit is used to perform synchronous working condition analysis on vibration data, soil hardness data, blade speed and torque data to generate physical compensation factors, including: Perform time stamp alignment on pre-processed vibration data, soil hardness data, blade speed, and torque data to generate unified synchronization timing parameters; According to the synchronous timing parameters, the physical characteristics of the speed data, torque data and soil hardness data are extracted and integrated to obtain the composite working condition index; Map the composite working condition indicators to generate physical compensation factors, including: Decouple the composite working condition indicators to obtain the dynamic working condition vector related only to the instantaneous load of the blade; Perform time integration on the dynamic working condition vector to obtain the cumulative fatigue coefficient of the blade under the current load; Analyze soil hardness data to derive dynamic weights; Based on the dynamic weight, the dynamic working condition vector and the cumulative fatigue coefficient are weighted and fused to generate the physical compensation factor; The stripping unit is used to perform interference stripping on the vibration data according to the physical compensation factor and generate a residual feature vector, including: Perform nonlinear transformation on the physical compensation factor and combine it with the torque data and soil hardness data to obtain the dynamic filter coefficient; Filtering the vibration data based on a dynamic filter coefficient to generate a stripped vibration signal; Wavelet packet analysis of the stripped vibration signal Kurtosis joint analysis to generate residual eigenvectors, including: Decompose the stripped vibration signal and soil hardness data to generate a feature subband set; After calculating the kurtosis value of the characteristic subband set and performing nonlinear weighting, the pulse sensitivity coefficient is obtained; The five frequency bands with the highest pulse sensitivity coefficients are selected for analysis to generate residual eigenvectors; A fusion unit is used to fuse the residual feature vector with the ambient temperature and humidity data to generate a coupling wear index; The decision unit is used to generate a dynamic state threshold based on historical fault data and ambient temperature and humidity data, compare the coupling wear index with the dynamic state threshold, and generate a state instruction.

2. The agricultural equipment status monitoring system based on the Internet of Things according to claim 1 is characterized in that: The residual eigenvector is fused with the ambient temperature and humidity data to generate the coupled wear index, including: Quantify the pre-processed ambient temperature and humidity data to obtain the environmental modulation factor; Analyze the residual eigenvector to generate the wear sensitivity vector; The environmental modulation factor is fused with the wear sensitivity vector to obtain the preliminary coupling index; The preliminary coupling index is calibrated based on historical failure data to generate a coupling wear index.

3. The agricultural equipment status monitoring system based on the Internet of Things according to claim 2 is characterized in that: Analyze the residual eigenvector to generate the wear sensitivity vector, including: Calculate the five frequency bands with the highest pulse sensitivity coefficients in the residual eigenvector separately to obtain the energy proportion of each core frequency band; Extract the peak sequence of the vibration signal in the core frequency band to obtain the peak fluctuation characteristics; Analyze the phase difference between core frequency bands and construct a phase difference matrix; Decompose the phase difference matrix to obtain the phase feature set; The energy proportion, peak fluctuation characteristics and phase feature set of each core frequency band are fused to generate a wear sensitivity vector.

4. The agricultural equipment status monitoring system based on the Internet of Things according to claim 1 is characterized in that: Generate dynamic status thresholds based on historical fault data and ambient temperature and humidity data, including: Perform cluster analysis on historical fault data to classify fault mode categories under different environmental temperature and humidity conditions; Count the ambient temperature and humidity ranges corresponding to each fault mode category and the equipment status parameter range when the fault occurs, and generate a fault characteristic parameter set; Calculate the frequency of historical faults based on the fault characteristic parameter set to obtain the basic threshold; The basic threshold is dynamically modified according to the ambient temperature and humidity data to generate a dynamic state threshold.

5. The agricultural equipment status monitoring system based on the Internet of Things according to claim 4 is characterized in that: Compare the coupling wear index with the dynamic state threshold to generate state instructions, including: The coupling wear index and the dynamic state threshold are calculated to generate the dynamic degradation degree; When the dynamic degradation When it is 0, it indicates that the equipment status is not safe enough and a replacement instruction is issued; When the dynamic degradation When it is 0, it indicates that the device status is in a safe range and no command is issued; The dynamic degradation degree is corrected in real time based on the ambient temperature and humidity data and historical failure mode categories.

6. The agricultural equipment status monitoring system based on the Internet of Things according to claim 4 is characterized in that: Based on the ambient temperature and humidity data and historical failure mode categories, the dynamic degradation degree is corrected in real time, including: According to the real-time environmental temperature and humidity data, the nearest fault mode category in the historical fault cluster is matched to generate the mode deviation factor; Calculate the current failure mode category to obtain the mean of the historical degradation slope; The dynamic degradation degree is corrected in real time according to the mode deviation factor and the average of the historical degradation slope.

Images