Mulching film continuous paving constant-speed adjusting method based on tension feedback

Abstract

The invention relates to the technical field of agricultural machinery automation control, and discloses a mulching film continuous paving constant speed adjusting method based on tension feedback, which comprises the following steps: collecting real-time tension feedback data and a historical database of mulching film paving equipment; classifying the historical tension data according to the paving state characteristics, and screening key tension dimensions by analyzing classification deviation; calculating data weight distribution based on the characteristic difference, and optimizing a constant speed adjustment model; and finally, inputting real-time tension data into the trained model, generating a speed adjusting instruction, and realizing uniform speed control of mulching film paving. According to the method, through data classification, key dimension screening and dynamic weight optimization, the response speed and the adjustment precision of tension fluctuation are remarkably improved, the method adapts to a complex farmland environment, the mulching film paving uniformity is ensured, and the phenomena of breakage and accumulation are reduced. The method has the characteristics of strong generalization ability and high automation degree, and can be widely applied to agricultural mechanized operation.

Description

Technical Field

[0001] The present invention relates to the technical field of agricultural machinery automation control, and in particular to a method for adjusting the uniform speed of continuous paving of ground film based on tension feedback. Background Art

[0002] As an important agronomic measure in modern agricultural production, mulching technology plays a key role in water and fertilizer conservation, temperature increase, and weed control, significantly promoting crop growth and increasing yields and incomes. With the acceleration of agricultural modernization, the demand for automated and efficient mulching operations is becoming increasingly urgent, and mulching equipment is constantly being updated and upgraded. However, in actual continuous mulching operations, maintaining a uniform mulching speed presents numerous challenges, becoming a key bottleneck restricting the quality and efficiency of mulching. Traditional methods of adjusting the speed of mulch film laying rely heavily on the operator's experience and manual adjustments. Operators must constantly monitor the film's progress and adjust the speed of the equipment based on their experience. This approach not only requires a high level of operator experience and skill, but also presents significant limitations. Due to errors and individual differences in subjective judgment, precise speed control is difficult to achieve, which can easily lead to uneven film tension during the laying process. Excessive tension can cause the film to overstretch or even rupture, impacting its lifespan and its ability to protect crops. Excessive tension can cause the film to sag and wrinkle, preventing it from adhering tightly to the ground. This reduces its thermal and water-retention properties, further impacting the growing environment for crops. While some existing mulch-film laying equipment is equipped with simple speed adjustment mechanisms, these devices often fail to comprehensively consider the various complex factors involved in the laying process. For example, factors such as soil texture, terrain undulations, and changes in the weight of the mulch roll all affect the tension during mulch laying. Traditional equipment struggles to detect these changes in real time and adjust the speed accordingly. This inability to adaptively adjust the laying speed in different operating environments results in poor versatility and applicability, making it unable to meet the diverse needs of agricultural production. With the gradual application of artificial intelligence and big data technologies in the agricultural field, some data-driven methods for regulating the speed of mulch film laying have begun to emerge. However, these existing methods have shortcomings in data processing and model training. They often do not fully explore the information contained in the historical laying data, the classification and analysis of the data are not detailed enough, and they cannot accurately identify the characteristic differences in the state of mulch film laying under different working conditions. During the model training process, no optimization is performed on the key influencing factors, resulting in low accuracy of the trained model, making it difficult to achieve precise regulation of the mulch film laying speed in practical applications. Therefore, the development of a method that can effectively utilize tension feedback data to achieve uniform speed regulation of continuous mulch film laying has become an important issue that needs to be urgently addressed in the current agricultural engineering field. Summary of the Invention

[0003] The purpose of the present invention is to provide a method for adjusting the uniform speed of continuous laying of ground film based on tension feedback, so as to solve the problems raised in the above background technology.

[0004] To achieve the above-mentioned object, the present invention provides the following technical solution: a method for adjusting the uniform speed of continuous paving of ground film based on tension feedback, the method comprising: Collect the real-time tension feedback data obtained by the ground film laying equipment and the database composed of multiple sets of tension feedback data recorded during the historical laying process; The historical tension data are divided into several types according to the ground film laying state characteristics corresponding to each group of tension data in the database; a uniform speed adjustment model is trained based on the database, and key tension dimensions are selected by analyzing data classification deviations; data weight distribution is calculated based on the characteristic differences of different data on the key tension dimensions; the uniform speed adjustment model is iteratively optimized based on the weight distribution to obtain a trained uniform speed adjustment model; The tension feedback data obtained in real time by the ground film laying equipment is input into the trained uniform speed adjustment model to obtain a speed adjustment instruction, and the ground film continuous laying uniform speed adjustment operation is performed based on the speed adjustment instruction.

[0005] Preferably, the historical tension data is divided into several types according to the ground film laying state characteristics corresponding to each group of tension data in the database, including: For the mth group of historical tension data: Each training stage of the mth group of historical tension data from the first misclassification to the correct classification is recorded as an analysis stage of the mth group of historical tension data; the state feature changes corresponding to the mth group of historical tension data in all analysis stages are arranged in chronological order to obtain a feature change sequence of the mth group of historical tension data; the position index of each value in the feature change sequence is used as the horizontal axis coordinate, and the value corresponding to each position index is used as the vertical axis coordinate to form no less than two coordinate points, all coordinate points are used as input for feature distribution analysis, and each two-dimensional coordinate vector and the corresponding mapping value are obtained; the two-dimensional coordinate vector corresponding to the maximum mapping value is used as the main direction vector, and the arc tangent value of the ratio of the vertical axis component to the horizontal axis component in the main direction vector is used as the feature distribution direction quantity; Calculate the classification evaluation value of the mth group of historical tension data based on the training stage corresponding to the first misclassification of the mth group of historical tension data, the stage interval between the first misclassification and the first correct classification, and the feature distribution direction; The type of the mth group of historical tension data is determined based on the classification evaluation amount.

[0006] Preferably, the calculating of the classification evaluation amount of the mth group of historical tension data based on the training stage corresponding to the first misclassification of the mth group of historical tension data, the stage interval between the first misclassification and the first correct classification, and the feature distribution direction amount includes: Recording the difference between the characteristic distribution direction and the preset reference direction as a first evaluation parameter; Calculate the inverse normalization result of the training stage corresponding to the first misclassification of the mth group of historical tension data, and determine the product of the stage interval between the first misclassification and the first correct classification, the inverse normalization result and the first evaluation parameter as the classification evaluation value of the mth group of historical tension data.

[0007] Preferably, the determining the type of the mth group of historical tension data based on the classification evaluation amount includes: If the classification evaluation value is greater than the preset type classification threshold, the mth group of historical tension data is determined to belong to the high-fluctuation tension type; otherwise, it is determined to belong to the low-fluctuation tension type.

[0008] Preferably, the screening of key tension dimensions by analyzing data classification deviations includes: For the nth type: Constructing a first feature matrix based on all correctly classified tension data in the nth type, wherein each row in the first feature matrix is a group of correctly classified tension data; processing the first feature matrix using a feature screening algorithm to obtain a first feature screening result, wherein each column of data in the first feature screening result constitutes a first candidate tension dimension; constructing a second feature matrix based on the misclassified tension data in the nth type, wherein each row in the second feature matrix is a group of misclassified tension data; processing the second feature matrix using a feature screening algorithm to obtain a second feature screening result, wherein each column of data in the second feature screening result constitutes a second candidate tension dimension; Matching and parsing the first candidate tension dimension and the second candidate tension dimension are performed, and key tension dimensions are screened based on the matching results.

[0009] Preferably, performing matching analysis on the first candidate tension dimension and the second candidate tension dimension, and screening the key tension dimension based on the matching result, includes: A feature association algorithm is used to match the first candidate tension dimension with the second candidate tension dimension to obtain multiple feature matching groups; the correlation between the two dimensions in each feature matching group is calculated respectively, and the feature matching group with a correlation greater than a preset correlation threshold is used as the target matching group; The average dimension of the two dimensions in each target matching group was taken as a key tension dimension.

[0010] Preferably, the step of calculating data weight distribution based on feature differences of different data on the key tension dimension includes: Correct tension data for group p: Record the misclassified data in the type of the correctly classified tension data of the pth group as the reference data of the correctly classified data of the pth group; calculate the mean of the characteristic values of all reference data of the correctly classified data of the pth group on each key tension dimension, and record the absolute difference between the characteristic value of the correctly classified data of the pth group on each key tension dimension and the corresponding mean as the difference index of the correctly classified data of the pth group on each key tension dimension; The sum of the inverse normalized results of the difference indicators of the p-th group of correctly classified data in all key tension dimensions is respectively used as the weight coefficient of the p-th group of correctly classified data; and the target weight of the p-th group of correctly classified data in each training stage is calculated based on the weight coefficient.

[0011] Preferably, obtaining the ground film laying state characteristics corresponding to each set of tension data in the database includes: The ground film laying tension trajectory corresponding to each group of tension data in the database is statistically analyzed to obtain the corresponding trajectory distribution histogram, and the boundary segmentation algorithm is used to segment the histogram to obtain no less than two segmentation intervals; The average value of the tension trajectory change rate of all tension data in each segmented interval is calculated and used as the ground film laying state feature of each group of tension data in the corresponding segmented interval.

[0012] Preferably, the calculating of the target weight of the p-th group of correctly classified data in each training stage based on the weight coefficient includes: For any group of correctly classified tension data: multiply the original weight of the group of data in each training stage by the corresponding weight coefficient, which is used as the target weight of the group of data in each training stage.

[0013] Preferably, the step of inputting the tension feedback data obtained in real time by the ground film laying equipment into the trained uniform speed adjustment model to obtain the speed adjustment instruction includes: The tension feedback data acquired in real time is subjected to noise filtering to extract effective tension characteristic values; the effective tension characteristic values are input into the uniform speed adjustment model, and corresponding speed adjustment instructions are generated through the tension-speed mapping rules within the model.

[0014] Compared with the prior art, the present invention has the following beneficial effects: At the data processing and analysis level, this method first collects the tension feedback data obtained in real time by the ground film laying equipment and a database consisting of multiple sets of tension feedback data recorded during the historical laying process. Through in-depth analysis of the historical tension data, it is divided into several types based on the ground film laying state characteristics corresponding to each set of tension data in the database. Specifically, for each set of historical tension data, by analyzing the various training stages from the first misclassification to the correct classification, a feature change sequence is constructed, and then parameters such as the feature distribution direction are determined. The classification evaluation amount is calculated in combination with the training stage and stage interval corresponding to the first misclassification to determine the type of data. This meticulous data classification method can accurately identify the characteristic differences of tension data in the ground film laying process under different working conditions, fully explore the information contained in the historical data, and provide a solid data foundation for subsequent model training and speed adjustment.

[0015] In terms of model training optimization, the uniform speed regulation model is trained based on the divided data types. By analyzing the data classification deviation, the key tension dimensions are screened. That is, feature matrices are constructed for the correctly and incorrectly classified tension data respectively. After processing with the feature screening algorithm, matching analysis is performed to screen out the key dimensions that have a greater impact on model classification and speed regulation. At the same time, the data weight distribution is calculated based on the feature differences of different data on the key tension dimensions, and the appropriate weight coefficient and target weight are determined for each group of correctly classified data. This weight distribution method based on data feature differences enables the model to pay more attention to important data information during the training process, effectively improving the model's adaptability and generalization ability to different working conditions, and improving the model's training accuracy and reliability. In practical application, real-time tension feedback data from mulch film laying equipment is fed into a trained constant-speed adjustment model. By filtering the data for noise and extracting effective tension eigenvalues, the model's internal tension-speed mapping rules are used to generate speed adjustment commands, achieving constant-speed adjustment for continuous mulch film laying. This approach detects tension changes during mulch film laying in real time and quickly and accurately adjusts speed accordingly, effectively avoiding problems such as excessive mulch stretching, loosening, and wrinkling caused by uneven tension, significantly improving the quality and effectiveness of mulch film laying. Compared to traditional methods that rely on manual experience or simple equipment adjustments, this method significantly reduces manual intervention, lowers operator workload and skill requirements, and improves the automation and efficiency of mulch film laying operations. Furthermore, through adaptive training for a variety of complex working conditions, this method enables mulch film laying equipment to maintain excellent performance in diverse soil textures, terrain, and other operating environments. This effectively improves the versatility and applicability of the equipment, better meeting the diverse needs of agricultural production and providing strong technical support for the efficient and precise development of modern agriculture. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 This is a working principle diagram of the method for adjusting the uniform speed of continuous paving of ground film based on tension feedback according to the present invention; Figure 2 Design diagrams divided for historical tension data types; Figure 3 Design diagrams screened for key tension dimensions; Figure 4 A design diagram for data weight assignment. DETAILED DESCRIPTION

[0017] 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.

[0018] See also Figure 1-Figure 4 The present invention relates to a method for adjusting the uniform speed of continuous paving of ground film based on tension feedback, and the specific implementation steps are as follows: Collect the tension feedback data obtained in real time by the ground film laying equipment and a database consisting of multiple sets of tension feedback data recorded in the historical laying process.

[0019] The historical tension data are divided into several types according to the ground film laying status characteristics corresponding to each group of tension data in the database; the uniform speed adjustment model is trained based on the database, and the key tension dimensions are screened by analyzing the data classification deviation; the data weight distribution is calculated according to the characteristic differences of different data on the key tension dimensions; the uniform speed adjustment model is iteratively optimized based on the weight distribution to obtain the trained uniform speed adjustment model.

[0020] The tension feedback data obtained in real time by the ground film laying equipment is input into the trained uniform speed adjustment model to obtain the speed adjustment instruction, and the ground film continuous laying uniform speed adjustment operation is performed based on the speed adjustment instruction.

[0021] Example 1: During the classification process of historical tension data, the analysis phase for the mth group of historical tension data must be determined. Each training phase of the mth group of historical tension data, from the initial misclassification to the final correct classification, is recorded as an analysis phase for that group of data. The key to this step is to clearly define the stages of data state change during the training process, providing a time-based basis for subsequent feature analysis.

[0022] Arrange the state feature changes corresponding to the mth set of historical tension data across all analysis phases in chronological order. State feature changes here refer to the numerical changes in state features associated with the tension data during each training phase. By arranging these changes in chronological order, a sequence is formed that reflects the characteristic change trends of this set of data during the training process, namely, the characteristic change sequence of the mth set of historical tension data.

[0023] Perform coordinate transformation on the feature change sequence. The position index of each value in the feature change sequence is used as the horizontal coordinate, and the value corresponding to each position index is used as the vertical coordinate, thus forming at least two coordinate points. These coordinate points constitute the input data for feature distribution analysis. By analyzing these coordinate points, each two-dimensional coordinate vector and the corresponding mapping value can be obtained. During this process, attention must be paid to the accuracy and completeness of the coordinate points to ensure the reliability of subsequent analysis.

[0024] After obtaining the 2D coordinate vectors and mapping values corresponding to all coordinate points, the principal direction vector needs to be determined. This is done by finding the 2D coordinate vector corresponding to the maximum mapping value and determining it as the principal direction vector. The principal direction vector reflects the primary distribution direction of the feature change sequence in 2D space and is a key parameter for subsequent calculations of the feature distribution direction.

[0025] After determining the principal direction vector, the feature distribution direction is calculated. Specifically, the inverse tangent of the ratio of the vertical component to the horizontal component of the principal direction vector is calculated and used as the feature distribution direction. This value quantifies the distribution direction of the feature change sequence and provides an important basis for the subsequent classification evaluation calculation.

[0026] After obtaining the characteristic distribution direction, a first evaluation parameter needs to be calculated. The difference between the characteristic distribution direction and the preset reference direction is recorded as the first evaluation parameter. The preset reference direction is a reference value set based on historical experience or relevant standards. It is used to compare the actual calculated characteristic distribution direction to determine the degree of difference between the two.

[0027] It is also necessary to calculate the inverse normalization result of the training phase corresponding to the first misclassification of the mth group of historical tension data. Inverse normalization is to convert the numerical value of the training phase into a relative value to facilitate the comprehensive calculation with other parameters.

[0028] Calculate the classification evaluation metric. Multiply the interval between the first misclassification and the first correct classification, the inverse normalization result, and the first evaluation parameter. The resulting product is the classification evaluation metric for the mth set of historical tension data. This calculation comprehensively considers the time interval between the data during training, the relative position of the misclassification stages, and the differences in feature distribution directions, providing a comprehensive reflection of the classification difficulty and feature distribution characteristics of the data set.

[0029] Determine the type of historical tension data based on the classification evaluation value. Set a preset classification threshold. If the classification evaluation value is greater than the threshold, the mth group of historical tension data is classified as high-fluctuation tension. If the classification evaluation value is less than or equal to the threshold, it is classified as low-fluctuation tension. The preset classification threshold should be determined based on actual application scenarios and historical data to ensure the accuracy and effectiveness of the classification.

[0030] Throughout the implementation process, each step must be strictly followed according to the above process to ensure accurate data processing and analysis. At the same time, attention must be paid to the calculation method and value range of each parameter to ensure that the final classification results truly reflect the actual characteristics of the historical tension data. For example, when determining the analysis phase, the time points of the first misclassification and the first correct classification must be accurately identified; when calculating coordinate points, the correct correspondence between position index and numerical value must be ensured; and when setting the preset reference direction and preset classification threshold, the actual application needs and data characteristics must be fully considered.

[0031] Example 2: During the implementation process of screening key tension dimensions, a first feature matrix is constructed for the nth category. This matrix is constructed based on all correctly classified tension data within the nth category, with each row representing a set of correctly classified tension data. Correctly classified tension data here refers to tension data sets that have been accurately classified into the nth category during the previous training or classification process. When constructing the first feature matrix, it is necessary to ensure that each row of data in the matrix fully and accurately reflects the characteristics of the corresponding tension data.

[0032] After constructing the first feature matrix, it is processed using a feature screening algorithm. The feature screening algorithm selects features that have a significant impact on the classification results from a large number of features. During this process, the algorithm evaluates and ranks the features in each column of the matrix according to specific rules and indicators, resulting in the first feature screening results. In this first feature screening result, each column of data constitutes a first candidate tension dimension. These first candidate tension dimensions are selected from correctly classified data and are likely to play an important role in the classification of that type.

[0033] After processing the first feature matrix, the next step is to process misclassified tension data. A second feature matrix is constructed based on the misclassified tension data in the nth category. Similarly, each row in the second feature matrix represents a set of misclassified tension data. Misclassified tension data refers to tension data that was incorrectly classified into a category other than the nth category during the previous classification process. When constructing the second feature matrix, ensure the accuracy and completeness of the data so that subsequent feature screening accurately reflects the characteristics of the misclassified data.

[0034] After constructing the second feature matrix, we process it using the same feature screening algorithm as the first feature matrix to obtain the second feature screening results. Each column of data in the second feature screening results constitutes a second candidate tension dimension. These second candidate tension dimensions are selected from misclassified data and are likely to be related to the cause of the misclassification.

[0035] After obtaining the first and second candidate tension dimensions, they need to be matched and analyzed. Specifically, a feature association algorithm is used to match the first and second candidate tension dimensions, thereby obtaining multiple feature matching groups. The feature association algorithm matches the first and second candidate tension dimensions pairwise based on metrics such as feature correlation and similarity, forming different feature matching groups.

[0036] For each feature matching group, the correlation between two dimensions is calculated. This correlation is calculated based on the inherent connections and mutual influence between features, using mathematical or statistical methods. Feature matching groups with a correlation greater than a preset correlation threshold are considered target matching groups. The preset correlation threshold is a pre-set standard based on actual application requirements and data characteristics, used to determine whether the correlation between feature matching groups is sufficiently high.

[0037] The key tension dimension is selected from each target matching group. Specifically, the average dimension of the two dimensions in each target matching group is used as the key tension dimension. This average dimension is calculated by averaging the feature values of the two dimensions. It combines the information of the two dimensions and can more comprehensively reflect the relationship and influence between the features.

[0038] Throughout the implementation process, attention must be paid to the accuracy and rigor of each step. For example, when constructing the feature matrix, the classification accuracy and completeness of the data must be ensured; when selecting feature screening algorithms and feature association algorithms, reasonable choices must be made based on the characteristics of the data and actual needs; when setting preset association thresholds, the distribution of the data and classification requirements must be fully considered. Through the detailed implementation methods described above, key tension dimensions that have a significant impact on the classification results can be screened out from correctly and incorrectly classified data, providing key feature support for subsequent uniform speed adjustment model training and optimization, enabling the model to more accurately adjust the speed based on the tension data, thereby achieving uniform speed adjustment for continuous paving of the ground film.

[0039] Example 3: During the implementation of calculating data weights, reference data for the correctly classified tension data in group p is required. The incorrectly classified data within the same category as the correctly classified tension data in group p is recorded as reference data for the correctly classified data in group p. Here, incorrectly classified data refers to tension data sets that are incorrectly classified into the same category under the same classification criteria. These data fall within the same category as the data in group p, but their classification results differ. Therefore, these data serve as a reference for analyzing the characteristic differences in the data in group p.

[0040] After determining the reference data, eigenvalue analysis is performed on each critical tension dimension. For all reference data from the pth group of correctly classified data, the mean of the eigenvalues along each critical tension dimension is calculated. The critical tension dimensions here are those that have a significant impact on the classification results, as determined through the previous feature screening step. Each critical tension dimension corresponds to a set of eigenvalues. To calculate the mean, the eigenvalues of all reference data along that dimension are summed and then divided by the number of reference data to obtain the average eigenvalue of the reference data along that dimension.

[0041] Calculate the difference index for each key tension dimension of the correctly classified data set p. Specifically, take the absolute difference between the eigenvalue and the corresponding mean for each key tension dimension of the correctly classified data set p. This difference index is recorded as the difference index for that dimension. The calculation of the absolute difference eliminates the influence of the magnitude and direction of the eigenvalue, reflecting only the degree of deviation between the data set and the reference data on that dimension. A larger difference index indicates a more significant difference in the characteristics of the data set compared to the reference data on that dimension.

[0042] After obtaining the difference index on each key tension dimension, these difference indexes need to be inverse normalized. Inverse normalization is the process of converting the difference index into a relative value. The specific operation is to map the value range of the difference index to a specific interval through a certain mathematical transformation, so as to facilitate the subsequent summation calculation. The inverse normalization results of the difference index of the p-th group of correctly classified data on all key tension dimensions are summed up respectively, and the sum obtained is the weight coefficient of the p-th group of correctly classified data. The weight coefficient reflects the comprehensive degree of difference between this group of data and the reference data in all key tension dimensions. The greater the degree of difference, the larger the weight coefficient, which means that this group of data may have a higher importance in model training.

[0043] After obtaining the weight coefficient, it is necessary to calculate the target weight of the data set in each training stage. For any group of correctly classified tension data, it originally has an original weight in each training stage, and the original weight may be determined according to the training rules or initial settings. Multiply the original weight of the data set in each training stage by the corresponding weight coefficient, and the product obtained is the target weight of the data set in this training stage. In this way, the weight coefficient is integrated into the original weight to realize the dynamic adjustment of the data weight in the training stage, so that the data with large differences from the reference data will obtain higher weights in training, thereby guiding the model to pay more attention to these representative data and improve the training effect of the model.

[0044] Throughout the implementation process, attention must be paid to the accuracy and logic of each step. For example, when determining the reference data, it is necessary to ensure that it is of the same type as the data in group p and that the classification results are correct to ensure the validity of the reference. When calculating the mean, the eigenvalues of all reference data must be accurately processed to avoid calculation errors. When performing inverse normalization, an appropriate normalization method must be selected to ensure that the converted values can reasonably reflect the degree of difference. When calculating the target weight, the corresponding relationship between the original weight value and the weight coefficient must be clearly defined to ensure that the calculation is accurate in each training phase.

[0045] The diversity and complexity of the data also need to be considered. Different groups of correctly classified data may have different patterns of feature differences. Therefore, when processing each data set, the calculations above need to be performed one by one to ensure that the weight assignment for each data set accurately reflects its unique characteristics. Furthermore, the number and selection of key tension dimensions will also affect the weight assignment results. Therefore, in the early feature screening step, the key dimensions must be determined strictly according to the screening rules to ensure the reliability of the subsequent weight calculation.

[0046] Through the above detailed implementation method, it is possible to calculate a reasonable weight distribution for each group of correctly classified tension data based on the characteristic differences of different data in the key tension dimension, and dynamically adjust its target weight in each training stage, so that the uniform speed adjustment model can learn data with different characteristics more specifically during the training process, improve the model's analysis and processing capabilities of tension data, and ultimately achieve more accurate speed adjustment instruction generation, thereby ensuring the uniformity of the continuous paving process of the ground film.

[0047] Example 4: To obtain the film paving status characteristics corresponding to each set of tension data in the database, the tension data in the database must be processed. Taking a specific set of historical tension data as an example, assuming that this data corresponds to the tension changes during a certain period of time during the film paving process, the corresponding film paving tension trajectory is first statistically analyzed. This tension trajectory refers to a curve showing the tension changes over time during the paving process, reflecting the tension fluctuations throughout the entire paving process.

[0048] After counting the tension trajectories, a corresponding trajectory distribution histogram will be obtained. The histogram construction process is to divide the tension value range into several intervals, and then count the number of times the tension data appears in each interval. The horizontal axis represents the tension interval and the vertical axis represents the number of times, thus forming a histogram. For example, if the tension value range is between 0 and 100, it can be divided into multiple intervals such as 0-20, 20-40, 40-60, 60-80, and 80-100. The number of times this group of tension data appears in each interval is then counted to draw a histogram.

[0049] A boundary segmentation algorithm is used to segment the histogram. The boundary segmentation algorithm finds appropriate segmentation points in the histogram, dividing the histogram into at least two intervals. Different boundary segmentation algorithms have different segmentation criteria, such as the peak and trough characteristics of the histogram, density changes in the data distribution, etc. For example, if the histogram exhibits two distinct peaks, the algorithm may segment the histogram at the trough between the two peaks, dividing the histogram into two intervals. If the histogram has multiple peaks, the algorithm may segment the histogram into more intervals.

[0050] Suppose that a boundary segmentation algorithm is used to process the histogram in the above example, resulting in two segmented intervals, interval A and interval B. Interval A corresponds to a lower tension value, while interval B corresponds to a higher tension value. In this case, the average rate of change of the tension trajectory of all tension data in each segmented interval needs to be calculated and used as the mulch film laying status feature for each set of tension data in the corresponding segmented interval.

[0051] The calculation of the tension trajectory change rate is for each tension data point, which represents the rate of change of tension at that point over time. Taking two adjacent tension data points as an example, assuming that the tension is F1 at time t1 and F2 at time t2, the time interval is , then the rate of change of the tension trajectory between these two points is For all tension data points in each segmented interval, it is necessary to calculate the rate of change between every two adjacent points, and then average these change rates to obtain the average value of the tension trajectory change rate in the interval.

[0052] Continuing with the example, if there are n tension data points in interval A, there will be n-1 pairs of adjacent points. Calculate the rate of change of these n-1 pairs, sum them, and divide by n-1 to obtain the average rate of change for interval A. Similarly, calculate the average rate of change for interval B. These two average rates of change serve as the film laying status characteristics for each set of tension data in intervals A and B, respectively.

[0053] In practice, the above steps need to be repeated for each set of tension data in the database. First, the tension trajectories are counted and a histogram is constructed. Then, a boundary segmentation algorithm is used to determine the segmented intervals. The average rate of change of the tension trajectory in each interval is calculated as the mulch film laying status characteristic for that set of data.

[0054] It is important to note that when calculating tension trajectories, the data must be in the correct chronological order to avoid inaccurate histogram construction due to incorrect data sorting. When selecting a boundary segmentation algorithm, it is important to make a reasonable choice based on the characteristics of the data and actual needs. Different algorithms may lead to different segmentation results, thus affecting the subsequent acquisition of state features. For example, for data with large tension fluctuations, a more sensitive segmentation algorithm may be required to capture its distribution characteristics; for data with smaller fluctuations, the choice of segmentation algorithm can be relatively relaxed.

[0055] The number of intervals should be at least two, but the specific number should also be determined based on the data distribution. If the histogram exhibits a multimodal distribution, more intervals may be needed to more accurately describe the distribution characteristics of the tension data. If the histogram distribution is relatively concentrated, two intervals may be sufficient.

[0056] When calculating the rate of change of the tension trajectory, it is also important to determine the time interval. The time interval can be determined based on the frequency of data collection. For example, if the data is collected once per second, the time interval can be is 1 second; if the acquisition frequency is higher, The corresponding value is smaller. Accurate time interval is the basis for calculating the rate of change. If the time interval is calculated incorrectly, the calculation result of the rate of change will be inaccurate, which will affect the accuracy of the state characteristics.

[0057] Through the detailed steps above, we can obtain the corresponding film laying state characteristics for each set of tension data in the database. These state characteristics reflect the changing trend and speed of tension data within different intervals, providing an important basis for subsequent steps such as historical tension data type classification and uniform speed adjustment model training. This enables the model to better understand and process tension data with different characteristics, thereby achieving more accurate uniform speed adjustment for continuous film laying.

[0058] Example 5: In the implementation process of inputting tension feedback data acquired in real time by mulch film laying equipment into a trained uniform speed control model to obtain speed control instructions, a specific mulch film laying scenario is used as an example to illustrate. Assume that the mulch film laying equipment is operating in a certain farmland, collecting real-time tension feedback data during the mulch film laying process. This data is continuously transmitted to the system in the form of a time series. For example, the tension value collected at 10:00:00 is F1, and at 10:00:01 is F2, and so on, forming a continuous tension data sequence.

[0059] The tension feedback data acquired in real time needs to be noise filtered. Since the equipment may be affected by factors such as mechanical vibration and electromagnetic interference in the actual working environment, the collected data may contain noise signals. These noises will interfere with the extraction of effective features, so the data needs to be filtered. For example, if the tension value F at a certain moment suddenly has a spike that is far higher than the normal range, such as the normal tension range is between 50-80, and 150 is collected at a certain moment, this is likely to be noise interference and needs to be processed by a filtering algorithm. Common filtering methods such as moving average filtering can set a window size. For example, the tension data at the current moment and the previous 5 moments are taken, and their average value is calculated as the effective data at the current moment, thereby smoothing the fluctuations caused by the noise.

[0060] After noise filtering, it is necessary to extract effective tension feature values from the data. Effective tension feature values refer to key features that can reflect the state of ground film laying. The extraction of these features is based on the key tension dimensions determined in the training phase. For example, the key tension dimensions determined in the training phase include the mean value, fluctuation amplitude, and rate of change of tension. Therefore, when extracting effective features, calculations need to be performed on these dimensions. For the tension data sequence in the above example, calculate its mean value within a certain time window, such as calculating the average value of all tension data within 1 minute; calculate the fluctuation amplitude of tension, that is, the difference between the maximum and minimum values in the time window; calculate the rate of change of tension, such as the speed of change of tension values at adjacent moments, etc. These calculated values are the effective tension feature values.

[0061] After extracting the effective tension feature values, they are fed into a trained constant-speed control model. This constant-speed control model is derived through previous training based on historical tension data. The model internally establishes mapping rules between tension features and speed control instructions. For example, the model may learn, after training, that when the mean tension value exceeds a certain threshold and fluctuates significantly, the paving speed should be reduced; when the mean tension value is within a normal range and the rate of change is small, the current speed should be maintained.

[0062] After receiving the effective tension characteristic value, the model processes it according to its internal mapping rules and generates corresponding speed adjustment instructions. For example, if the input effective tension characteristic value shows that the current tension average is 75, which is within the normal range of 50-80, but the fluctuation amplitude reaches 30, exceeding the preset fluctuation threshold of 20, the model's mapping rules will generate an instruction to reduce the paving speed, such as adjusting the current speed from 10 meters / minute to 8 meters / minute.

[0063] After generating a speed adjustment command, the system transmits it to the mulch-laying equipment's actuators. Based on the speed adjustment command, the actuators perform a uniform speed adjustment operation for continuous mulch laying. For example, the equipment's drive motor adjusts its speed based on the command, thereby varying the mulch-laying speed and achieving uniform speed adjustment.

[0064] In actual operation, attention must be paid to the effectiveness of noise filtering. Different noise types require different filtering methods. For example, for periodic noise, bandpass filtering can be used; for random noise, moving average filtering or median filtering may be more effective. Therefore, during implementation, it is necessary to select an appropriate filtering algorithm based on the equipment's operating environment and noise characteristics to ensure that the filtered data accurately reflects the tension state of the mulch film.

[0065] Furthermore, the extraction of effective tension feature values must be tailored to the specific application scenario and model requirements. For example, the tension characteristics of mulch film laying may vary in different farmland environments. Therefore, the feature extraction time window size and calculation method must be adjusted based on the actual situation to ensure that the extracted features accurately reflect the current paving status.

[0066] The tension-velocity mapping rules within the model are obtained through training with a large amount of historical data in the early stage. Therefore, during the application of the model, if the adjustment effect is found to be unsatisfactory, it may be necessary to collect data again to optimize the model training in order to update the mapping rules and improve the accuracy and adaptability of the model.

[0067] Through the above detailed implementation method, the tension feedback data collected in real time can be effectively processed and analyzed, and the trained model can be input to generate reasonable speed adjustment instructions, and the equipment can execute the adjustment operation, thereby achieving uniform speed adjustment during the continuous laying of the ground film and ensuring the quality and efficiency of the ground film laying.

[0068] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "includes," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0069] 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. A method for adjusting the uniform speed of continuous paving of ground film based on tension feedback, characterized in that: The method comprises the following steps: Collect the real-time tension feedback data obtained by the ground film laying equipment and the database composed of multiple sets of tension feedback data recorded during the historical laying process; The historical tension data are divided into several types according to the ground film laying state characteristics corresponding to each group of tension data in the database; a uniform speed adjustment model is trained based on the database, and key tension dimensions are selected by analyzing data classification deviations; data weight distribution is calculated based on the characteristic differences of different data on the key tension dimensions; the uniform speed adjustment model is iteratively optimized based on the weight distribution to obtain a trained uniform speed adjustment model; The tension feedback data obtained in real time by the ground film laying equipment is input into the trained uniform speed adjustment model to obtain a speed adjustment instruction, and the ground film continuous laying uniform speed adjustment operation is performed based on the speed adjustment instruction.

2. The method for adjusting the uniform speed of continuous paving of ground film based on tension feedback according to claim 1, characterized in that: The historical tension data is divided into several types according to the ground film laying state characteristics corresponding to each group of tension data in the database, including: For the mth group of historical tension data: Each training stage of the mth group of historical tension data from the first misclassification to the correct classification is recorded as an analysis stage of the mth group of historical tension data; the state feature changes corresponding to the mth group of historical tension data in all analysis stages are arranged in chronological order to obtain a feature change sequence of the mth group of historical tension data; the position index of each value in the feature change sequence is used as the horizontal axis coordinate, and the value corresponding to each position index is used as the vertical axis coordinate to form no less than two coordinate points, all coordinate points are used as input for feature distribution analysis, and each two-dimensional coordinate vector and the corresponding mapping value are obtained; the two-dimensional coordinate vector corresponding to the maximum mapping value is used as the main direction vector, and the arc tangent value of the ratio of the vertical axis component to the horizontal axis component in the main direction vector is used as the feature distribution direction quantity; Calculate the classification evaluation value of the mth group of historical tension data based on the training stage corresponding to the first misclassification of the mth group of historical tension data, the stage interval between the first misclassification and the first correct classification, and the feature distribution direction; The type of the mth group of historical tension data is determined based on the classification evaluation amount.

3. The method for adjusting the uniform speed of continuous paving of ground film based on tension feedback according to claim 2, characterized in that: The calculating of the classification evaluation value of the mth group of historical tension data based on the training stage corresponding to the first misclassification of the mth group of historical tension data, the stage interval between the first misclassification and the first correct classification, and the feature distribution direction quantity includes: Recording the difference between the characteristic distribution direction and the preset reference direction as a first evaluation parameter; Calculate the inverse normalization result of the training stage corresponding to the first misclassification of the mth group of historical tension data, and determine the product of the stage interval between the first misclassification and the first correct classification, the inverse normalization result and the first evaluation parameter as the classification evaluation value of the mth group of historical tension data.

4. The method for adjusting the uniform speed of continuous paving of ground film based on tension feedback according to claim 2, characterized in that: The determining the type of the mth group of historical tension data based on the classification evaluation amount includes: If the classification evaluation value is greater than the preset type classification threshold, the mth group of historical tension data is determined to belong to the high-fluctuation tension type; otherwise, it is determined to belong to the low-fluctuation tension type.

5. The method for adjusting the uniform speed of continuous paving of ground film based on tension feedback according to claim 1, characterized in that: The key tension dimensions are screened by analyzing the data classification deviation, including: For the nth type: Constructing a first feature matrix based on all correctly classified tension data in the nth type, wherein each row in the first feature matrix is a group of correctly classified tension data; processing the first feature matrix using a feature screening algorithm to obtain a first feature screening result, wherein each column of data in the first feature screening result constitutes a first candidate tension dimension; constructing a second feature matrix based on the misclassified tension data in the nth type, wherein each row in the second feature matrix is a group of misclassified tension data; processing the second feature matrix using a feature screening algorithm to obtain a second feature screening result, wherein each column of data in the second feature screening result constitutes a second candidate tension dimension; Matching and parsing the first candidate tension dimension and the second candidate tension dimension are performed, and key tension dimensions are screened based on the matching results.

6. The method for adjusting the uniform speed of continuous paving of ground film based on tension feedback according to claim 5, characterized in that: The matching and parsing of the first candidate tension dimension and the second candidate tension dimension, and screening the key tension dimension based on the matching result, includes: A feature association algorithm is used to match the first candidate tension dimension with the second candidate tension dimension to obtain multiple feature matching groups; the correlation between the two dimensions in each feature matching group is calculated respectively, and the feature matching group with a correlation greater than a preset correlation threshold is used as the target matching group; The average dimension of the two dimensions in each target matching group was taken as a key tension dimension.

7. The method for adjusting the uniform speed of continuous paving of ground film based on tension feedback according to claim 1, characterized in that: The calculation of data weight distribution based on the feature differences of different data on the key tension dimension includes: Correct tension data for group p: Record the misclassified data in the type of the correctly classified tension data of the pth group as the reference data of the correctly classified data of the pth group; calculate the mean of the characteristic values of all reference data of the correctly classified data of the pth group on each key tension dimension, and record the absolute difference between the characteristic value of the correctly classified data of the pth group on each key tension dimension and the corresponding mean as the difference index of the correctly classified data of the pth group on each key tension dimension; The sum of the inverse normalized results of the difference indicators of the p-th group of correctly classified data in all key tension dimensions is respectively used as the weight coefficient of the p-th group of correctly classified data; and the target weight of the p-th group of correctly classified data in each training stage is calculated based on the weight coefficient.

8. The method for adjusting the uniform speed of continuous paving of ground film based on tension feedback according to claim 1, characterized in that: Acquisition of the ground film laying state characteristics corresponding to each set of tension data in the database includes: The ground film laying tension trajectory corresponding to each group of tension data in the database is statistically analyzed to obtain the corresponding trajectory distribution histogram, and the boundary segmentation algorithm is used to segment the histogram to obtain no less than two segmentation intervals; The average value of the tension trajectory change rate of all tension data in each segmented interval is calculated and used as the ground film laying state feature of each group of tension data in the corresponding segmented interval.

9. The method for adjusting the uniform speed of continuous paving of ground film based on tension feedback according to claim 7, characterized in that: The calculating the target weight of the p-th group of correctly classified data in each training stage based on the weight coefficient includes: For any group of correctly classified tension data: multiply the original weight of the group of data in each training stage by the corresponding weight coefficient, which is used as the target weight of the group of data in each training stage.

10. The method for adjusting the uniform speed of continuous paving of ground film based on tension feedback according to claim 1, characterized in that: The method of inputting the tension feedback data obtained in real time by the ground film laying equipment into the trained uniform speed adjustment model to obtain the speed adjustment instruction includes: The tension feedback data acquired in real time is subjected to noise filtering to extract effective tension characteristic values; the effective tension characteristic values are input into the uniform speed adjustment model, and corresponding speed adjustment instructions are generated through the tension-speed mapping rules within the model.

Images