Intelligent monitoring management system and method applied to manure scraper

Abstract

The invention discloses an intelligent monitoring management system and method applied to a manure scraper, and relates to the technical field of artificial intelligence. According to historical work records, a manure quality prediction model and a path generation model are constructed, and a work scoring model is generated; calculating to obtain an operation score through an operation scoring model, screening out an execution path of the manure scraper, collecting operation parameters of the manure scraper on the execution path, calculating to obtain an actual operation score, calculating to obtain an absolute difference value with the operation score, and generating an analysis adjustment model based on the absolute difference value and an analysis time period; real-time working records are collected and analyzed, the updating time of the manure scraper management platform is determined, and through accurate real-time feedback and adjustment, the service life of equipment can be effectively prolonged, the fault occurrence rate is reduced, and long-term efficient operation of the manure scraper is ensured.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, specifically to an intelligent monitoring and management system and method for manure scrapers. Background Technology

[0002] In large-scale sheep farms, manure scraping equipment is a key facility for cleaning manure in the feeding area. Its operation mode is gradually developing from the traditional fixed manure scraping device to unmanned manure scraping vehicles with autonomous walking ability. Compared with the traditional manure scraper that runs along a fixed track or steel wire, the unmanned manure scraper is usually equipped with drive, power supply and basic sensing unit. It can move autonomously in the breeding area according to the preset route to complete the manure cleaning operation, providing a hardware foundation for the flexibility of manure cleaning. In practical applications, existing unmanned manure scraping vehicles only support route cruising and timed operation based on manual settings or simple rules. They lack in-depth analysis of the differences in manure load in different areas. The scraping path, cleaning frequency and operating power are often configured uniformly, making it difficult to dynamically adjust according to changes in manure distribution in the breeding area. Although some unmanned manure scrapers have introduced environmental or load sensors to monitor their status, their control strategies are mostly at the level of passive response. They fail to predict and plan subsequent operation strategies based on the trend of manure production. At the same time, the adjustment of operating parameters and routes is usually carried out according to fixed cycles or manual triggering. When the environment changes steadily or the load changes suddenly, it is easy to cause problems such as over-adjustment or delayed response, which affects cleaning efficiency and equipment energy consumption control. Therefore, how to improve the ability of the manure scraper to perceive and predict changes in manure, and to achieve intelligent and coordinated control of operating path, power configuration and adjustment rhythm, has become an urgent technical problem to be solved in the field of manure scraper management. Summary of the Invention

[0003] The purpose of this invention is to provide an intelligent monitoring and management system and method for manure scrapers, so as to solve the problems raised in the prior art.

[0004] To solve the above-mentioned technical problems, the present invention provides the following technical solution: an intelligent monitoring and management method applied to a manure scraper, the method comprising: Step S100: In the manure scraper management platform, obtain the planar spatial model of the feeding area in the work record, and extract the working space model of the manure scraper according to the functional area of ​​the manure scraper operation, divide it into several subspaces, and generate the time series collection data corresponding to each subspace. Step S200: Based on historical work records, construct a fecal quality prediction model and a path generation model; Step S300: Set the training period, obtain the work records within the training period, calculate the predicted fecal quality of each feature space through the fecal quality prediction model, generate several candidate paths based on the path generation model, collect the operation parameters of the scraper executing the candidate paths, calculate the operation score, and combine it with the feature parameters of the candidate paths to generate an operation score model. Step S400: Calculate the work score using the work scoring model, select the execution path of the manure scraper, collect the work parameters of the manure scraper on the execution path, calculate the actual work score, and calculate the absolute difference between the actual work score and the work score. Based on the absolute difference and the analysis time period, generate an analysis and adjustment model. Step S500: Collect and analyze real-time work records to determine the update time of the manure scraper management platform.

[0005] Furthermore, step S100 includes: Step S101: Obtain the site information of the feeding area where the manure scraper works, construct a planar spatial model of the feeding area based on the site information, collect the amount of feed put into the feeding area and the corresponding amount of feed consumed, collect the operating parameters of the manure scraper during its operation in the feeding area, combine the amount of feed put into the feeding area, the feed consumption parameters and the operating parameters to generate the working record of the manure scraper, and upload it to the manure scraper management platform. Step S102: In the planar space model, divide the functional areas within each feeding area to obtain the functional areas where the manure scraper operates, extract the working space model of the manure scraper, divide the working space model into several subspaces and label them. Step S103: Collect the operation timestamp corresponding to the operation of the manure scraper in each subspace, obtain the manure quality collected by the manure scraper in each subspace, combine the operation timestamp and manure quality to generate time-series collection data corresponding to each subspace; By acquiring information about the feeding area and constructing a planar spatial model, the feed input, feed consumption, and manure scraper operating parameters are combined to generate a complete work record and upload it to the management platform. This enables the data-driven and structured storage of the manure scraper operation process, facilitating subsequent querying, statistics, and traceability analysis. Dividing the feeding area into functional zones in the planar spatial model and further subdividing the manure scraper's operating space into multiple subspaces helps to clarify the operating range of the manure scraper in different functional zones, providing a spatial basis for subsequent accurate analysis of manure generation in different areas.

[0006] Furthermore, step S200 includes: Step S201: In the historical work record set, construct a fecal quality prediction model based on the time-series data collected in each subspace of the historical work record; Step S202: Collect the operating parameters of the manure scraper from the historical work records, and construct the manure scraper path generation model based on the manure quality prediction model; A fecal quality prediction model is constructed based on the time-series data collected in each subspace of the historical work record set. This model not only considers changes in fecal quality but also comprehensively reflects the temporal characteristics and historical operational patterns of different subspaces, thereby effectively improving the prediction accuracy of the amount and distribution of feces in the breeding area and reducing prediction volatility. By collecting the operating parameters of the manure scraper from historical work records and combining them with the manure quality prediction model, a manure scraper path generation model is established. This model ensures that the manure scraper's walking path and operating sequence match the actual manure distribution characteristics, avoiding unreasonable operation problems caused by relying solely on fixed routes or experience settings.

[0007] Furthermore, step S201 includes: Step S201-1: Extract the manure quality collected by the manure scraper in each subspace from the historical work records, collect the number of livestock in the feeding area from the historical work records, normalize the manure quality and the number of livestock respectively, and sum the normalized manure quality and the number of livestock according to preset weights to calculate the manure feature value of each subspace. Summarize the manure feature values ​​corresponding to the historical work records of each subspace to calculate the average manure feature value of each subspace. Set a preset average manure feature threshold, and mark the subspaces that exceed the average manure feature threshold as feature spaces. Step S201-2: In the historical work record, collect the timestamp of the last feed delivery and calculate the feed delivery duration by comparing it with the timestamp of the historical work record. Obtain the weather parameters for the delivery duration and normalize the weather parameters. Calculate the weather score by weighting the normalized weather parameters according to preset weights. Extract the timestamp of the historical work record. Preset a day into several time periods and mark them. Determine the time period corresponding to the historical work record. Obtain the feed consumption quantity within the delivery duration and divide it by the delivery duration to calculate the feed consumption frequency. Summarize the feed consumption frequency of the feed corresponding to the historical work record in each time period. Calculate the average feed consumption frequency for each time period and sort the time periods from low to high according to the average consumption frequency. Collect the position corresponding to each time period and assign values ​​to the time periods according to the position. Step S201-3: Summarize the historical work records of a certain feature space, and normalize the deployment duration, time period value, and weather score respectively. Use the normalized deployment duration, time period value, and weather score as input and the fecal quality of the feature space as output. Train the model through a random forest regression model. The training process is to generate multiple decision trees by random sampling, learn the mapping relationship between input features and fecal quality, and integrate the prediction results of each decision tree to generate a fecal quality prediction model. By normalizing the quality of feces and the number of livestock in historical work records and performing weighted calculations according to preset weights, feces feature values ​​that can reflect the characteristics of feces generation in the subspace are obtained. Then, the feature space is selected by judging the feature average value and threshold, so that the subsequent modeling focuses on the subspace with obvious feces generation characteristics, thereby reducing the interference of irrelevant data and improving the efficiency of data utilization. By incorporating feed delivery duration, weather parameters, and time period divisions, and by normalizing and weighting the relevant parameters, the model can comprehensively reflect the feed consumption rhythm, external environmental conditions, and the impact of different time periods on livestock feeding and excretion behavior, thereby improving the fecal quality prediction model's adaptability to changes in the actual feeding environment. By statistically analyzing the frequency of feed consumption in different time periods, sorting and assigning numerical values, discrete time information is transformed into numerical features with regular patterns. This facilitates machine learning models in effectively learning time factors and improves the model's ability to characterize the temporal patterns of fecal production.

[0008] Furthermore, step S202 includes: Step S202-1: Extract the operating parameters of the manure scraper from the historical work records. The operating parameters include the operating speed, single running time, and motor load parameters of the manure scraper. Multiply the operating speed by the single running time to calculate the theoretical running distance of a single run. Multiply the theoretical running distance by the cleaning width of the manure scraper to calculate the theoretical cleaning area of ​​a single run. Calculate the load correction coefficient based on the motor load parameters. Multiply the theoretical cleaning area by the load correction coefficient to calculate the effective cleaning area of ​​a single run. Summarize the effective cleaning areas corresponding to each single run in the historical work records and calculate the average effective cleaning area. Use the average effective cleaning area as the capacity threshold of the manure scraper within a unit operating cycle. Step S202-2: Based on the fecal predicted quality of each feature space output by the fecal quality prediction model, obtain the spatial positional relationship between each feature space, preset the fecal quality threshold, set the feature space with the fecal predicted quality higher than the fecal quality threshold as the priority cleaning space, and combine the priority cleaning spaces under the condition of meeting the capacity threshold of the manure scraper to construct the operation path model of the manure scraper. By extracting the operating speed, single run duration, and motor load parameters of the manure scraper from historical work records, the theoretical cleaning area per run is calculated. This is then corrected using a load correction factor to obtain the effective cleaning area, thus accurately reflecting the true cleaning capacity of the manure scraper under different operating conditions. The average effective cleaning area obtained by summarizing these parameters serves as a capacity threshold, providing a precise capacity constraint basis for path planning. By combining the fecal quality and spatial location relationships of each feature space output by the fecal quality prediction model, the key cleaning areas can be dynamically determined based on the actual fecal distribution. Under the condition of meeting the capacity threshold of the fecal scraper, the priority cleaning spaces can be combined to make the generated operation path model more in line with the working capacity of the fecal scraper and the fecal distribution characteristics, thus avoiding over-cleaning or omission of cleaning. By combining capacity threshold constraints with priority cleaning space, cleaning efficiency can be maximized within a single operation cycle, reducing unnecessary travel and repeated cleaning, lowering energy consumption and mechanical wear, thereby improving the overall economy and reliability of the system.

[0009] Furthermore, step S300 includes: Step S301: Select several consecutive days as the training period, obtain the work records within the training period, collect the weather parameters, the timestamp of the last feed delivery, and the current timestamp from the work records, calculate the weather score and the duration of feed delivery, determine the value of the time period corresponding to the work records, normalize the duration of feed delivery, the time period value, and the weather score, and input them into the fecal quality prediction model of each feature space to calculate the predicted fecal quality of each feature space. Step S302: Input the predicted quality of feces in each feature space into the running path model to generate several candidate paths, obtain the candidate path selected in the current working path, collect the operating parameters of the manure scraper, including the actual cleaning area, cleaning amount per unit time, motor load, and power consumption, perform normalization calculation on the operating parameters, and perform weighted summation of the normalized operating parameters according to preset weights to calculate the operating score; Step S303: Extract the feature parameters of the candidate path. The feature parameters include path length, number of covered feature spaces, and total predicted fecal quality. The feature parameters of the candidate path are used as input and the job score is used as output. The model is trained by a deep learning model. The training process uses the mean squared error loss function and uses an optimizer to iteratively update the model parameters to minimize the error between the job score output by the model and the actual job score, thereby generating a job score model. By combining factors such as weather parameters, delivery duration, and time period, and inputting them into the fecal quality prediction model, the fecal quality of each feature space can be accurately predicted, providing scientific fecal distribution data for subsequent cleaning path planning. Taking into account the influence of different environmental factors, the cleaning path planning becomes more targeted and reasonable, improving the accuracy of path optimization and operational effectiveness. By combining the feature parameters of each candidate path with the job score and training a deep learning model, a job score model for different job conditions can be generated. This model can adaptively optimize path selection, improve job efficiency, and reduce unnecessary travel and resource waste. By training deep learning models, the error between the output of the assignment score and the actual assignment score is minimized, which ensures the consistency and stability of path selection and assignment evaluation. It can efficiently process a large number of assignment parameters, enhance the accuracy of the score, and help with real-time feedback and dynamic adjustment of assignment strategies.

[0010] Furthermore, step S400 includes: Step S401: Input the feature parameters of the candidate paths for each work record into the job scoring model, calculate the job score, and select the candidate path corresponding to the highest job score as the execution path of the manure scraper. Step S402: Collect the operation parameters of the manure scraper on the execution path, calculate the actual operation score, and calculate the absolute difference with the operation score. Set a threshold for the absolute difference. If the absolute difference exceeds the threshold, mark the work record as an abnormal analysis record. Step S403: Set the time period between two adjacent anomaly analysis records as the analysis time period, obtain the absolute difference of the job score corresponding to each work record within the analysis time period, construct the absolute difference time series set, use the absolute difference time series set as input and the analysis time period as output, train through the Long Short-Term Memory Network model. The training process is to use a loss function to calculate the error between the model output result and the actual analysis result, and update the model parameters through the backpropagation algorithm to generate the analysis adjustment model. By calculating the job score and selecting the candidate path with the highest score as the execution path of the manure scraper, it is possible to ensure that the manure scraper selects the optimal path during operation, maximizing the work efficiency. By combining the characteristic parameters of the path with the scoring model, the accuracy and rationality of path selection are effectively improved, thereby reducing unnecessary waste and time loss in the operation. By calculating the absolute difference between the actual work score and the preset score, and filtering abnormal records according to the threshold, the operating status of the manure scraper can be monitored in real time. If there is a significant deviation or abnormality in the execution path, the system can mark the work record in time and start further abnormal analysis. This can ensure the efficiency and stability of the manure scraper operation and avoid potential failures or operational errors in the long term. By using the time interval between adjacent anomaly analysis records as the analysis time interval, and training the absolute difference time series based on a long short-term memory network, the system can automatically learn the patterns and rules of anomaly occurrence, thereby optimizing the adjustment mechanism. By continuously adjusting and updating the model parameters, the anomaly analysis and adjustment mechanism have stronger adaptive capabilities, improving the system's dynamic adjustment capabilities and decision-making level.

[0011] Furthermore, step S500 includes: Step S501: Based on the normalized real-time delivery duration, real-time time period value, and real-time weather score from the real-time working data, input them into the fecal quality prediction model to calculate the real-time fecal predicted quality for each feature space. Step S502: Input the real-time predicted quality of feces in each feature space into the operation path model to generate several real-time candidate paths. Input the feature parameters of each real-time candidate path into the operation scoring model to calculate the real-time operation score. Select the highest real-time operation score as the real-time execution path of the manure scraper. Step S503: After the manure scraper completes its work, the real-time actual operation score of the manure scraper is calculated, and the real-time absolute difference between the score and the actual operation score is calculated. If the real-time absolute difference exceeds the threshold, the manure scraper management platform is updated immediately. If the real-time absolute difference does not exceed the threshold, a real-time absolute difference time series set is constructed and input into the analysis and adjustment model to calculate the update time duration. The manure scraper management platform is then updated according to the update time duration. By inputting the real-time delivery duration, real-time time period value, and real-time weather score into the fecal quality prediction model, the predicted fecal quality of each feature space can be calculated in real time, providing the scraper with accurate fecal distribution data. The real-time prediction mechanism makes path optimization more flexible and effective, ensuring that the scraper selects the optimal execution path in a dynamically changing environment, thereby improving operational efficiency and accuracy. The feature parameters of each real-time candidate path are input into the job scoring model to calculate the real-time job score. The path with the highest score is selected as the execution path of the manure scraper, so that the manure scraper can be dynamically adjusted according to the real-time job score during the execution process, ensuring the optimal selection of the job path and reducing unnecessary time waste and resource consumption in the operation. By calculating the real-time absolute difference between the real-time operation score and the actual operation score, the operation status of the manure scraper can be monitored in real time. If the difference exceeds the preset threshold, the system will immediately update the management platform and take timely measures to prevent the expansion of operation deviation or potential faults. The early warning mechanism ensures the stability of equipment operation and avoids long-term inefficient operation or equipment damage. When the real-time absolute difference does not exceed the threshold, the update time is calculated by constructing a real-time absolute difference time series set, combining it with the analysis and adjustment model for training, and automatically updating the manure scraper management platform. This enables real-time adjustment of operation strategies and path selection, ensuring the efficient operation of the manure scraper under different operating environments and conditions, and improving the system's adaptability and intelligence level. Real-time performance evaluation and adjustment mechanisms can reduce errors during manure scraper operation, improve operational accuracy and quality, and effectively extend equipment lifespan, reduce failure rate, and ensure long-term efficient operation of the manure scraper through precise real-time feedback and adjustments.

[0012] To better implement the above methods, an intelligent monitoring and management system for manure scrapers is proposed. The system includes a time-series data collection module, a manure quality prediction model and a path generation model module, an operation scoring model module, an analysis and adjustment model module, and an update time module. Time-series data collection module: In the manure scraper management platform, the planar spatial model of the feeding area in the work record is obtained, and the working space model of the manure scraper is extracted according to the functional area of ​​the manure scraper operation, and divided into several subspaces, generating time-series data collection data corresponding to each subspace; Fecal quality prediction model and path generation model module: Construct fecal quality prediction model and path generation model based on historical work records; The job scoring model module sets a training period, acquires work records within the training period, calculates the predicted quality of feces in each feature space through the feces quality prediction model, generates several candidate paths based on the path generation model, collects the job parameters of the scraper executing the candidate paths, calculates the job score, and combines it with the feature parameters of the candidate paths to generate the job scoring model. Analysis and Adjustment Model Module: Through the operation scoring model, the operation score is calculated, the execution path of the manure scraper is selected, the operation parameters of the manure scraper on the execution path are collected, the actual operation score is calculated, and the absolute difference between the actual operation score and the operation score is calculated. Based on the absolute difference and the analysis time period, the analysis and adjustment model is generated. Update Time Module: Collects and analyzes real-time work records to determine the update time of the manure scraper management platform.

[0013] Furthermore, the update time module includes a unit for determining the real-time execution path and a unit for determining the update time: Determine the real-time execution path unit: Based on the normalized real-time delivery time, real-time time period value, and real-time weather score from the real-time work data, input them into the fecal quality prediction model to calculate the real-time fecal prediction quality for each feature space. Input the real-time fecal prediction quality for each feature space into the operation path model to generate several real-time candidate paths. Input the feature parameters of each real-time candidate path into the operation scoring model to calculate the real-time operation score. Select the highest real-time operation score as the real-time execution path of the manure scraper. Determine the update time unit: After the manure scraper completes its work, calculate the real-time actual operation score of the manure scraper, calculate the real-time absolute difference between the score and the actual operation score, and if the real-time absolute difference exceeds the threshold, immediately update the manure scraper management platform. If the real-time absolute difference does not exceed the threshold, construct a real-time absolute difference time series set and input it into the analysis and adjustment model to calculate the update time duration, and update the manure scraper management platform according to the update time duration.

[0014] Compared with the prior art, the beneficial effects of the present invention are as follows: by constructing a fecal quality prediction model, the fecal quality of each functional area can be accurately predicted, providing precise data support for the path planning of the fecal scraper. By combining it with the path generation model, the working path of the fecal scraper is optimized, ensuring that it can prioritize the treatment of areas with large amounts of feces during the cleaning process, thereby improving cleaning efficiency. The use of historical work records and fecal quality prediction makes the path planning more intelligent, enabling dynamic adjustments for different environments and working conditions, rather than a single fixed path, which improves the working efficiency of the fecal scraper. By evaluating each candidate path through a task scoring model and adjusting the task path in real time based on the differences in the actual task process, abnormal tasks can be quickly detected, and task parameters and path planning can be adjusted based on the analysis results, thereby reducing errors and resource waste. By leveraging big data and deep learning models, patterns can be extracted from historical work records. Combined with real-time data, accurate predictions and optimizations can be made, gradually improving the operational quality and intelligence of the manure scraper. Compared to traditional manual intervention or single mechanical control, this method is more efficient and precise. Attached Figure Description

[0015] Figure 1 This is a flowchart illustrating an intelligent monitoring and management method for a manure scraper according to the present invention. Figure 2 This is a schematic diagram of the structure of an intelligent monitoring and management system for a manure scraper according to the present invention. Detailed Implementation

[0016] 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.

[0017] Please see Figure 1 and Figure 2 This invention provides a technical solution: an intelligent monitoring and management method for manure scrapers, the method comprising: Step S100: In the manure scraper management platform, obtain the planar spatial model of the feeding area in the work record, and extract the working space model of the manure scraper according to the functional area of ​​the manure scraper operation, divide it into several subspaces, and generate the time series collection data corresponding to each subspace. Step S100 includes: Step S101: Obtain the site information of the feeding area where the manure scraper works, construct a planar spatial model of the feeding area based on the site information, collect the amount of feed put into the feeding area and the corresponding amount of feed consumed, collect the operating parameters of the manure scraper during its operation in the feeding area, combine the amount of feed put into the feeding area, the feed consumption parameters and the operating parameters to generate the working record of the manure scraper, and upload it to the manure scraper management platform. Step S102: In the planar space model, divide the functional areas within each feeding area to obtain the functional areas where the manure scraper operates, extract the working space model of the manure scraper, divide the working space model into several subspaces and label them. Step S103: Collect the operation timestamp corresponding to the operation of the manure scraper in each subspace, obtain the manure quality collected by the manure scraper in each subspace, and combine the operation timestamp and manure quality to generate time-series collection data corresponding to each subspace.

[0018] Step S200: Based on historical work records, construct a fecal quality prediction model and a path generation model; Step S200 includes: Step S201: In the historical work record set, construct a fecal quality prediction model based on the time-series data collected in each subspace of the historical work record; Step S201 includes: Step S201-1: Extract the manure quality collected by the manure scraper in each subspace from the historical work records, collect the number of livestock in the feeding area from the historical work records, normalize the manure quality and the number of livestock respectively, and sum the normalized manure quality and the number of livestock according to preset weights to calculate the manure feature value of each subspace. Summarize the manure feature values ​​corresponding to the historical work records of each subspace to calculate the average manure feature value of each subspace. Set a preset average manure feature threshold, and mark the subspaces that exceed the average manure feature threshold as feature spaces. Step S201-2: In the historical work record, collect the timestamp of the last feed delivery and calculate the feed delivery duration by comparing it with the timestamp of the historical work record. Obtain the weather parameters for the delivery duration and normalize the weather parameters. Calculate the weather score by weighting the normalized weather parameters according to preset weights. Extract the timestamp of the historical work record. Preset a day into several time periods and mark them. Determine the time period corresponding to the historical work record. Obtain the feed consumption quantity within the delivery duration and divide it by the delivery duration to calculate the feed consumption frequency. Summarize the feed consumption frequency of the feed corresponding to the historical work record in each time period. Calculate the average feed consumption frequency for each time period and sort the time periods from low to high according to the average consumption frequency. Collect the position corresponding to each time period and assign values ​​to the time periods according to the position. Step S201-3: Summarize the historical work records of a certain feature space, and normalize the deployment duration, time period value, and weather score respectively. Use the normalized deployment duration, time period value, and weather score as input and the fecal quality of the feature space as output. Train the model through a random forest regression model. The training process is to generate multiple decision trees by random sampling, learn the mapping relationship between input features and fecal quality, and integrate the prediction results of each decision tree to generate a fecal quality prediction model. For example, select 7 days from May 1st to May 7th, perform 4 manure scraping operations per day, and count the total mass of manure in each subspace each day; On May 1, the fecal mass of subspace s1 was 245, the fecal mass of subspace s2 was 230, the fecal mass of subspace s3 was 260, the fecal mass of subspace s4 was 240, the fecal mass of subspace s5 was 215, and the fecal mass of subspace s6 was 235. On May 2nd, the fecal mass of subspace s1 is 250, the fecal mass of subspace s2 is 228, the fecal mass of subspace s3 is 268, the fecal mass of subspace s4 is 242, the fecal mass of subspace s5 is 220, and the fecal mass of subspace s6 is 238. On May 3rd, the fecal mass of subspace s1 was 248, the fecal mass of subspace s2 was 232, the fecal mass of subspace s3 was 270, the fecal mass of subspace s4 was 245, the fecal mass of subspace s5 was 218, and the fecal mass of subspace s6 was 240. On May 4th, the fecal mass of subspace s1 was 252, the fecal mass of subspace s2 was 235, the fecal mass of subspace s3 was 275, the fecal mass of subspace s4 was 248, the fecal mass of subspace s5 was 222, and the fecal mass of subspace s6 was 243. On May 5th, the fecal mass of subspace s1 is 255, the fecal mass of subspace s2 is 238, the fecal mass of subspace s3 is 278, the fecal mass of subspace s4 is 250, the fecal mass of subspace s5 is 225, and the fecal mass of subspace s6 is 245. On May 6th, the fecal mass of subspace s1 was 258, the fecal mass of subspace s2 was 240, the fecal mass of subspace s3 was 280, the fecal mass of subspace s4 was 252, the fecal mass of subspace s5 was 228, and the fecal mass of subspace s6 was 248. The fecal mass of subspace s1 on the date 05-07 is 260, the fecal mass of subspace s2 is 242, the fecal mass of subspace s3 is 285, the fecal mass of subspace s4 is 255, the fecal mass of subspace s5 is 230, and the fecal mass of subspace s6 is 250. Each subspace corresponds to a pen of 50 Hu sheep. Taking the subspace s3 of dates 05-07 as an example, the calculated manure quality is normalized to 1, the livestock number is normalized to 1, and the calculated manure characteristic value is 1; The average data of fecal characteristic values ​​over 7 days are as follows: The average fecal characteristic value of subspace s1 is 0.84; The average fecal characteristic value of subspace s2 is 0.76; The average fecal characteristic value of subspace s3 is 0.95; The average fecal characteristic value of subspace s4 is 0.88; The average fecal characteristic value of subspace s5 is 0.70; The average fecal characteristic value of subspace s6 is 0.82; If the average threshold for fecal features is preset to 0.90, then s3 is the feature space; Assume the most recent feed delivery time was 06:00:00 on May 7th; The time range of time period T1 is 00:00–06:00; The time range for time period T2 is 06:00–12:00; The time range for time period T3 is 12:00–18:00; The time range for time period T4 is 18:00–24:00; Historical feed consumption statistics are as follows: The average feed consumption in time period T1 was 120, the feeding time was 6, and the consumption frequency was 20. The average feed consumption in time period T2 was 520, the feeding duration was 6, and the consumption frequency was 86.7. The average feed consumption during time period T3 was 480, the feeding duration was 6, and the consumption frequency was 80. The average feed consumption during time period T4 was 260, the feeding duration was 6, and the consumption frequency was 43.3. Sort by consumption frequency from low to high: T1 < T4 < T3 < T2; Then assign values ​​to the time periods: T1 is 1, T4 is 2, T3 is 3, and T2 is 4. Assuming that the normalized duration of delivery, time period values, and weather scores are used as inputs, and the fecal quality is used as output, the original training dataset is constructed. A subset of size Y is randomly selected from the original training dataset, and this subset will be used as the training data for a certain tree. During the construction of each tree, each time a node is split, a subset of all features is randomly selected to determine the best splitting feature. Each tree grows continuously through recursive splitting until a stopping condition is met, such as the depth of the tree or the number of samples in each leaf node. At each split, features and split points that minimize the prediction error are selected. Each tree is trained as a regression tree. In the regression task, each leaf node stores a value, which is the mean of the leaf node samples. This value is used as the predicted value for the corresponding tree. The input features are fed into each regression tree, and the predicted value of each regression tree is calculated. The predicted values ​​are then aggregated to calculate the average prediction value, which is used as the predicted fecal quality to generate a fecal quality prediction model.

[0019] Step S202: Collect the operating parameters of the manure scraper from the historical work records, and construct the manure scraper path generation model based on the manure quality prediction model; Step S202 includes: Step S202-1: Extract the operating parameters of the manure scraper from the historical work records. The operating parameters include the operating speed, single running time, and motor load parameters of the manure scraper. Multiply the operating speed by the single running time to calculate the theoretical running distance of a single run. Multiply the theoretical running distance by the cleaning width of the manure scraper to calculate the theoretical cleaning area of ​​a single run. Calculate the load correction coefficient based on the motor load parameters. Multiply the theoretical cleaning area by the load correction coefficient to calculate the effective cleaning area of ​​a single run. Summarize the effective cleaning areas corresponding to each single run in the historical work records and calculate the average effective cleaning area. Use the average effective cleaning area as the capacity threshold of the manure scraper within a unit operating cycle. Step S202-2: Based on the predicted fecal quality of each feature space output by the fecal quality prediction model, obtain the spatial relationship between each feature space, preset the fecal quality threshold, set the feature space with the predicted fecal quality higher than the fecal quality threshold as the priority cleaning space, and combine the priority cleaning spaces under the condition of meeting the capacity threshold of the sludge scraper to construct the operation path model of the sludge scraper.

[0020] Step S300: Set the training period, obtain the work records within the training period, calculate the predicted fecal quality of each feature space through the fecal quality prediction model, generate several candidate paths based on the path generation model, collect the operation parameters of the scraper executing the candidate paths, calculate the operation score, and combine it with the feature parameters of the candidate paths to generate an operation score model. Step S300 includes: Step S301: Select several consecutive days as the training period, obtain the work records within the training period, collect the weather parameters, the timestamp of the last feed delivery, and the current timestamp from the work records, calculate the weather score and the duration of feed delivery, determine the value of the time period corresponding to the work records, normalize the duration of feed delivery, the time period value, and the weather score, and input them into the fecal quality prediction model of each feature space to calculate the predicted fecal quality of each feature space. Step S302: Input the predicted quality of feces in each feature space into the running path model to generate several candidate paths, obtain the candidate path selected in the current working path, collect the operating parameters of the manure scraper, including the actual cleaning area, cleaning amount per unit time, motor load, and power consumption, perform normalization calculation on the operating parameters, and perform weighted summation of the normalized operating parameters according to preset weights to calculate the operating score; Step S303: Extract the feature parameters of the candidate path. The feature parameters include path length, number of covered feature spaces, and total predicted fecal quality. The feature parameters of the candidate path are used as input and the job score is used as output. The model is trained by a deep learning model. The training process uses the mean squared error loss function and uses an optimizer to iteratively update the model parameters to minimize the error between the job score output by the model and the actual job score, thereby generating a job score model. For example, a neural network is a feedforward neural network, which consists of several layers. The output of each layer is calculated by the output of the previous layer and the weights and biases of the current layer through an activation function. Assume that the neural network has L layers, the input of each layer l is h(l-1), and the output of each layer is h(l). For the calculation process of the l-th layer, assuming that this layer has Nl neurons, the calculation formula is as follows: ; Where wl is the weight matrix of the l-th layer, and bl is the bias term of the l-th layer. [] represents the activation function; After L layers of computation, the output of the last layer is the predicted job score y1. Assuming the last layer of the neural network is a fully connected layer with a single neuron, the final output can be expressed as: y1 = wL × h(L-1) + bL, where wL represents the weight matrix of the last layer and bL represents the bias term of the last layer; To train the neural network, we use the mean squared error loss function to analyze the error between the predicted value y1 and the actual job score y2, and update the parameters of the neural network by minimizing the loss function. The gradient descent algorithm is used for parameter optimization. The optimization steps are to calculate the gradient of the loss function with respect to each parameter, adjust the parameter values ​​according to the gradient, calculate the gradient through the backpropagation algorithm, and update the weights and biases of each layer of the neural network until the loss function converges to the minimum value.

[0021] Step S400: Calculate the work score using the work scoring model, select the execution path of the manure scraper, collect the work parameters of the manure scraper on the execution path, calculate the actual work score, and calculate the absolute difference between the actual work score and the work score. Based on the absolute difference and the analysis time period, generate an analysis and adjustment model. Step S400 includes: Step S401: Input the feature parameters of the candidate paths for each work record into the job scoring model, calculate the job score, and select the candidate path corresponding to the highest job score as the execution path of the manure scraper. Step S402: Collect the operation parameters of the manure scraper on the execution path, calculate the actual operation score, and calculate the absolute difference with the operation score. Set a threshold for the absolute difference. If the absolute difference exceeds the threshold, mark the work record as an abnormal analysis record. Step S403: Set the time period between two adjacent anomaly analysis records as the analysis time period, obtain the absolute difference of the job score corresponding to each work record within the analysis time period, construct the absolute difference time series set, use the absolute difference time series set as input and the analysis time period as output, train through the Long Short-Term Memory Network model. The training process is to use a loss function to calculate the error between the model output result and the actual analysis result, and update the model parameters through the backpropagation algorithm to generate the analysis adjustment model. For example, a long short-term memory network can be constructed to predict and analyze the duration of time periods using time-series data; In an LSTM network, the forget gate, input gate, candidate memory units, memory state update, output gate, and hidden state output are computed. Through these operations, the network completes the modeling of temporal features. After processing the complete absolute difference time series, the hidden state h1 of the last time step is taken, and the predicted update time is obtained through a fully connected layer: T = Wy × h1 + by, where T represents the predicted update time, and Wy and by represent the weight matrix and bias of the output layer, respectively. During training, mean squared error is used as the loss function. The loss function is iteratively optimized by an optimizer to adjust the parameters of the long short-term memory network model and generate an analysis and adjustment model.

[0022] Step S500: Collect and analyze real-time work records to determine the update time of the manure scraper management platform; Step S500 includes: Step S501: Based on the normalized real-time delivery duration, real-time time period value, and real-time weather score from the real-time working data, input them into the fecal quality prediction model to calculate the real-time fecal predicted quality for each feature space. Step S502: Input the real-time predicted quality of feces in each feature space into the operation path model to generate several real-time candidate paths. Input the feature parameters of each real-time candidate path into the operation scoring model to calculate the real-time operation score. Select the highest real-time operation score as the real-time execution path of the manure scraper. Step S503: After the manure scraper completes its work, the real-time actual operation score of the manure scraper is calculated, and the real-time absolute difference between the score and the actual operation score is calculated. If the real-time absolute difference exceeds the threshold, the manure scraper management platform is updated immediately. If the real-time absolute difference does not exceed the threshold, a real-time absolute difference time series set is constructed and input into the analysis and adjustment model to calculate the update time duration. The manure scraper management platform is then updated according to the update time duration.

[0023] To better implement the above methods, an intelligent monitoring and management system for manure scrapers is proposed. The system includes a time-series data collection module, a manure quality prediction model and a path generation model module, an operation scoring model module, an analysis and adjustment model module, and an update time module. Time-series data collection module: In the manure scraper management platform, the planar spatial model of the feeding area in the work record is obtained, and the working space model of the manure scraper is extracted according to the functional area of ​​the manure scraper operation, and divided into several subspaces, generating time-series data collection data corresponding to each subspace; Fecal quality prediction model and path generation model module: Construct fecal quality prediction model and path generation model based on historical work records; The job scoring model module sets a training period, acquires work records within the training period, calculates the predicted quality of feces in each feature space through the feces quality prediction model, generates several candidate paths based on the path generation model, collects the job parameters of the scraper executing the candidate paths, calculates the job score, and combines it with the feature parameters of the candidate paths to generate the job scoring model. Analysis and Adjustment Model Module: Through the operation scoring model, the operation score is calculated, the execution path of the manure scraper is selected, the operation parameters of the manure scraper on the execution path are collected, the actual operation score is calculated, and the absolute difference between the actual operation score and the operation score is calculated. Based on the absolute difference and the analysis time period, the analysis and adjustment model is generated. Update Time Module: Collects and analyzes real-time work records to determine the update time of the manure scraper management platform; The update time module includes a unit for determining the real-time execution path and a unit for determining the update time. Determine the real-time execution path unit: Based on the normalized real-time delivery time, real-time time period value, and real-time weather score from the real-time work data, input them into the fecal quality prediction model to calculate the real-time fecal prediction quality for each feature space. Input the real-time fecal prediction quality for each feature space into the operation path model to generate several real-time candidate paths. Input the feature parameters of each real-time candidate path into the operation scoring model to calculate the real-time operation score. Select the highest real-time operation score as the real-time execution path of the manure scraper. Determine the update time unit: After the manure scraper completes its work, calculate the real-time actual operation score of the manure scraper, calculate the real-time absolute difference between the score and the actual operation score, and if the real-time absolute difference exceeds the threshold, immediately update the manure scraper management platform. If the real-time absolute difference does not exceed the threshold, construct a real-time absolute difference time series set and input it into the analysis and adjustment model to calculate the update time, and update the manure scraper management platform according to the update time.

[0024] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. An intelligent monitoring and management method for manure scrapers, characterized in that, The methods include: Step S100: In the manure scraper management platform, obtain the planar spatial model of the feeding area in the work record, and extract the working space model of the manure scraper according to the functional area of ​​the manure scraper operation, divide it into several subspaces, and generate the time series collection data corresponding to each subspace. Step S200: Based on historical work records, construct a fecal quality prediction model and a path generation model; Step S300: Set the training period, obtain the work records within the training period, calculate the predicted fecal quality of each feature space through the fecal quality prediction model, generate several candidate paths based on the path generation model, collect the operation parameters of the scraper executing the candidate paths, calculate the operation score, and combine it with the feature parameters of the candidate paths to generate an operation score model. Step S400: Calculate the work score using the work scoring model, select the execution path of the manure scraper, collect the work parameters of the manure scraper on the execution path, calculate the actual work score, and calculate the absolute difference between the actual work score and the work score. Based on the absolute difference and the analysis time period, generate an analysis and adjustment model. Step S500: Collect and analyze real-time work records to determine the update time of the manure scraper management platform.

2. The intelligent monitoring and management method for a manure scraper according to claim 1, characterized in that, Step S100 includes the following steps: Step S101: Obtain the site information of the feeding area where the manure scraper works, construct a planar spatial model of the feeding area based on the site information, collect the amount of feed put into the feeding area and the corresponding amount of feed consumed, collect the operating parameters of the manure scraper during its operation in the feeding area, combine the amount of feed put into the feeding area, the feed consumption parameters and the operating parameters to generate the working record of the manure scraper, and upload it to the manure scraper management platform. Step S102: In the planar space model, divide the functional areas within each feeding area to obtain the functional areas where the manure scraper operates, extract the working space model of the manure scraper, divide the working space model into several subspaces and label them. Step S103: Collect the operation timestamp corresponding to the operation of the manure scraper in each subspace, obtain the manure quality collected by the manure scraper in each subspace, and combine the operation timestamp and manure quality to generate time-series collection data corresponding to each subspace.

3. The intelligent monitoring and management method for a manure scraper according to claim 2, characterized in that, Step S200 includes the following steps: Step S201: In the historical work record set, construct a fecal quality prediction model based on the time-series data collected in each subspace of the historical work record; Step S202: Collect the operating parameters of the manure scraper from the historical work records, and construct the manure scraper path generation model based on the manure quality prediction model.

4. The intelligent monitoring and management method for a manure scraper according to claim 3, characterized in that, The fecal quality prediction model constructed in step S201 includes the following steps: Step S201-1: Extract the manure quality collected by the manure scraper in each subspace from the historical work records, collect the number of livestock in the feeding area from the historical work records, normalize the manure quality and the number of livestock respectively, and sum the normalized manure quality and the number of livestock according to preset weights to calculate the manure feature value of each subspace. Summarize the manure feature values ​​corresponding to the historical work records of each subspace to calculate the average manure feature value of each subspace. Set a preset average manure feature threshold, and mark the subspaces that exceed the average manure feature threshold as feature spaces. Step S201-2: In the historical work record, collect the timestamp of the last feed delivery and calculate the feed delivery duration by comparing it with the timestamp of the historical work record. Obtain the weather parameters for the delivery duration and normalize the weather parameters. Calculate the weather score by weighting the normalized weather parameters according to preset weights. Extract the timestamp of the historical work record. Preset a day into several time periods and mark them. Determine the time period corresponding to the historical work record. Obtain the feed consumption quantity within the delivery duration and divide it by the delivery duration to calculate the feed consumption frequency. Summarize the feed consumption frequency of the feed corresponding to the historical work record in each time period. Calculate the average feed consumption frequency for each time period and sort the time periods from low to high according to the average consumption frequency. Collect the position corresponding to each time period and assign values ​​to the time periods according to the position. Step S201-3: Summarize the historical work records of a certain feature space, and normalize the deployment duration, time period value, and weather score respectively. Use the normalized deployment duration, time period value, and weather score as input and the fecal quality of the feature space as output. Train the model through a random forest regression model. The training process involves generating multiple decision trees by random sampling, learning the mapping relationship between input features and fecal quality, and integrating the prediction results of each decision tree to generate a fecal quality prediction model.

5. The intelligent monitoring and management method for a manure scraper according to claim 3, characterized in that, The construction of the path generation model in step S202 includes the following steps: Step S202-1: Extract the operating parameters of the manure scraper from the historical work records. The operating parameters include the operating speed, single running time, and motor load parameters of the manure scraper. Multiply the operating speed by the single running time to calculate the theoretical running distance of a single run. Multiply the theoretical running distance by the cleaning width of the manure scraper to calculate the theoretical cleaning area of ​​a single run. Calculate the load correction coefficient based on the motor load parameters. Multiply the theoretical cleaning area by the load correction coefficient to calculate the effective cleaning area of ​​a single run. Summarize the effective cleaning areas corresponding to each single run in the historical work records and calculate the average effective cleaning area. Use the average effective cleaning area as the capacity threshold of the manure scraper within a unit operating cycle. Step S202-2: Based on the predicted fecal quality of each feature space output by the fecal quality prediction model, obtain the spatial relationship between each feature space, preset the fecal quality threshold, set the feature space with the predicted fecal quality higher than the fecal quality threshold as the priority cleaning space, and combine the priority cleaning spaces under the condition of meeting the capacity threshold of the sludge scraper to construct the operation path model of the sludge scraper.

6. The intelligent monitoring and management method for a manure scraper according to claim 1, characterized in that, Step S300 includes the following steps: Step S301: Select several consecutive days as the training period, obtain the work records within the training period, collect the weather parameters, the timestamp of the last feed delivery, and the current timestamp from the work records, calculate the weather score and the duration of feed delivery, determine the value of the time period corresponding to the work records, normalize the duration of feed delivery, the time period value, and the weather score, and input them into the fecal quality prediction model of each feature space to calculate the predicted fecal quality of each feature space. Step S302: Input the predicted quality of feces in each feature space into the running path model to generate several candidate paths, obtain the candidate path selected in the current working path, collect the operating parameters of the manure scraper, including the actual cleaning area, cleaning amount per unit time, motor load, and power consumption, perform normalization calculation on the operating parameters, and perform weighted summation of the normalized operating parameters according to preset weights to calculate the operating score; Step S303: Extract the feature parameters of the candidate paths, including path length, number of covered feature spaces, and total predicted fecal quality. Using the feature parameters of the candidate paths as input and the job score as output, train the model through a deep learning model. The training process uses the mean squared error loss function and an optimizer to iteratively update the model parameters to minimize the error between the job score output by the model and the actual job score, thereby generating a job score model.

7. The intelligent monitoring and management method for a manure scraper according to claim 6, characterized in that, Step S400 includes the following steps: Step S401: Input the feature parameters of the candidate paths for each work record into the job scoring model, calculate the job score, and select the candidate path corresponding to the highest job score as the execution path of the manure scraper. Step S402: Collect the operation parameters of the manure scraper on the execution path, calculate the actual operation score, and calculate the absolute difference with the operation score. Set a threshold for the absolute difference. If the absolute difference exceeds the threshold, mark the work record as an abnormal analysis record. Step S403: Set the time period between two adjacent anomaly analysis records as the analysis time period, obtain the absolute difference of the job score corresponding to each work record within the analysis time period, construct the absolute difference time series set, use the absolute difference time series set as input and the analysis time period as output, train through a long short-term memory network model. The training process is to use a loss function to calculate the error between the model output result and the actual analysis result, and update the model parameters through the backpropagation algorithm to generate an analysis and adjustment model.

8. The intelligent monitoring and management method for a manure scraper according to claim 7, characterized in that, Step S500 includes the following steps: Step S501: Based on the normalized real-time delivery duration, real-time time period value, and real-time weather score from the real-time working data, input them into the fecal quality prediction model to calculate the real-time fecal predicted quality for each feature space. Step S502: Input the real-time predicted quality of feces in each feature space into the operation path model to generate several real-time candidate paths. Input the feature parameters of each real-time candidate path into the operation scoring model to calculate the real-time operation score. Select the highest real-time operation score as the real-time execution path of the manure scraper. Step S503: After the manure scraper completes its work, the real-time actual operation score of the manure scraper is calculated, and the real-time absolute difference between the score and the actual operation score is calculated. If the real-time absolute difference exceeds the threshold, the manure scraper management platform is updated immediately. If the real-time absolute difference does not exceed the threshold, a real-time absolute difference time series set is constructed and input into the analysis and adjustment model to calculate the update time duration. The manure scraper management platform is then updated according to the update time duration.

9. An intelligent monitoring and management system for a manure scraper, used to implement the intelligent monitoring and management method for a manure scraper as described in any one of claims 1-8, characterized in that, The system includes a time-series data collection module, a fecal quality prediction model and a path generation model module, a job scoring model module, an analysis and adjustment model module, and an update time module. The time-series data collection module: In the manure scraper management platform, it obtains the planar spatial model of the feeding area from the work record, and extracts the working space model of the manure scraper according to the functional area of ​​the manure scraper operation, and divides it into several subspaces, generating time-series data collection data corresponding to each subspace; The fecal quality prediction model and path generation model module: constructs a fecal quality prediction model and a path generation model based on historical work records; The job scoring model module: sets a training period, acquires work records within the training period, calculates the predicted fecal quality for each feature space through the fecal quality prediction model, generates several candidate paths based on the path generation model, collects the job parameters of the scraper executing the candidate paths, calculates the job score, and combines it with the feature parameters of the candidate paths to generate the job scoring model. The analysis and adjustment model module calculates the operation score through the operation scoring model, selects the execution path of the manure scraper, collects the operation parameters of the manure scraper on the execution path, calculates the actual operation score, calculates the absolute difference between the actual operation score and the operation score, and generates the analysis and adjustment model based on the absolute difference and the analysis time period. The update time module collects and analyzes real-time work records to determine the update time of the manure scraper management platform.

10. The intelligent monitoring and management system for a manure scraper according to claim 9, characterized in that, The update time module includes a real-time execution path determination unit and an update time determination unit: The real-time execution path determination unit: Based on the normalized real-time delivery time, real-time time period value, and real-time weather score from the real-time work data, input them into the fecal quality prediction model to calculate the real-time fecal prediction quality for each feature space. Input the real-time fecal prediction quality for each feature space into the operation path model to generate several real-time candidate paths. Input the feature parameters of each real-time candidate path into the operation scoring model to calculate the real-time operation score. Select the highest real-time operation score as the real-time execution path of the manure scraper. The determination of the update time unit is as follows: After the manure scraper completes its work, the real-time actual operation score of the manure scraper is calculated, and the real-time absolute difference between the score and the actual operation score is calculated. If the real-time absolute difference exceeds the threshold, the manure scraper management platform is updated immediately. If the real-time absolute difference does not exceed the threshold, a real-time absolute difference time series set is constructed and input into the analysis and adjustment model to calculate the update time duration. The manure scraper management platform is then updated according to the update time duration.

Images