A corn harvester remote operation and maintenance data collection optimization method

Abstract

The present application relates to the technical field of agricultural machinery operation and maintenance, and discloses a corn harvester remote operation and maintenance data acquisition optimization method. The method comprises the following steps: acquiring a real-time operation area distribution map of the corn harvester, and determining a core acquisition area according to the map; collecting working condition characteristics of each operation sub-area in the operation area distribution map to obtain crop density characteristics and terrain undulation characteristics corresponding to each operation sub-area; then performing fault prediction analysis on the two kinds of characteristics based on the core acquisition area to generate a static fault characteristic set; subsequently acquiring a transmissible path of a remote operation and maintenance terminal, combining each static fault characteristic value in the static fault characteristic set and the corresponding operation sub-area, and controlling the acquisition frequency of operation and maintenance data on the transmissible path; and finally monitoring the change amount of the operating parameters of the current operation sub-area in real-time data acquisition, combining the static fault characteristic value corresponding to the current operation sub-area, and dynamically calibrating the sensing direction of the acquisition terminal.

Description

Technical Field

[0001] This invention relates to the field of agricultural machinery operation and maintenance technology, specifically to a method for optimizing remote operation and maintenance data collection for corn harvesters. Background Technology

[0002] In the rapid development of agricultural mechanization, corn harvesters, as key agricultural equipment, directly impact the economic benefits of corn cultivation through their operational efficiency and stability. With the gradual application of remote operation and maintenance technology in agricultural machinery, collecting operational data from corn harvesters for fault warning and maintenance management has become an important means of improving equipment reliability. However, current methods for collecting remote operation and maintenance data from corn harvesters still face many unresolved issues.

[0003] Current data acquisition methods mostly adopt a uniform, all-area acquisition model, failing to consider the actual differences in the operating areas of corn harvesters. Corn planting areas often exhibit uneven crop density, with significant variations in crop density in some areas due to differences in plot planning or planting management. Furthermore, field terrain is not entirely flat, with some areas featuring gentle slopes, depressions, and other undulating terrain. In this context, using a uniform, all-area data acquisition approach can lead to the collection of excessive redundant data in areas with low crop density, flat terrain, and low equipment failure risk. Conversely, in core areas with high crop density, complex terrain, and high equipment failure risk, insufficient acquisition resources may prevent the acquisition of enough critical data, impacting the accuracy of failure prediction.

[0004] Existing methods lack specificity in data transmission path utilization and acquisition frequency control. The available transmission paths between remote maintenance terminals and corn harvesters are affected by factors such as field network signals and terrain obstructions, resulting in variations in transmission stability and bandwidth across different paths. However, current acquisition methods do not consider the fault characteristics of each operating sub-area of ​​the equipment, applying the same acquisition frequency to all areas. This not only easily leads to a waste of transmission resources but may also result in insufficient acquisition frequency in high-fault-risk areas, causing a failure to promptly detect abnormal changes in equipment operating parameters and delaying fault warnings.

[0005] During real-time data acquisition, existing methods lack the ability to dynamically adjust the calibration of the sensor direction of the acquisition terminal. During corn harvesting operations, crop density, terrain undulations, and other operating conditions change as the work sub-area shifts. The acquisition terminal needs to adjust its sensor direction according to the current operating conditions to ensure accurate acquisition of equipment operating parameters. However, current methods mostly use a fixed sensor direction calibration mode, which cannot be dynamically adjusted according to real-time operating conditions. This leads to reduced accuracy of the acquired operating parameter data in areas with significant changes in operating conditions due to sensor direction deviation, thus affecting the reliability of subsequent fault analysis and maintenance decisions. These problems severely restrict the efficiency and quality of remote maintenance data acquisition for corn harvesters, making it difficult to meet the needs of modern agriculture for efficient operation and maintenance management of agricultural machinery. Summary of the Invention

[0006] The purpose of this invention is to provide a method for optimizing remote operation and maintenance data collection for corn harvesters, so as to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides a method for optimizing remote operation and maintenance data collection for corn harvesters, the method comprising:

[0008] Obtain a real-time operating area distribution map of the corn harvester, and determine the core collection area based on the operating area distribution map;

[0009] Based on the operation area distribution map, the operation condition characteristics of each operation sub-region are collected to obtain the crop density characteristics and terrain undulation characteristics corresponding to each operation sub-region.

[0010] Based on the core collection area, the crop density characteristics and terrain undulation characteristics are analyzed for fault prediction, and a static fault feature set is generated.

[0011] Obtain the transmittable path of the remote operation and maintenance terminal, and control the collection frequency of operation and maintenance data on the transmittable path based on each static fault feature value in the static fault feature set and its corresponding operation sub-region.

[0012] In real-time data acquisition, the changes in operating parameters of the current working sub-region are monitored, and the sensing direction of the acquisition terminal is dynamically calibrated in combination with the static fault characteristic value corresponding to the current working sub-region.

[0013] Preferably, the step of collecting working condition characteristics based on each sub-region of the working area distribution map to obtain the crop density characteristics and terrain undulation characteristics corresponding to each sub-region includes:

[0014] Based on the area boundary coordinates in the aforementioned work area distribution map, grid division is performed to generate multiple work sub-areas;

[0015] Obtain the harvesting direction of travel, select the starting sub-region based on the harvesting direction of travel to collect crop density characteristics, and collect crop density characteristics of subsequent sub-regions along the direction of travel in sequence;

[0016] Obtain the ground slope vector, reselect the starting sub-region based on the ground slope vector to collect terrain undulation features, and collect the terrain undulation features of the subsequent sub-regions in sequence along the slope change direction.

[0017] Preferably, the fault prediction analysis based on the crop density characteristics and terrain undulation characteristics of the core acquisition area includes:

[0018] Determine the depth of the associated region based on the core acquisition area;

[0019] Extract the crop density and terrain undulation features of the first sub-region. Based on the location coordinates of the first sub-region and the depth of the associated region, select adjacent sub-regions for fault feature association analysis. The fault feature association analysis adopts a multiple linear regression model to generate the first static fault feature value.

[0020] Static fault feature values ​​for each sub-region of the operation are generated sequentially using the same method, and then integrated into a static fault feature set based on the operation area distribution map.

[0021] Preferably, the collection frequency of the control and maintenance data on the transmittable path includes:

[0022] Extract each static fault feature value and its associated operation sub-region from the static fault feature set;

[0023] Identify all path segments of the transmittable path and their corresponding adjacent operational sub-regions;

[0024] The static fault characteristic values ​​of adjacent work sub-regions corresponding to each path segment are weighted and superimposed to generate the data acquisition priority of each path segment.

[0025] Configure the data sampling interval of the sensors on the corresponding path segment based on the data acquisition priority.

[0026] Preferably, the dynamic calibration of the sensing direction of the acquisition terminal includes:

[0027] Calculate the deviation between the changes in operating parameters of the current sub-region and the static fault characteristic values;

[0028] When the deviation exceeds the preset tolerance threshold, the calibration angle is determined based on the equipment vibration spectrum of the current operating sub-region.

[0029] Based on the calibration angle, adjust the pitch angle of the image acquisition device and the orientation of the vibration sensor in the corresponding work sub-area.

[0030] Preferably, after generating the static fault feature set, the method further includes:

[0031] Statistical analysis of the dispersion of static fault characteristic values ​​in each sub-region of operation;

[0032] When the distribution dispersion is lower than a preset dispersion threshold, the operation and maintenance data collection tasks of adjacent operation sub-regions are merged.

[0033] Preferably, the frequency of collecting the control and maintenance data further includes:

[0034] Monitor the bandwidth fluctuation of the transmittable path;

[0035] When the bandwidth fluctuation exceeds the preset fluctuation threshold, the original data resolution of low-priority path segments is dynamically compressed according to the data acquisition priority.

[0036] Preferably, obtaining the change in the operating parameters includes:

[0037] Real-time acquisition of engine speed fluctuation, cutting table hydraulic pressure offset, and travel mechanism slippage rate;

[0038] The speed fluctuation value, hydraulic pressure offset and slippage rate are normalized and weighted to generate a comprehensive change in operating parameters.

[0039] Preferably, the dynamic calibration is followed by:

[0040] Record the calibration count and historical deviation data for each sub-region of operation;

[0041] When the calibration frequency of a specific sub-region exceeds a preset frequency threshold, the feature weight coefficient of that sub-region in the static fault feature set is increased.

[0042] Preferably, the method further includes:

[0043] An environmental fitness model is constructed based on historical calibration data, and the random forest algorithm is used to train the environmental fitness model.

[0044] When a new operational area distribution map is accessed, the initial acquisition parameters for each sub-region are pre-configured according to the environmental adaptability model.

[0045] Compared with the prior art, the beneficial effects of the present invention are:

[0046] This optimized method for remote operation and maintenance data acquisition of corn harvesters effectively solves the problems of data redundancy and missing key data in the traditional uniform acquisition mode by precisely dividing the work area and identifying core areas. In practical applications, determining the core acquisition area based on the real-time work area distribution map allows data acquisition resources to be more accurately focused on areas with high crop density, large terrain undulations, and high equipment failure risk. This avoids investing too many acquisition resources in areas with low failure risk, reducing the generation of redundant data, while ensuring that sufficient key data can be obtained in the core area, providing a more comprehensive and effective data foundation for subsequent failure prediction and analysis. This targeted area division and resource allocation method not only improves the effectiveness of data acquisition but also reduces unnecessary data storage and processing costs, making the data acquisition process more economical and efficient.

[0047] Regarding data transmission and acquisition frequency control, this method combines the transmittable paths of the remote maintenance terminal with the correlation between the static fault characteristic values ​​in the static fault feature set and the corresponding work sub-regions to differentiate the acquisition frequency, significantly improving the utilization efficiency of transmission resources and the timeliness of fault early warning. Since the static fault characteristic values ​​differ across different work sub-regions, corresponding to varying degrees of equipment fault risk, adjusting the acquisition frequency based on the fault characteristic values ​​allows for increased acquisition frequency in high-fault-risk work sub-regions, ensuring timely capture of subtle changes in equipment operating parameters and timely detection of potential faults; while appropriately reducing the acquisition frequency in low-fault-risk regions avoids wasting transmission resources. Simultaneously, controlling the acquisition frequency based on the actual conditions of the transmittable paths fully utilizes paths with high transmission stability and sufficient bandwidth, reducing data loss or delays caused by transmission problems, further ensuring the reliability of data acquisition and transmission, and providing strong support for the remote maintenance terminal to obtain accurate data in a timely manner.

[0048] In the real-time data acquisition sensor direction calibration stage, this method significantly improves the accuracy of data acquisition by monitoring changes in operating parameters in the current working sub-region and combining this with the corresponding static fault characteristic value for dynamic calibration. During corn harvesting operations, differences in crop density, terrain undulations, and other working conditions in different working sub-regions can cause changes in the equipment's operating status. If the sensor direction of the acquisition terminal remains fixed, data acquisition deviations are likely to occur. This method, however, can determine the difference between the current working condition and the static fault characteristic value based on the real-time monitored changes in operating parameters, and then dynamically adjust the sensor direction, ensuring that the acquisition terminal always acquires equipment operating parameters at the optimal angle, reducing data errors caused by sensor direction deviations. This dynamic calibration method ensures that the acquired operating parameter data accurately reflects the actual operating status of the equipment under different working conditions, providing more reliable data support for subsequent fault analysis and maintenance decisions, and helping to improve the accuracy and effectiveness of remote maintenance.

[0049] From an overall operation and maintenance management perspective, this method optimizes each stage of data collection, forming a complete optimization system from area division and frequency control to sensor calibration. This significantly improves the overall quality and efficiency of remote operation and maintenance data collection for corn harvesters. In actual agricultural production scenarios, this method helps maintenance personnel to grasp the equipment's operating status more quickly and accurately, identify potential fault risks in advance, formulate timely maintenance plans, reduce operational delays caused by equipment downtime due to malfunctions, improve the operating efficiency and service life of corn harvesters, and thus provide strong support for the efficient development of agricultural production, meeting the needs of modern agricultural mechanization and intelligent development. Attached Figure Description

[0050] Figure 1 This is a schematic diagram illustrating the working principle of the remote operation and maintenance data acquisition and optimization method for corn harvesters described in this invention.

[0051] Figure 2 A flowchart for collecting operational characteristics of each sub-region;

[0052] Figure 3 A flowchart for controlling the frequency of operation and maintenance data collection;

[0053] Figure 4 This is a flowchart for dynamic calibration of the sensing direction of the data acquisition terminal. Detailed Implementation

[0054] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0055] Please see Figure 1 This invention provides a method for optimizing remote operation and maintenance data collection for corn harvesters, the method comprising:

[0056] By acquiring a real-time operational area distribution map of the corn harvester, a core data collection area is determined based on this map. Operating condition characteristics are collected for each sub-region in the operational area distribution map, yielding crop density and terrain undulation characteristics for each sub-region. Fault prediction analysis is performed on the crop density and terrain undulation characteristics based on the core data collection area, generating a static fault feature set. The transmittable path of the remote maintenance terminal is obtained, and the data collection frequency along the transmittable path is controlled based on the static fault feature values ​​in the static fault feature set and their corresponding operational sub-regions. During real-time data acquisition, the changes in operating parameters of the current operational sub-region are monitored, and the sensing direction of the data collection terminal is dynamically calibrated based on the corresponding static fault feature values ​​of the current operational sub-region.

[0057] Example 1: See Figure 2 The specific implementation process of a remote operation and maintenance data acquisition optimization method for corn harvesters is as follows. This system integrates multi-source sensor data and geospatial information to achieve refined monitoring and predictive maintenance of the corn harvester's working status. The implementation process begins with acquiring a high-precision distribution map of the operating area. This map typically comes from pre-stored path planning files of the agricultural machinery's automatic driving system or real-time GPS trajectory data. The map data uses the WGS-84 coordinate system, is stored in latitude and longitude format, and is visualized using GIS software. Based on this distribution map, the system identifies core acquisition areas using density clustering algorithms (such as DBSCAN). These areas typically correspond to continuous large-area operating blocks or areas with a high incidence of historical faults.

[0058] After the core data acquisition area is determined, the system enters the operational feature acquisition phase. The operational area distribution map is loaded into the processing unit, where grid division is performed first. Based on the map scale and the accuracy requirements of agricultural machinery operations, the entire area is divided into several standard-sized square sub-regions. For example, for a map covering a 1 square kilometer operational area, a 100m × 100m grid is used to generate 100 operational sub-regions. Each sub-region is assigned a unique code, using a "row number-column number" format, and the latitude and longitude coordinates of its four corner points are recorded.

[0059] The harvesting direction is acquired by reading the heading angle data from the agricultural machinery navigation system via the CAN bus, or by analyzing the movement vector of GPS trajectory points. The system selects the first sub-region along the direction of travel as the starting acquisition point. Crop density characteristics are acquired using a multispectral camera and a lidar sensor mounted on the harvester. The multispectral camera acquires images of the crop canopy, and the number of plants per unit area is calculated using image processing algorithms; the lidar scans the crop height distribution, and the vegetation cover density is calculated by combining it with point cloud data. Data is sampled at a rate of 10 frames per second, and the data acquisition duration for each sub-region is the entire time the agricultural machinery passes through that region. The acquired values ​​are calculated as a time-weighted average.

[0060] After completing the initial sub-region data collection, the system processes subsequent sub-regions sequentially along the direction of travel. The direction of travel is dynamically updated using real-time positioning data to ensure that the data collection order matches the actual movement path of the agricultural machinery. The crop density characteristic data of each sub-region is stored in conjunction with its region code.

[0061] The acquisition of terrain undulation features was conducted independently. The system acquired triaxial acceleration and angular velocity data through an IMU (Inertial Measurement Unit) mounted on the agricultural machinery chassis, which was then processed into a ground slope vector using a sensor fusion algorithm. The slope vector contains two parameters: slope value and direction angle. Based on the vector direction, the system reselected the starting sub-region for acquisition, typically located at the point of most significant slope change (such as the bottom or top of the slope). Terrain data was acquired using RTK-GPS and a high-precision tilt sensor. GPS provided elevation data, and the tilt sensor provided local tilt angles. Acquisition proceeded along the direction of slope change, and a gradient tracking algorithm was used to determine the acquisition sequence. For example, when the slope vector indicated a direction of 30° east of north, the system started from the current sub-region and sequentially acquired terrain data from adjacent sub-regions in a direction of 30° east of north. Table 1 shows the feature acquisition results for some of the operating sub-regions.

[0062] Table 1: Feature collection data of the sub-region of the operation.

[0063] Sub-region coding Longitude of the center point (°) Latitude of the center point (°) Average crop density (plants / m²) Topographic slope angle (°) Azimuth of slope (°) A-1 116.3521 39.8765 6.8 2.3 45.2 A-2 116.3528 39.8767 7.2 1.8 42.1 A-3 116.3535 39.8769 5.9 3.1 48.5 B-1 116.3523 39.8772 6.5 4.2 52.3 B-2 116.3530 39.8774 7.8 2.9 46.7 B-3 116.3537 39.8776 6.2 3.5 49.8 In the fault prediction and analysis phase, the system first calculates the depth of the associated region based on the core acquisition area. The core acquisition area is identified through historical operation and maintenance data, typically selecting continuous areas with a fault frequency higher than a threshold. The depth of the associated region is calculated using an adaptive algorithm, taking into account the size and shape complexity of the core area. For example, for an approximately circular core area, the depth value is taken as 1.5 times the area radius; for irregularly shaped areas, the minimum bounding rectangle is used to calculate the reference depth.

[0064] After extracting the crop density and terrain undulation features of the first sub-region (e.g., coded A-1), the system determines the range of adjacent sub-regions based on the coordinates of this sub-region and the depth of the associated region. Adjacency relationships are quickly queried using a spatial indexing algorithm (e.g., R-tree), selecting all sub-regions whose distance from the center point of the current sub-region is within the depth range. A correlation analysis is performed on the crop density and terrain features of these adjacent regions, and a multiple linear regression model is used to calculate the failure risk index. The model input includes parameters such as the density difference, slope difference, and aspect consistency between the current sub-region and adjacent sub-regions, and the output is the first static failure feature value. This value is a normalized value between 0 and 1; a higher value indicates a greater failure risk.

[0065] The system traverses all operational sub-regions, performing the same analysis process for each. During the analysis, the set of adjacent sub-regions is dynamically updated to ensure that each sub-region calculates its characteristic values ​​based on its actual surrounding environment. The static fault characteristic values ​​of all sub-regions are ultimately integrated into a feature set and stored as a structured database table. Each record in the table contains the sub-region code, latitude and longitude coordinates, characteristic value, and timestamp.

[0066] The entire implementation process adopts a distributed computing architecture. The data acquisition layer uses embedded sensor nodes, the processing layer uses an edge computing gateway to perform real-time analysis, and the data storage layer uses a cloud platform for persistent storage. The system ensures the consistency of the collected data through a time synchronization mechanism and guarantees the integrity of data transmission through redundancy checks.

[0067] Example 2: See Figure 3 This system optimizes and controls the frequency of operation and maintenance data collection based on static fault feature sets, and adaptively adjusts it to address network bandwidth fluctuations. By analyzing the fault risk level of each operational sub-region and dynamically configuring the collection parameters along the data transmission path, the system achieves precise matching between network resources and collection requirements.

[0068] The implementation process begins with extracting static fault feature values ​​and their associated operational sub-regions from the static fault feature set. The static fault feature set is stored in a database table, containing fields such as: sub-region code, latitude and longitude coordinates, feature value collection timestamp, and static fault feature value (range 0.0-1.0). The system retrieves all records through an SQL query interface and establishes a mapping table between sub-region codes and feature values. Associated operational sub-regions are managed using a spatial relational database, utilizing PostGIS extended storage for the geographic polygon data of the sub-regions, and determining the adjacency relationships between sub-regions through spatial join queries.

[0069] Transmittable path identification is achieved through a network topology discovery module. This module employs Link Layer Discovery Protocol (LLDP) or route tracing technology to obtain the complete transmission path from the data acquisition terminal to the remote operation and maintenance center. The path is decomposed into multiple consecutive path segments, each corresponding to a network hop or physical link. The system creates a descriptor for each path segment, including: path segment ID, starting network device IP, ending network device IP, transmission medium type (e.g., 4G / 5G, satellite link, WiFi), and baseline bandwidth value. The mapping between adjacent operational sub-regions and path segments is established using geofencing technology. Each path segment is assigned a service area range, and the set of operational sub-regions it covers is determined by GPS coordinate matching.

[0070] The weighted calculation employs a distributed processing framework. For each path segment, the system queries all covered operational sub-regions to obtain the corresponding static fault characteristic values. Weight coefficients are set based on the relative distance between the sub-region and the path segment's base station; the closer the distance, the higher the weight. The calculation method is as follows: multiply the characteristic value of each sub-region by the weight coefficient, sum the results, and then divide by the total weight coefficient to obtain the data acquisition priority score (range 0.0-1.0) for that path segment. The system maintains a priority configuration table, recording each path segment ID and its corresponding real-time priority score.

[0071] When configuring sensor sampling intervals based on data acquisition priorities, the system employs a hierarchical control strategy. Priority scores are divided into four levels: 0.0-0.3 is low priority, with a sampling interval of 2000 milliseconds; 0.3-0.6 is medium priority, with a sampling interval of 1000 milliseconds; 0.6-0.8 is high priority, with a sampling interval of 500 milliseconds; and 0.8-1.0 is high priority, with a sampling interval of 200 milliseconds. Sampling interval configuration commands are published to each acquisition terminal via the MQTT protocol. The embedded system on the terminal device receives the commands and adjusts the timer parameters of the sensor driver program.

[0072] Network bandwidth fluctuation monitoring is achieved through real-time traffic probing. Lightweight probes are deployed at the starting network device of each path segment to measure the actual available bandwidth once per second. Bandwidth fluctuation is calculated as the absolute difference between the current bandwidth value and the baseline bandwidth value, divided by the baseline bandwidth value. A preset fluctuation threshold is set to 0.3 (i.e., 30% fluctuation). When the fluctuation exceeds this threshold for three consecutive sampling periods, a dynamic compression mechanism is triggered.

[0073] The dynamic compression process first sorts the path segments according to their data acquisition priority. For path segments with a priority below 0.5, the system initiates a data compression process: for image sensor data, JPEG2000 encoding is used to replace the original RAW format, reducing the color depth from 24 bits to 16 bits; for vibration sensor data, a lossy compression algorithm is used, reducing the sampling rate from 10kHz to 5kHz and the quantization bits from 16 bits to 12 bits; for operating parameter data, a floating-point precision reduction strategy is used, converting 32-bit floating-point numbers to 16-bit fixed-point numbers. The compression algorithm is implemented through programmable logic devices on the acquisition terminal, and the compression level is dynamically adjusted according to the degree of bandwidth fluctuation; the greater the fluctuation, the higher the compression ratio. Simultaneously, the system maintains the data transmission quality of high-priority path segments (priority > 0.7), only applying light compression to them in the most extreme cases (bandwidth drop exceeding 70%).

[0074] The entire implementation process employs a closed-loop control architecture, recalculating the priority scores and bandwidth status of each path segment every 30 seconds and dynamically adjusting the collection parameters. The system records all configuration change operations, including timestamps, path segment IDs, old parameter values, new parameter values, and reasons for adjustment, forming a complete operation and maintenance log. This method ensures the quality of critical data collection while effectively utilizing limited network bandwidth resources, even under fluctuating network conditions.

[0075] Example 3: See Figure 4 During real-time data acquisition, the sensing direction is dynamically calibrated, and the implementation method of feature weights is optimized based on historical calibration data. This process achieves precise adjustment of the acquisition terminal's orientation by continuously monitoring the deviation between operating parameters and expected characteristics, and optimizes the fault prediction model by learning from historical behavior.

[0076] The implementation process is illustrated using sub-area B-3 as an example. When the corn harvester enters this sub-area, the data acquisition system begins real-time monitoring of operating parameters. These parameters include engine speed, hydraulic system pressure, and travel motor torque, acquired from the CAN bus at 100-millisecond intervals. Simultaneously, the system retrieves pre-generated static fault characteristic values ​​(e.g., 0.72) from the static fault characteristic set for this sub-area. Changes in operating parameters are obtained by calculating the relative deviation between the current parameter value and the baseline parameter. Engine speed deviation is the difference between the actual speed and the rated speed; hydraulic pressure deviation is the percentage change between the current pressure and the set pressure; and travel slippage rate is calculated by comparing the theoretical travel distance with the GPS measured distance.

[0077] The deviation calculation employs a multi-parameter fusion method. Each parameter is first normalized, converting parameters with different dimensions into dimensionless values ​​within the 0-1 range. The normalized value for engine speed deviation is the ratio of the actual deviation to the maximum permissible deviation; hydraulic pressure deviation is processed using a similar method; and the travel slippage rate is directly normalized as a percentage value. These normalized values ​​are then weighted and summed to obtain the overall operating parameter change. The weighting coefficients are set according to the parameter importance: engine speed weight 0.4, hydraulic pressure weight 0.3, and travel slippage rate weight 0.3. The calculated overall change is compared with the static fault characteristic values ​​to determine the absolute deviation.

[0078] The preset tolerance threshold is set according to the equipment type and working environment, for example, 0.15. When the calculated deviation exceeds this threshold, the system initiates the calibration process. First, vibration spectrum data of the equipment is collected. Vibration signals are acquired at a sampling rate of 5kHz using triaxial vibration sensors installed on key parts such as the engine, cutting table, and traveling mechanism. The vibration signals are converted into frequency domain representation through fast Fourier transform, and the main frequency components and their amplitudes are extracted.

[0079] The calibration angle is determined based on the results of spectrum analysis. The system identifies the main frequency bands where vibration energy is concentrated and calculates the optimal sensing direction based on the mechanical structural characteristics of the equipment. For image acquisition equipment, the calibration angle includes two components: pitch and azimuth. The pitch angle is calculated based on the vertical energy distribution in the vibration spectrum, taking the phase angle corresponding to the main vibration frequency; the azimuth angle is determined based on the directional characteristics of vibration energy in the horizontal plane, obtained by calculating the average direction of the vibration vector. The orientation calibration of the vibration sensor is based on the mechanical vibration propagation characteristics, adjusting the sensor's main sensing direction to the direction of maximum vibration energy.

[0080] During calibration, the system generates control commands and sends them to the actuators. The tilt angle of the image acquisition device is adjusted by a stepper motor driving the pan-tilt unit, with an adjustment accuracy of 0.1 degrees. The orientation adjustment of the vibration sensor is accomplished by rotating the sensor base using a miniature servo motor, with an adjustment range of 360 degrees and an accuracy of 1 degree. The entire calibration process is completed in real time while the equipment is running, with a single calibration taking no more than 2 seconds.

[0081] The calibration operation history is meticulously saved. The calibration record table in the database includes fields such as: timestamp, sub-region code, deviation, calibration angle, calibration type (imaging equipment or vibration sensor), and parameter comparison before and after calibration. The system counts the calibration frequency for each sub-region and calculates the number of calibrations per unit time. The preset frequency threshold is set according to the workload, for example, 5 calibrations per work hour.

[0082] When the calibration frequency of a specific sub-region (e.g., B-3) exceeds a threshold, the system initiates a weight adjustment process. First, the calibration records for that sub-region are analyzed to calculate the average deviation and calibration angle distribution characteristics. The feature weight coefficients are adjusted using an incremental update algorithm, determining the weight increase based on the calibration frequency and deviation magnitude. The weight coefficient update formula considers a time decay factor, giving higher weights to more recent calibration records. The updated feature weight coefficients are written back to the static fault feature set, affecting subsequent fault prediction calculations.

[0083] This system employs a distributed architecture for calibration data management. Edge computing nodes handle real-time data acquisition and preliminary analysis, while cloud servers store historical data and execute weight optimization algorithms. Data transmission utilizes encryption protocols for security, and control commands employ redundant verification to ensure reliability. Through this implementation, the system can adaptively optimize sensing direction, improve data acquisition quality, and continuously enhance fault prediction accuracy through machine learning mechanisms.

[0084] Example 4: This system analyzes the distribution characteristics of static fault feature values, dynamically adjusts the data acquisition strategy, and achieves comprehensive monitoring of equipment operating status through multi-parameter fusion method.

[0085] At the start of the implementation process, the system reads the static fault characteristic values ​​of each work sub-region from the static fault characteristic set. These characteristic values ​​are stored in floating-point form, with a value range of [0,1], representing the fault risk level of each sub-region. The system uses a sliding window statistical method, selecting nine adjacent sub-regions as a statistical unit (3×3 grid) centered on the current work location. For each statistical unit, the dispersion index of its characteristic values ​​is calculated.

[0086] The dispersion is calculated using a statistical method based on absolute deviation, and the formula is as follows:

[0087] ;

[0088] Where: M represents the distribution dispersion, N represents the number of sub-regions within the statistical unit, s_i represents the static fault characteristic value of the i-th sub-region, and μ represents the arithmetic mean of all characteristic values ​​within the statistical unit. This formula calculates the average level of the absolute deviation of each characteristic value from the mean, effectively reflecting the degree of data dispersion.

[0089] The preset discrete threshold is set based on historical data analysis, and is usually a specific value within the range of 0.05-0.15. When the calculated M value is lower than this threshold, it indicates that the fault risk characteristics of these sub-regions are relatively similar, and data collection tasks can be merged.

[0090] The specific implementation process of the merging operation is as follows: The system first identifies adjacent sub-region groups that meet the merging conditions. These regions are not only spatially adjacent but also have similar fault characteristic values. Then, the system creates a new acquisition task unit for each merged group, which contains the geographical extent information of all sub-regions within the group. The original individual sub-region acquisition tasks are paused, and the merged acquisition tasks are executed instead. At the data storage level, sensor data within the merged area are marked as the same acquisition group and share the same acquisition parameter configuration.

[0091] Regarding operational parameter monitoring, the system collects three key parameters in real time: engine speed fluctuation, cutter head hydraulic pressure offset, and travel mechanism slippage rate. Engine speed fluctuation is collected by a speed sensor and calculated as the absolute difference between the actual and set speeds. Cutter head hydraulic pressure offset is obtained by a pressure sensor and expressed as the percentage deviation between the current pressure and the rated pressure. The travel mechanism slippage rate is obtained by comparing the theoretical travel distance (calculated based on drive wheel speed) with the actual travel distance (based on GPS positioning data).

[0092] These parameters are normalized to convert the original data of different dimensions into dimensionless values ​​within the range of [0,1]. The normalization method uses a min-max scaling algorithm. For each parameter type, the system maintains a normal range of values ​​and maps the measured values ​​to this range.

[0093] The normalized parameters are then weighted and fused, with weighting coefficients set according to the importance and reliability of each parameter. Engine speed fluctuation has a weight of 0.4, cutter head hydraulic pressure offset has a weight of 0.3, and travel mechanism slippage rate has a weight of 0.3. The weighted summation yields the overall operating parameter variation, which reflects the overall operating status of the equipment.

[0094] The entire implementation process employs a distributed computing architecture, with discrete calculations and merging decisions completed on edge computing nodes, enabling real-time optimization of the data acquisition task. Comprehensive evaluation of operating parameters is performed at the data acquisition terminal, reducing data transmission volume. The system maintains a task configuration database, recording the merging status and acquisition parameters of each sub-region to ensure the consistency and integrity of data acquisition.

[0095] This implementation method enables the system to optimize resource utilization efficiency and reduce redundant data acquisition while ensuring data acquisition quality, and simultaneously achieve comprehensive monitoring of equipment operating status. Dynamic merging of data acquisition tasks reduces system overhead, while the multi-parameter fusion method provides a more comprehensive assessment of operating status.

[0096] Example 5: An environmental adaptability model is constructed based on historical calibration data, and intelligent pre-configuration of acquisition parameters is achieved when a new work area is accessed. This implementation method analyzes historical work data using machine learning methods to establish a mapping relationship between sub-regional environmental characteristics and optimal acquisition parameters, thereby improving the adaptability and efficiency of the data acquisition system.

[0097] The implementation process begins with the collection and organization of historical calibration data. The system extracts calibration records from the operations and maintenance database, including timestamps, sub-region codes, environmental parameters (temperature, humidity, soil type), equipment operating parameters (operating speed, cutter head height), calibration angle deviation, calibration type, and other information. This data undergoes cleaning and preprocessing to remove outliers and duplicate records, forming a standardized training dataset. Each data sample includes input features (environmental parameters and equipment parameters) and output labels (optimal acquisition parameter configuration).

[0098] The environmental adaptability model was constructed using a supervised learning method. The system employed a random forest algorithm for model training, which can handle high-dimensional features and capture nonlinear relationships. Input features included continuous variables such as average crop density in the sub-region, terrain slope, soil moisture content, ambient temperature, and operation time, as well as soil type (categorical variables encoded as numerical values). Output labels included acquisition parameters such as the initial pitch angle of the image acquisition equipment, the foundation orientation of the vibration sensor, and the data sampling interval. During training, five-fold cross-validation was used to adjust model hyperparameters, including the number of trees, maximum depth, and minimum number of leaf samples.

[0099] After model training is complete, the system saves the model parameters and feature importance evaluation results. Feature importance analysis shows that crop density and terrain slope have the most significant impact on the configuration of the collected parameters, which is consistent with the actual situation of agricultural machinery operations. The model is deployed on a cloud server and provides a RESTful API interface for real-time calls.

[0100] When a new operational area distribution map is received, the system initiates a parameter pre-configuration process. First, it parses the geographic information of the new distribution map, including boundary coordinates, grid division scheme, and geographic features of each sub-region. For each sub-region, its environmental feature vector is extracted, including vegetation indices obtained from remote sensing data, slope information derived from the digital elevation model, and soil type distribution data. These features are standardized according to model input requirements to ensure they share the same feature space as the training data.

[0101] The system invokes the environmental fitness model for prediction, generating initial acquisition parameter configurations for each sub-region. The prediction process employs batch processing, processing the feature data of all sub-regions at once and outputting the corresponding parameter configuration schemes. These configurations include: the suggested pitch angle of the image acquisition device (range -15° to +15°), the reference orientation of the vibration sensor (0° to 360°), and the suggested sampling interval (200ms to 2000ms). The configuration results are stored in JSON format and mapped to the sub-region encoding.

[0102] Pre-configured parameters are sent to each acquisition terminal via a wireless network. After receiving the configuration command, the terminal device automatically adjusts the sensor parameters: the image acquisition device adjusts to the specified pitch angle via a servo motor, the vibration sensor adjusts the orientation angle via a rotation mechanism, and the data acquisition module sets the corresponding sampling timer. The system establishes a configuration verification mechanism, collecting actual data after the parameters are sent and comparing it with the expected results. If a significant deviation occurs, a reconfiguration process is triggered.

[0103] During implementation, continuous model optimization is emphasized. The system records the actual data collected and the effects of use in new work areas, updates the training dataset regularly, and retrains the environmental adaptability model. Model updates employ incremental learning, incorporating new work data while retaining existing knowledge, enabling the model to adapt to environmental changes in different seasons and regions.

[0104] The entire implementation process adopts a microservice architecture, with model training, prediction services, and configuration management running as independent services, exchanging data through message queues. This architecture ensures the system's scalability and reliability, capable of handling simultaneous access requests from large-scale job areas. Data encryption and access control mechanisms ensure the security of operational data and prevent unauthorized access.

[0105] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0106] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A corn harvester remote operation and maintenance data collection optimization method, characterized in that, The method comprises the following steps: acquiring a real-time working area distribution map of a corn harvester, determining a core collection area based on the working area distribution map; collecting working condition characteristics according to each working sub-area in the working area distribution map to obtain crop density characteristics and terrain undulation characteristics corresponding to each working sub-area respectively; performing fault prediction analysis on the crop density characteristics and the terrain undulation characteristics based on the core collection area to generate a static fault feature set; acquiring a transmissible path of a remote operation and maintenance terminal, controlling the collection frequency of operation and maintenance data on the transmissible path based on each static fault feature value in the static fault feature set and the corresponding working sub-area; monitoring the change amount of the operating parameters of the current working sub-area in real-time data collection, and dynamically calibrating the sensing direction of the collection terminal in combination with the static fault feature value corresponding to the current working sub-area; the fault prediction analysis on the crop density characteristics and the terrain undulation characteristics based on the core collection area comprises: determining an associated area depth according to the core collection area; extracting the crop density characteristics and the terrain undulation characteristics of a first working sub-area, selecting adjacent working sub-areas for fault feature correlation analysis according to the position coordinates of the first working sub-area and the associated area depth, the fault feature correlation analysis adopting a multivariate linear regression model to generate a first static fault feature value; generating the static fault feature values of each working sub-area in the same way in sequence, and integrating them into a static fault feature set according to the working area distribution map; the control of the collection frequency of operation and maintenance data on the transmissible path comprises: extracting each static fault feature value in the static fault feature set and its associated working sub-area; identifying all path segments of the transmissible path and their corresponding adjacent working sub-areas; weighting and superimposing the static fault feature values of the corresponding adjacent working sub-areas of each path segment to generate the data collection priority of each path segment; configuring the data sampling interval of the sensor on the corresponding path segment based on the data collection priority.

2. The corn harvester remote operation and maintenance data collection optimization method of claim 1, wherein, the collection of working condition characteristics according to each working sub-area in the working area distribution map to obtain crop density characteristics and terrain undulation characteristics corresponding to each working sub-area respectively comprises: grid division based on the region boundary coordinates in the working area distribution map to generate a plurality of working sub-areas; acquiring a harvesting travel direction, collecting crop density characteristics of the starting sub-area according to the harvesting travel direction, and collecting crop density characteristics of subsequent working sub-areas in sequence along the travel direction; acquiring a ground slope vector, reselecting a starting sub-area to collect terrain undulation characteristics according to the ground slope vector, and collecting terrain undulation characteristics of subsequent working sub-areas in sequence along the slope change direction.

3. The corn harvester remote operation and maintenance data collection optimization method of claim 1, wherein, the dynamic calibration of the sensing direction of the collection terminal comprises: calculating the deviation amount between the change amount of the operating parameters of the current working sub-area and the static fault feature value; when the deviation amount exceeds a preset tolerance threshold, determining a calibration angle according to the equipment vibration spectrum of the current working sub-area; adjusting the pitch angle of the image collection equipment and the orientation of the vibration sensor of the corresponding working sub-area based on the calibration angle.

4. The corn harvester remote operation and maintenance data collection optimization method of claim 3, wherein, The generating static fault feature set further comprises: Statistics of the static fault feature value distribution dispersion of each work sub-region; When the distribution dispersion is lower than the preset dispersion threshold, the operation and maintenance data collection tasks of adjacent work sub-regions are combined.

5. The corn harvester remote operation and maintenance data collection optimization method of claim 4, wherein, The control of the collection frequency of operation and maintenance data further comprises: Monitoring the bandwidth fluctuation of the transmissible path; When the bandwidth fluctuation is greater than the preset fluctuation threshold, the original data resolution of the low-priority path segment is dynamically compressed according to the data collection priority.

6. The corn harvester remote operation and maintenance data collection optimization method of claim 5, wherein, The acquisition of the running parameter change amount comprises: Real-time collection of engine speed fluctuation, header hydraulic pressure offset and walking mechanism slip rate; The speed fluctuation, hydraulic pressure offset and slip rate are normalized and weighted to generate a comprehensive running parameter change amount.

7. The corn harvester remote operation and maintenance data collection optimization method of claim 6, wherein, The dynamic calibration further comprises: Recording the calibration frequency and deviation history data of each work sub-region; When the calibration frequency of a specific work sub-region exceeds the preset frequency threshold, the feature weight coefficient of the work sub-region in the static fault feature set is increased.

8. The corn harvester remote operation and maintenance data collection optimization method of claim 7, wherein, The method further comprises: Based on the historical calibration data, an environmental fitness model is constructed, and a random forest algorithm is used for training the environmental fitness model; When a new work area distribution map is accessed, the initial collection parameters of each sub-region are preconfigured according to the environmental fitness model.

Images