Distributed automatic batching control system and method

Abstract

The invention relates to the field of distributed batching control, and particularly discloses a distributed automatic batching control system and method. The method comprises the following steps: firstly, acquiring a feed crushing monitoring video acquired by a feed monitoring module in the distributed system, crushed raw material state data acquired by a raw material information extraction module in the distributed system and rotating speed values of a crusher at a plurality of preset time points acquired by a crusher rotating speed extraction module in the distributed system; and finally, a classification result is obtained through a classifier to judge whether the rotating speed of the crusher needs to be adjusted, so that the raw materials are crushed to the expected granularity, and the quality and the consistency of the feed are improved.

Description

Technical Field

[0001] The present application relates to the field of distributed batching control, and more specifically, to a distributed automatic batching control system and method. Background Art

[0002] Feed is a mixture that provides the nutrients required for animals. Its main purpose is to support the health, growth and production performance of animals. Depending on the type, age and production goals of the animal, the feed formula will vary, and usually includes grains, protein sources, minerals, vitamins and other nutritional additives. The feed batching steps usually include crushing, screening, mixing and metering of raw materials. First, the raw materials are refined into uniform particles through a grinder for better mixing. The crushed raw materials are then passed through a screening device to remove particles that do not meet the specifications. Then, the system automatically weighs each raw material according to the set formula ratio, accurately batches the ingredients, and finally mixes the batched raw materials evenly through a mixer to ensure the quality and consistency of the feed.

[0003] Among them, crushing is an important process in the feed batching process. In traditional processing and crushing, the raw materials are poured into the crusher and crushed according to the preset program. If the crushing situation is not understood and the crushing particle size of the raw materials is not well controlled, the quality of the produced feed is often not high.

[0004] Therefore, a distributed automatic batching control system and method are desired. Summary of the invention

[0005] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides a distributed automatic batching control system and method, which first obtains the feed crushing monitoring video collected by the feed monitoring module in the distributed system, the crushing raw material status data collected by the raw material information extraction module in the distributed system, and the speed values ​​of the grinder at multiple predetermined time points collected by the grinder speed extraction module in the distributed system, and then uses deep learning technology to perform feature extraction and association analysis on the three, and finally obtains the classification result through a classifier to determine whether the grinder speed needs to be adjusted, thereby ensuring that the raw material is crushed to the expected particle size and improving the quality and consistency of the feed.

[0006] According to one aspect of the present application, a distributed automatic batching control system is provided, comprising:

[0007] A distributed batching data acquisition module is used to acquire feed crushing monitoring video collected by a feed monitoring module in a distributed system, crushed raw material status data collected by a raw material information extraction module in a distributed system, and the speed values ​​of a crusher at multiple predetermined time points collected by a crusher speed extraction module in a distributed system;

[0008] A distributed batching data extraction module is used to extract a feed crushing state associated feature vector and a grinder speed feature vector from the feed crushing monitoring video collected by the feed monitoring module in the distributed system, the crushing raw material state data collected by the raw material information extraction module in the distributed system, and the grinder speed values ​​at multiple predetermined time points collected by the grinder speed extraction module in the distributed system;

[0009] The grinder speed adjustment analysis module is used to determine whether the grinder speed needs to be adjusted based on the feed grinding state associated feature vector and the grinder speed feature vector.

[0010] According to another aspect of the present application, a distributed automatic batching control method is provided, comprising:

[0011] Acquire feed crushing monitoring video collected by a feed monitoring module in a distributed system, crushed raw material status data collected by a raw material information extraction module in a distributed system, and speed values ​​of a crusher at multiple predetermined time points collected by a crusher speed extraction module in a distributed system;

[0012] Extracting a feed crushing state associated feature vector and a grinder speed feature vector from the feed crushing monitoring video collected by the feed monitoring module in the distributed system, the crushing raw material state data collected by the raw material information extraction module in the distributed system, and the grinder speed values ​​at multiple predetermined time points collected by the grinder speed extraction module in the distributed system;

[0013] Based on the feed crushing state associated feature vector and the grinder speed feature vector, it is determined whether the grinder speed needs to be adjusted.

[0014] Compared with the prior art, the present application provides a distributed automatic batching control system and method, which first obtains the feed crushing monitoring video collected by the feed monitoring module in the distributed system, the crushed raw material status data collected by the raw material information extraction module in the distributed system, and the grinder speed values ​​at multiple predetermined time points collected by the grinder speed extraction module in the distributed system, and then uses deep learning technology to perform feature extraction and association analysis on the three, and finally obtains the classification result through a classifier to determine whether the grinder speed needs to be adjusted, thereby ensuring that the raw material is crushed to the expected particle size and improving the quality and consistency of the feed. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0016] Figure 1 It is a block diagram of a distributed automatic batching control system according to an embodiment of the present application.

[0017] Figure 2 It is a block diagram of a distributed batching data extraction module in a distributed automatic batching control system according to an embodiment of the present application.

[0018] Figure 3 It is a block diagram of a feed monitoring data feature extraction unit in a distributed automatic batching control system according to an embodiment of the present application.

[0019] Figure 4 It is a block diagram of a grinder speed adjustment analysis module in a distributed automatic batching control system according to an embodiment of the present application.

[0020] Figure 5 Flow chart of a distributed automatic batching control method according to an embodiment of the present application.

[0021] Figure 6 is a block diagram of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0022] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0023] Figure 1 FIG. 1 is a block diagram of a distributed automatic batching control system according to an embodiment of the present application. Figure 1As shown, according to the distributed automatic batching control system 100 of the embodiment of the present application, it includes: a distributed batching data acquisition module 110, which is used to acquire the feed crushing monitoring video collected by the feed monitoring module in the distributed system, the crushing raw material state data collected by the raw material information extraction module in the distributed system, and the speed value of the crusher at multiple predetermined time points collected by the crusher speed extraction module in the distributed system; a distributed batching data extraction module 120, which is used to extract the feed crushing state associated feature vector and the crusher speed feature vector from the feed crushing monitoring video collected by the feed monitoring module in the distributed system, the crushing raw material state data collected by the raw material information extraction module in the distributed system, and the speed value of the crusher at multiple predetermined time points collected by the crusher speed extraction module in the distributed system; a crusher speed adjustment analysis module 130, which is used to determine whether the crusher speed needs to be adjusted based on the feed crushing state associated feature vector and the crusher speed feature vector.

[0024] In the above-mentioned distributed automatic batching control system 100, the distributed batching data acquisition module 110 is used to obtain the feed crushing monitoring video collected by the feed monitoring module in the distributed system, the crushing raw material state data collected by the raw material information extraction module in the distributed system, and the speed value of the crusher at multiple predetermined time points collected by the crusher speed extraction module in the distributed system. It should be understood that feed is a mixture specially provided for animals to provide the required nutrients, aiming to promote their health, growth and production performance. The formula of feed will be adjusted according to the type, age and production goals of different animals, and usually contains grains, protein sources, minerals, vitamins and other nutritional additives. The process of making feed generally includes several steps: crushing, screening, mixing and metering of raw materials. First, the raw materials are placed in a crusher for crushing to form uniform particles for subsequent mixing. The crushed materials will pass through a screening device to remove particles that do not meet the specifications. Next, the system will automatically weigh various raw materials according to the set formula, and accurately batch the ingredients, and finally fully mix all the raw materials through a mixer to ensure the consistency and quality of the feed. Among these steps, crushing is a key link. The traditional crushing process is to put the raw materials into the crusher and operate it according to a predetermined procedure. However, due to insufficient control of the crushing effect and imprecise control of the crushing particle size, this method may lead to unstable or low feed quality. Therefore, in the technical solution of the present application, by obtaining the feed crushing monitoring video collected by the feed monitoring module in the distributed system, the crushed raw material status data collected by the raw material information extraction module in the distributed system, and the speed values ​​of the crusher at multiple predetermined time points collected by the crusher speed extraction module in the distributed system, and combining with deep learning technology, it is determined whether the speed of the crusher needs to be adjusted, thereby ensuring that the raw material reaches the expected crushing particle size and improving the quality and uniformity of the feed.

[0025] Specifically, a distributed system is a network composed of multiple independent computing nodes that work together to complete tasks. In a specific application scenario, such as a feed production system, it can include a feed monitoring module for real-time tracking and management of feed status, a raw material information extraction module responsible for collecting and analyzing raw material data, and a grinder speed extraction module to monitor the grinder speed to ensure that it is within the optimal performance range. These modules work together through a distributed architecture to improve the reliability, flexibility and efficiency of the system. Among them, the feed crushing monitoring video can capture visual information during the crushing process in real time, which is crucial for identifying and monitoring the uniformity of crushing, raw material status, and possible abnormal conditions. Secondly, the crushed raw material status data provided by the raw material information extraction module helps to understand the specific status and properties of the raw materials. Finally, the speed data at multiple predetermined time points provided by the grinder speed extraction module is the basis for evaluating the performance of the grinder. Through the comprehensive analysis of monitoring video, raw material status data and speed information, the system can adjust the operation of the grinder in real time and optimize the production process, thereby improving the uniformity and quality of feed to meet the nutritional needs of different animals.

[0026] In the above-mentioned distributed automatic batching control system 100, the distributed batching data extraction module 120 is used to extract the feed crushing state associated feature vector and the grinder speed feature vector from the feed crushing monitoring video collected by the feed monitoring module in the distributed system, the crushing raw material state data collected by the raw material information extraction module in the distributed system, and the grinder speed values ​​at multiple predetermined time points collected by the grinder speed extraction module in the distributed system. It should be understood that by analyzing the associated characteristics of the crushing state and the speed, the system can adjust the operating parameters in real time to optimize the crushing effect and feed quality. This not only improves the accuracy and efficiency of production, but also ensures the stability and consistency of the feed, thereby meeting the nutritional needs of different animals and improving the overall production efficiency.

[0027] Figure 2 FIG. 1 is a block diagram of a distributed batching data extraction module in a distributed automatic batching control system according to an embodiment of the present application. Figure 2As shown, in a specific embodiment of the present application, the distributed ingredient data extraction module 120 includes: a feed monitoring data feature extraction unit 121, which is used to perform feature extraction on the feed crushing monitoring video collected by the feed monitoring module in the distributed system to obtain a feed crushing monitoring target efficient attention feature vector; a raw material information feature extraction unit 122, which is used to perform feature extraction on the crushed raw material state data collected by the raw material information extraction module in the distributed system to obtain a crushed raw material state semantic association feature vector; a feed crushing state feature association unit 123, which is used to associate the feed crushing monitoring target efficient attention feature vector with the crushing raw material state semantic association feature vector to obtain the feed crushing state association feature vector; a grinder speed feature extraction unit 124, which is used to perform feature extraction on the grinder speed values ​​at multiple predetermined time points collected by the grinder speed extraction module in the distributed system to obtain the grinder speed feature vector.

[0028] It should be understood that the visual data of the feed grinding process contains a lot of information about the state of the raw materials, the uniformity of the grinding, and the operation of the equipment. By performing feature extraction on these videos, key visual features can be extracted from a large amount of image data to help identify important details in the grinding process. For example, the video can show the particle size during the grinding process, the operating status of the grinder, and possible failures or abnormalities. These features represent key information in the grinding process and can be combined with other data sources (such as raw material status data and speed data) to build a comprehensive process model. Through this model, the system can evaluate the effect of the grinding process in real time and make corresponding adjustments to optimize production parameters and ensure the consistency and stability of feed quality.

[0029] Furthermore, the crushed raw material status data contains a wealth of raw material information, such as raw material particle size, humidity, density and other physical and chemical properties. These data directly affect the crushing effect and the final quality of the feed. For example, the humidity and particle size of the raw material will affect the uniformity and effect of the crushing, and then affect the nutritional value and production efficiency of the feed. Through feature extraction, these complex data can be converted into feature vectors with semantic meaning, thereby refining the key information of the raw material status.

[0030] Furthermore, the feed crushing monitoring target efficient attention feature vector is derived from the visual data analysis of the crushing process. It focuses on key visual features in the crushing process, such as the morphology of raw material particles, the operating status of the crusher, and possible abnormalities. The crushing raw material state semantic association feature vector is derived from the feature extraction of raw material information. It contains in-depth information about the physical and chemical properties of the raw material, such as particle size, moisture, density, etc. By combining visual monitoring data and raw material state data, a panoramic view of the crushing process is provided, allowing the system to identify and understand the relationship between raw material characteristics and the crushing process.

[0031] In particular, the speed of the pulverizer determines the shear force and impact force to which the material is subjected during the pulverization process, thereby affecting the degree and uniformity of the raw material pulverization. Too high or too low a speed may lead to poor pulverization. For example, too high a speed may cause the raw material to overheat, affecting the pulverization quality; too low a speed may lead to incomplete pulverization, affecting the uniformity and nutritional value of the feed. The feature extraction process converts the speed values ​​at multiple predetermined time points into speed feature vectors. This process can fully describe the operating status of the pulverizer by extracting important information from the speed data, such as the average speed, fluctuation amplitude, and change trend. By performing feature extraction on the speed data, complex raw data can be converted into structured information, making it easier to analyze and process. These feature vectors can be used to monitor the operating status of the pulverizer in real time, helping operators to discover and solve problems in a timely manner.

[0032] Figure 3 FIG. 1 is a block diagram of a feed monitoring data feature extraction unit in a distributed automatic batching control system according to an embodiment of the present application. Figure 3 As shown, in a specific embodiment of the present application, the feed monitoring data feature extraction unit 121 includes: a feed monitoring key frame extraction subunit 1211, which is used to extract key frames from the feed crushing monitoring video collected by the feed monitoring module in the distributed system to obtain multiple feed crushing monitoring key frames; a feed monitoring target detection subunit 1212, which is used to pass the multiple feed crushing monitoring key frames through a feed crushing monitoring target detection network model to obtain multiple feed crushing monitoring regions of interest; a feed monitoring convolutional coding subunit 1213, which is used to pass the multiple feed crushing monitoring regions of interest through a feed crushing monitoring target efficient attention mechanism as a feature extractor to obtain the feed crushing monitoring target efficient attention feature vector.

[0033] It should be understood that video data usually contains a large number of consecutive frames, most of which are redundant in information content and highly repetitive. Extracting key frames can significantly reduce the amount of data that needs to be processed while retaining important information in the video. Key frames are frames that represent major events or scenes in the video and can capture key moments in the pulverization process, such as the start and stop of equipment, the input of raw materials, and the effect of pulverization. By extracting these frames, data redundancy can be effectively reduced and the efficiency of subsequent analysis can be improved.

[0034] Furthermore, the feed crushing monitoring keyframe may contain a large amount of information, of which only a part of the area is important for analysis. These regions of interest (ROI) may include key elements such as the working status of the grinder, the distribution of raw materials, and the crushing effect. The target detection network model can automatically identify and locate these important areas, thereby focusing resources on analyzing the most relevant parts. This method reduces the processing requirements for the entire image and improves computational efficiency. The target detection network model is usually trained with a large amount of labeled data and can learn to identify the characteristics of target objects in the crushing process, such as equipment components, particles in the crushing process, etc. Through this training, the model can accurately detect and locate these target areas in the keyframe, thereby extracting valuable information. After extracting the regions of interest, the details in these areas can be further analyzed. Specifically, the feed crushing monitoring target detection network model is used to process the multiple feed crushing monitoring keyframes with the following target detection formula to obtain the multiple feed crushing monitoring regions of interest;

[0035] Among them, the target detection formula is:

[0036] ROI=(cls(ψ det ,B),Regr(ψ det ,B))

[0037] Among them, ψ det are multiple feed crushing monitoring key frames, B is the anchor frame, ROI is the region of interest for multiple feed crushing monitoring, cls(ψ det ,B) indicates classification, Regr(ψ det ,B) indicates regression.

[0038] Furthermore, the region of interest (ROI) is the important area located by the object detection model in the key frames extracted from the feed crushing monitoring video. These regions contain key information directly related to the crushing process, such as equipment status, raw material distribution and crushing effect. In order to extract useful information from these regions, deep learning network models, especially convolutional neural networks (CNNs), are widely used to process image data. Convolutional coding is the process of processing image data through convolution operations. Convolutional neural networks extract different levels of features in the image through multiple convolutional layers, from low-level edges and textures to high-level shapes and objects. These convolutional layers can effectively capture important local features in the image by applying different convolution kernels (filters) and integrate them into higher-level feature representations. This feature extraction method can help the system accurately understand the details of the crushing process and make corresponding adjustments.

[0039] It should be understood that the region of interest (ROI) represents the important parts of the key frames in the video that are related to the crushing process. However, directly analyzing these areas may be affected by data redundancy and noise. Therefore, the use of an efficient attention mechanism can better focus on the most important features, thereby improving the accuracy of the analysis. The efficient attention mechanism is a deep learning technology that improves the effect of feature extraction by automatically learning and focusing on the most relevant information areas. Specifically, the attention mechanism is able to assign different weights to each area by calculating the relative importance between regions. In this way, the model can pay more attention to the parts that have a direct impact on the task and ignore irrelevant or redundant information. In feed crushing monitoring, this mechanism can help the system identify and focus on key states and abnormal conditions in the crushing process. Specifically, the convolutional coding part of the feed crushing monitoring target efficient attention mechanism as a feature extractor is used to perform deep convolutional coding on the multiple feed crushing monitoring interest regions to obtain an initial convolutional feature map; the initial convolutional feature map is input into the spatial attention part of the feed crushing monitoring target efficient attention mechanism as a feature extractor to obtain a spatial attention map; the spatial attention map is activated by a Softmax function to obtain a spatial attention feature map; and the spatial attention feature map and the initial convolutional feature map are multiplied by the position point to obtain the feed crushing monitoring target efficient attention feature vector. More specifically, the initial convolutional feature map is input into the spatial attention part of the feed crushing monitoring target efficient attention mechanism as a feature extractor to obtain the spatial attention map, including: performing average pooling and maximum pooling along the channel dimension on the initial convolutional feature map to obtain an average feature matrix and a maximum feature matrix; cascading and channel-adjusting the average feature matrix and the maximum feature matrix to obtain a channel feature matrix; and using the convolutional layer of the spatial attention feature map to perform convolutional coding on the channel feature matrix to obtain the spatial attention map.

[0040] In a specific embodiment of the present application, the raw material information feature extraction unit 122 is used to: segment the crushed raw material status data collected by the raw material information extraction module in the distributed system to obtain a crushed raw material status word sequence; pass the crushed raw material status word sequence through a crushed raw material status context semantic encoder to obtain multiple crushed raw material status semantic feature vectors; and cascade the multiple crushed raw material status semantic feature vectors into the crushed raw material status semantic association feature vector.

[0041] It should be understood that the crushed raw material status data is usually recorded in text or numerical form. However, these data may be messy in the original form, and direct analysis may encounter problems such as data redundancy and information confusion. Word segmentation technology can break down these data into smaller, meaningful units (words), thereby improving the processability and analyzability of the data. The word segmentation process involves splitting the raw material status data into meaningful words or phrases. This segmentation method can convert data from larger text blocks into structured word sequences, allowing the subsequent analysis process to focus more on each specific status feature.

[0042] Furthermore, after the crushed raw material state word sequence is obtained through word segmentation, although the data is converted into a structured vocabulary form, the existence of these words alone may not be able to fully express the complex semantic relationship of the raw material state. The crushed raw material state context semantic encoder can use context information to perform in-depth semantic understanding of these words. Through the context semantic encoder, these words will be converted into semantic feature vectors with context information. These vectors not only reflect the meaning of the words themselves, but also include their specific functions and relationships in a specific context. The generated crushed raw material state semantic feature vector is a high-dimensional, comprehensive representation that can effectively capture and express the complex characteristics of the raw material state. Specifically, the embedding layer of the crushed raw material state context semantic encoder is used to map each crushed raw material state word in the crushed raw material state word sequence to a word embedding vector to obtain a sequence of crushed raw material state word embedding vectors; the converter-based Bert model of the crushed raw material state context semantic encoder is used to perform global context semantic encoding on the sequence of crushed raw material state word embedding vectors to obtain multiple crushed raw material state semantic feature vectors.

[0043] Furthermore, in practical applications, each semantic feature vector of crushed raw material status provides a deep semantic representation of a single word or phrase, but the information represented by each vector may not be comprehensive enough. When these feature vectors are cascaded together, they can present the overall raw material status information in a unified and comprehensive way. The cascade operation connects each individual feature vector into a longer vector, thereby integrating the full information of multiple semantic features. This cascade method can significantly improve the expressiveness of information.

[0044] It should be understood that by associating the feed crushing monitoring target efficient attention feature vector and the crushing raw material state semantic association feature vector, the information of the two can be combined to more comprehensively reflect the state of the feed crushing process. This can more accurately determine whether the current crusher speed is suitable for the current crushing task, and whether the crusher speed needs to be adjusted to ensure crushing efficiency and product quality.

[0045] In a specific embodiment of the present application, the grinder speed feature extraction unit 124 includes: arranging the grinder speed values ​​at multiple predetermined time points collected by the grinder speed extraction module in the distributed system into a grinder speed input vector; passing the grinder speed input vector through a grinder speed time series feature extractor to obtain the grinder speed feature vector.

[0046] It should be understood that the speed data of the pulverizer is usually in the form of a time series, which contains speed values ​​at multiple time points. Directly using these time point data for analysis may be challenging because the time series characteristics of the speed data need to be effectively captured and utilized. Arranging this data as a pulverizer speed input vector helps to systematically process and analyze this time series information. Specifically, arranging the speed values ​​as an input vector can convert the dynamic time series data into a fixed-length feature vector. This vector will include the speed values ​​of multiple time points, arranged in chronological order, thereby retaining the time series characteristics of the speed change.

[0047] Furthermore, the speed data of the pulverizer is recorded in the form of a time series, which usually includes the speed values ​​at multiple time points. However, the original time series data is often very large and complex, and it may be challenging to use it directly for analysis and modeling. In order to effectively extract and utilize this information, the data needs to be processed by a feature extractor. The task of the pulverizer speed time series feature extractor is to extract representative features from the time series data in order to better capture the dynamic characteristics and potential patterns of the speed data. These features can include statistics of the speed (such as mean, standard deviation), frequency domain features (such as power spectral density), time domain features (such as autocorrelation coefficient), etc. Through these features, important patterns in the speed data can be identified, such as periodic changes, trend changes, or sudden anomalies. Specifically, the feature extractor can apply various techniques, such as sliding windows, Fourier transforms, wavelet transforms, etc., to analyze different aspects of time series data. Through the sliding window method, the time series data can be decomposed into local features within multiple windows to capture short-term changes in the speed. The Fourier transform can convert the time series data into the frequency domain to identify the periodic components and frequency characteristics of the speed. Wavelet transform can provide multi-scale time-frequency analysis to capture more complex time-varying features. In a specific embodiment of the present application, the fully connected layer of the grinder speed time series feature extractor is used to perform full-connection encoding on the grinder speed input vector to extract high-dimensional implicit features of the eigenvalues ​​at each position in the grinder speed input vector; and the one-dimensional convolution layer of the grinder speed time series feature extractor is used to perform one-dimensional encoding on the grinder speed input vector to extract high-dimensional implicit correlation features of the correlation between the eigenvalues ​​at each position in the grinder speed input vector.

[0048] In the above-mentioned distributed automatic batching control system 100, the grinder speed adjustment analysis module 130 is used to determine whether the grinder speed needs to be adjusted based on the feed crushing state associated feature vector and the grinder speed feature vector. It should be understood that the feed crushing state associated feature vector contains comprehensive characteristics related to the feed crushing process, such as particle size distribution, humidity change, crushing efficiency, etc. On the other hand, the grinder speed feature vector provides time series characteristics about the grinder speed, such as speed change trend, fluctuation frequency, etc. By comprehensively analyzing these two feature vectors, the relationship between the working state of the grinder and the crushing effect can be more comprehensively understood, thereby reducing manual intervention and improving production efficiency and safety.

[0049] Figure 4 FIG. 1 is a block diagram of a pulverizer speed adjustment analysis module in a distributed automatic batching control system according to an embodiment of the present application. Figure 4 As shown, in a specific embodiment of the present application, the grinder speed adjustment analysis module 130 includes: a distributed ingredient feature fusion unit 131, which is used to perform feature alignment driven by essential characteristics on the feed crushing state associated feature vector and the grinder speed feature vector to obtain a grinder speed adjustment judgment feature vector; a grinder speed adjustment judgment classification unit 132, which is used to pass the grinder speed adjustment judgment feature vector through a classifier to obtain a classification result, and the classification result is used to judge whether the grinder speed needs to be adjusted.

[0050] It should be understood that fusing the feed crushing state associated feature vector and the grinder speed feature vector into a grinder speed adjustment judgment feature vector can integrate these two types of information to form a more comprehensive feature representation. This fusion can capture the complex relationship in the crushing process, so as to more accurately determine whether the speed needs to be adjusted and determine the optimal adjustment range.

[0051] In particular, in the technical solution of the present application, it is considered that the feed crushing state-related feature vector (such as raw material moisture, particle size, etc. obtained by video analysis) and the grinder speed feature vector (such as speed value) may have different numerical ranges. If directly merged, one feature vector may dominate the learning process of the model due to its larger numerical range, masking the importance of another feature, thereby affecting the accuracy of the final analysis. The feed crushing state-related feature vector usually captures information about the visual or physical state, such as the appearance and texture changes of the raw materials; while the grinder speed feature vector focuses more on mechanical performance data, such as speed, power consumption, etc. These two features represent information from different perspectives and levels. Direct merging without considering their semantic differences may confuse the model's understanding of this information and fail to accurately distinguish the relationship between state information and mechanical properties. The feed crushing monitoring video provides continuous time series data that can reflect the changes in the raw material state over time; while the grinder speed may be data collected at discrete time points. Directly merging these two types of data with different time characteristics and frequencies may destroy the original time series relationship, cause information distortion, and affect the correct understanding of the dynamic characteristics of the crushing process. Due to different sources, there may be information overlap or contradictions between the two feature vectors. For example, the state of the raw materials in a certain video frame may have implicitly reflected the machine operation status at that time, and direct merging may lead to repeated calculations or cancel each other out effective information, making it difficult for the model to extract meaningful decision boundaries from it. In order to solve this problem, in the technical solution of this application, the feed crushing state associated feature vector and the crusher speed feature vector are aligned with the essential characteristics driven feature.

[0052] Specifically, in an embodiment of the present application, the distributed ingredient feature fusion unit includes: a feature alignment subunit, which is used to perform feature alignment based on norm calibration on the feed crushing state associated feature vector and the grinder speed feature vector to obtain a pre-aligned feed crushing state associated feature vector and a pre-aligned grinder speed feature vector; a principal component feature generation subunit, which is used to perform intrinsic characteristic analysis on the pre-aligned feed crushing state associated feature vector and the pre-aligned grinder speed feature vector to obtain a first principal component pre-aligned feed crushing state associated feature vector and a second principal component pre-aligned grinder speed feature vector; a grinder speed adjustment judgment feature generation subunit, which is used to perform fine-grained feature integration on the first principal component pre-aligned feed crushing state associated feature vector and the second principal component pre-aligned grinder speed feature vector to obtain the grinder speed adjustment judgment feature vector.

[0053] More specifically, in an embodiment of the present application, the feature alignment subunit is used to: calculate the square of the binary norm of the feed crushing state associated feature vector to obtain a feed crushing state feature intensity value; calculate the natural exponential function value for each eigenvalue of the feed crushing state associated feature vector to obtain an indexed feed crushing state associated feature vector; divide each eigenvalue of the indexed feed crushing state associated feature vector by the sum of the indexed feed crushing state associated feature vectors to obtain a normalized feed crushing state associated feature vector; multiply each eigenvalue of the normalized feed crushing state associated feature vector by the feed crushing state feature intensity value to obtain the pre-aligned feed crushing state associated feature vector.

[0054] More specifically, in an embodiment of the present application, the principal component feature generation subunit is used to: perform eigenvalue-based decomposition on the pre-aligned feed crushing state associated feature vector to obtain a first eigenvalue-based decomposition sequence of pre-aligned feed crushing state associated feature vectors and a first diagonal matrix; and cascade the first eigenvector and the second eigenvector in the sequence of the first eigenvalue-based decomposition pre-aligned feed crushing state associated feature vector to obtain the first principal component pre-aligned feed crushing state associated feature vector.

[0055] In the embodiment of the present application, the formula implementation expression of the distributed batching feature fusion unit can be: the feed crushing state associated feature vector and the grinder speed feature vector are processed by the following formula to obtain the grinder speed adjustment judgment feature vector; wherein the formula is:

[0056]

[0057]

[0058] Among them, V 1 Represents the associated feature vector of feed crushing state, V 2 represents the pulverizer speed feature vector, represents the square of the vector's second norm, v 1i represents the eigenvalue of the ith position of the eigenvector associated with the feed crushing state, v 2i represents the eigenvalue of the ith position of the eigenvector of the mill speed, exp represents the natural exponential function, V 1 ′ represents the associated feature vector of the pre-aligned feed crushing state, V 2 ′ represents the pre-aligned mill speed eigenvector, PCA(·) represents the intrinsic characteristic analysis, U 1 represents the sequence of associated eigenvectors of the first eigendecomposition pre-aligned feed crushing state, Λ 1 represents the first diagonal matrix, λ 11 and λ1m denote the eigenvalues ​​of the first and mth positions of the first diagonal matrix, respectively. T denotes the transpose of the vector. concat{·;·} denotes vector concatenation. v 11 、v 12 、v 1m denote the first, second and mth eigenvectors of the sequence of the associated eigenvectors of the first eigendecomposition pre-aligned feed crushing state, V k represents the first principal component pre-aligned feed crushing state associated eigenvector, U 2 represents the sequence of the second eigendecomposition pre-aligned mill speed eigenvectors, Λ 2 represents the second diagonal matrix, λ 21 and λ 2m denote the eigenvalues ​​of the first and mth positions of the second diagonal matrix, respectively. concat{·;·} denotes vector concatenation. v 21 、v 22 、v 2m denote the first, second and mth eigenvectors of the sequence of the second eigendecomposition pre-aligned mill speed eigenvectors, V s represents the second principal component pre-aligned mill speed eigenvector, represents positional subtraction, ⊙ represents positional multiplication, Indicates positional addition, Conv 1×1 represents point convolution, f vp (V k ,V s ) represents the crusher speed adjustment judgment feature vector.

[0059] That is, in the technical solution of the present application, the feed crushing state associated feature vector and the grinder speed feature vector are subjected to feature alignment driven by essential characteristics to obtain the grinder speed adjustment judgment feature vector, which first performs feature alignment based on norm calibration on the feed crushing state associated feature vector and the grinder speed feature vector to obtain a pre-aligned feed crushing state associated feature vector and a pre-aligned grinder speed feature vector. This process is not just for simple scale adjustment, but an important technology in feature engineering, which aims to solve the problem of inconsistent scales between features in the data set. By calculating the modulus of the feature vector and dividing each vector by its own modulus, it can be ensured that all feature vectors have the same length, thereby eliminating the impact of scale differences between different feature vectors. Normalization processing not only helps numerical stability in subsequent processing, but also makes the comparison between feature vectors more fair and reasonable, laying the foundation for subsequent feature alignment.

[0060] Next, the pre-aligned feed crushing state associated feature vector and the pre-aligned grinder speed feature vector are subjected to intrinsic characteristic analysis to obtain the first principal component pre-aligned feed crushing state associated feature vector and the second principal component pre-aligned grinder speed feature vector. That is, after the normalization modulation is completed, the pre-aligned feed crushing state associated feature vector and the grinder speed feature vector are subjected to intrinsic characteristic analysis. As an unsupervised linear dimensionality reduction method, intrinsic characteristic analysis is widely used in data compression and feature extraction. This process determines the principal components by finding the direction of the maximum variance of the data. These principal components are new coordinate axes in the original feature space and can capture the structure of the data with minimal information loss. It is worth mentioning that by selecting the eigenvectors corresponding to the first few largest eigenvalues, the dimension of the feature space can be effectively reduced while maintaining the key information of the data. This dimensionality reduction can not only accelerate the model training process, but also help the model generalize better and reduce the impact of noise, thereby improving the performance of the model.

[0061] Finally, fine-grained feature integration is performed on the first principal component pre-aligned feed crushing state associated feature vector and the second principal component pre-aligned grinder speed feature vector to obtain the grinder speed adjustment judgment feature vector. The goal of this step is to find the best matching point or area between the two principal component feature vectors to achieve alignment between the two. Fine-grained alignment goes beyond the traditional rough fusion methods such as simple weighted average or splicing, and needs to consider the time series characteristics or spatial distribution characteristics of the feature vector. The fusion after alignment is not a simple numerical operation, but requires the design of a specific mechanism to determine how to combine the information of the two feature vectors. This content-based intelligent fusion method can more effectively utilize multi-source information and improve the representation ability and decision-making quality of the model.

[0062] Furthermore, through the classifier, these complex feature data can be automatically processed and analyzed to obtain classification results, such as "needs adjustment" or "no adjustment". The advantage of this method lies in its efficiency and accuracy. The classifier can maintain high performance when processing complex data, evaluate the operating status of the pulverizer in real time, and make adjustment suggestions. Through automated classification decisions, manual operation errors can be reduced and the system can be ensured to respond quickly to changes in production, thereby optimizing the working efficiency and pulverization effect of the pulverizer.

[0063] In summary, the embodiment of the present application first obtains the feed crushing monitoring video collected by the feed monitoring module in the distributed system, the crushing raw material status data collected by the raw material information extraction module in the distributed system, and the grinder speed values ​​at multiple predetermined time points collected by the grinder speed extraction module in the distributed system, and then uses deep learning technology to perform feature extraction and association analysis on the three, and finally obtains the classification result through the classifier to determine whether the grinder speed needs to be adjusted, so as to ensure that the raw material is crushed to the expected particle size and improve the quality and consistency of the feed.

[0064] As described above, the distributed automatic batching control system 100 according to the embodiment of the present application can be implemented in various terminal devices. In one example, the distributed automatic batching control system 100 can be integrated into the terminal device as a software module and / or a hardware module. For example, the distributed automatic batching control system 100 can be a software module in the operating system of the terminal device, or can be an application developed for the terminal device; of course, the distributed automatic batching control system 100 can also be one of the many hardware modules of the terminal device.

[0065] Alternatively, in another example, the distributed automatic batching control system 100 and the terminal device may also be separate devices, and the distributed automatic batching control system 100 may be connected to the terminal device via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

[0066] Figure 5 FIG. 1 is a flow chart of a distributed automatic batching control method according to an embodiment of the present application. Figure 5 As shown, according to the distributed automatic batching control method of the embodiment of the present application, it includes: S110, acquiring the feed crushing monitoring video collected by the feed monitoring module in the distributed system, the crushing raw material status data collected by the raw material information extraction module in the distributed system, and the speed values ​​of the crusher at multiple predetermined time points collected by the crusher speed extraction module in the distributed system; S120, extracting the feed crushing state associated feature vector and the crusher speed feature vector from the feed crushing monitoring video collected by the feed monitoring module in the distributed system, the crushing raw material status data collected by the raw material information extraction module in the distributed system, and the speed values ​​of the crusher at multiple predetermined time points collected by the crusher speed extraction module in the distributed system; S130, judging whether the crusher speed needs to be adjusted based on the feed crushing state associated feature vector and the crusher speed feature vector.

[0067] Here, those skilled in the art will appreciate that the specific operations of each step in the above-mentioned distributed automatic batching control method have been described in detail above. Figures 1 to 4The description of the distributed automatic batching control system has been introduced in detail, and therefore, its repeated description will be omitted.

[0068] Below, reference Figure 6 To describe an electronic device according to an embodiment of the present application.

[0069] Figure 6 is a block diagram of an electronic device according to an embodiment of the present application. Figure 6 , which shows a schematic diagram of the structure of an electronic device suitable for implementing a terminal device or a server in an embodiment of the present application.

[0070] like Figure 6 As shown, the electronic device includes one or more processors, a communication unit, etc. The one or more processors are, for example: one or more central processing units (CPUs), and / or one or more graphics processing units (GPUs), etc. The processor can perform various appropriate actions and processes according to executable instructions stored in a read-only memory (ROM) or executable instructions loaded from a storage part to a random access memory (RAM). The communication unit may include but is not limited to a network card, and the network card may include but is not limited to an IB (Infiniband) network card. The processor may communicate with a read-only memory and / or a random access memory to execute executable instructions, and is connected to the communication unit through a bus and communicates with other target devices through the communication unit, thereby completing the operation corresponding to any method provided in the embodiments of the present application, for example, the primary shard manager shards the features in the feature queue; for any shard, the primary shard manager stores the any shard in the corresponding shard feature library, and records the mapping relationship between the any shard and the corresponding shard feature library in a mapping table; in response to a failure of the primary shard manager, a shard manager is reselected from at least one slave shard manager as the new primary shard manager; the new primary shard manager obtains shard information of the shard currently being written from all shard feature libraries, and recreates the mapping relationship of the shard currently being written based on the shard information of the shard currently being written in all shard feature libraries.

[0071] In addition, various programs and data required for the operation of the device can also be stored in the RAM. The CPU, ROM and RAM are connected to each other via a bus. In the case of RAM, the ROM is an optional module. The RAM stores executable instructions, or writes executable instructions into the ROM at runtime, and the executable instructions enable the processor to perform operations corresponding to any of the above methods of the present invention. The input / output (I / O) interface is also connected to the bus. The communication unit can be integrated or can be configured to have multiple submodules (e.g., multiple IB network cards) and be linked on the bus.

[0072] The following components are connected to the I / O interface: an input section including a keyboard, a mouse, etc.; an output section including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage section including a hard disk, etc.; and a communication section including a network interface card such as a LAN card, a modem, etc. The communication section performs communication processing via a network such as the Internet. A drive is also connected to the I / O interface as needed. Removable media, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., are installed on the drive as needed so that a computer program read therefrom is installed into the storage section as needed.

[0073] If necessary, Figure 6 The architecture shown is only an optional implementation method. In the specific practice, the above can be modified according to actual needs. Figure 6 The number and type of components can be selected, deleted, added or replaced; different functional components can be set up by separate settings or integrated settings, for example, the GPU and CPU can be set up separately or the GPU can be integrated on the CPU, the communication unit can be set up separately or integrated on the CPU or GPU, etc. These replaceable implementations all fall within the protection scope disclosed in the present invention.

[0074] In addition, an embodiment of the present invention further provides a computer program, including computer-readable code. When the computer-readable code runs on a device, a processor in the device executes instructions for implementing each step in the data processing method of any of the above embodiments of the present invention.

Claims

1. A distributed automatic batching control system, characterized in that: include: A distributed batching data acquisition module is used to acquire feed crushing monitoring video collected by a feed monitoring module in a distributed system, crushed raw material status data collected by a raw material information extraction module in a distributed system, and the speed values ​​of a crusher at multiple predetermined time points collected by a crusher speed extraction module in a distributed system; A distributed batching data extraction module is used to extract a feed crushing state associated feature vector and a grinder speed feature vector from the feed crushing monitoring video collected by the feed monitoring module in the distributed system, the crushing raw material state data collected by the raw material information extraction module in the distributed system, and the grinder speed values ​​at multiple predetermined time points collected by the grinder speed extraction module in the distributed system; The grinder speed adjustment analysis module is used to determine whether the grinder speed needs to be adjusted based on the feed grinding state associated feature vector and the grinder speed feature vector.

2. The distributed automatic batching control system according to claim 1, characterized in that: The distributed batching data extraction module comprises: A feed monitoring data feature extraction unit, used for extracting features from the feed crushing monitoring video collected by the feed monitoring module in the distributed system to obtain a feed crushing monitoring target efficient attention feature vector; A raw material information feature extraction unit, used for extracting features from the crushed raw material state data collected by the raw material information extraction module in the distributed system to obtain a crushed raw material state semantic association feature vector; A feed crushing state feature association unit, used for associating the feed crushing monitoring target efficient attention feature vector with the crushing raw material state semantic association feature vector to obtain the feed crushing state association feature vector; The grinder speed feature extraction unit is used to extract features from the grinder speed values ​​at multiple predetermined time points collected by the grinder speed extraction module in the distributed system to obtain the grinder speed feature vector.

3. The distributed automatic batching control system according to claim 2, characterized in that: The feed monitoring data feature extraction unit comprises: A feed monitoring key frame extraction subunit, used for extracting key frames from the feed crushing monitoring video collected by the feed monitoring module in the distributed system to obtain a plurality of feed crushing monitoring key frames; A feed monitoring target detection subunit, configured to obtain a plurality of feed crushing monitoring regions of interest by passing the plurality of feed crushing monitoring key frames through a feed crushing monitoring target detection network model; The feed monitoring convolutional encoding subunit is used to obtain the feed crushing monitoring target efficient attention feature vector by using the feed crushing monitoring target efficient attention mechanism as a feature extractor for the multiple feed crushing monitoring interest regions.

4. The distributed automatic batching control system according to claim 3, characterized in that: The raw material information feature extraction unit is used to: Segmenting the crushed raw material state data collected by the raw material information extraction module in the distributed system to obtain a crushed raw material state word sequence; Passing the crushed raw material state word sequence through a crushed raw material state context semantic encoder to obtain a plurality of crushed raw material state semantic feature vectors; The multiple crushed raw material state semantic feature vectors are cascaded into the crushed raw material state semantic association feature vector.

5. The distributed automatic batching control system according to claim 4, characterized in that: The pulverizer speed feature extraction unit is used to: Arranging the speed values ​​of the pulverizer at a plurality of predetermined time points collected by the pulverizer speed extraction module in the distributed system as a pulverizer speed input vector; The pulverizer speed input vector is passed through a pulverizer speed time series feature extractor to obtain the pulverizer speed feature vector.

6. The distributed automatic batching control system according to claim 5, characterized in that: The pulverizer speed adjustment and analysis module comprises: A distributed batching feature fusion unit, used for performing feature alignment driven by essential characteristics on the feed crushing state associated feature vector and the grinder speed feature vector to obtain a grinder speed adjustment judgment feature vector; The grinder speed adjustment judgment classification unit is used to pass the grinder speed adjustment judgment feature vector through a classifier to obtain a classification result, and the classification result is used to judge whether the grinder speed needs to be adjusted.

7. The distributed automatic batching control system according to claim 6, characterized in that: The distributed batching feature fusion unit comprises: A feature alignment subunit, used for performing feature alignment based on norm calibration on the feed crushing state associated feature vector and the grinder speed feature vector to obtain a pre-aligned feed crushing state associated feature vector and a pre-aligned grinder speed feature vector; A principal component feature generation subunit is used to perform an intrinsic characteristic analysis on the pre-aligned feed pulverizing state associated feature vector and the pre-aligned pulverizer speed feature vector to obtain a first principal component pre-aligned feed pulverizing state associated feature vector and a second principal component pre-aligned pulverizer speed feature vector; The grinder speed adjustment judgment feature generation subunit is used to perform fine-grained feature integration on the first principal component pre-aligned feed crushing state associated feature vector and the second principal component pre-aligned grinder speed feature vector to obtain the grinder speed adjustment judgment feature vector.

8. The distributed automatic batching control system according to claim 7, characterized in that: The feature alignment subunit is used to: Calculating the square of the second norm of the feed crushing state associated feature vector to obtain the feed crushing state feature intensity value; Calculating the natural exponential function value for each eigenvalue of the feed crushing state associated eigenvector to obtain an indexed feed crushing state associated eigenvector; Dividing each characteristic value of the indexed feed crushing state associated characteristic vector by the sum of the indexed feed crushing state associated characteristic vectors to obtain a normalized feed crushing state associated characteristic vector; Each eigenvalue of the normalized feed crushing state associated feature vector is multiplied by the feed crushing state feature intensity value to obtain the pre-aligned feed crushing state associated feature vector.

9. The distributed automatic batching control system according to claim 8, characterized in that: The principal component feature generating subunit is used for: Performing eigenvalue-based decomposition on the pre-aligned feed pulverization state associated feature vector to obtain a sequence of first eigen-decomposed pre-aligned feed pulverization state associated feature vectors and a first diagonal matrix; The first eigenvector and the second eigenvector in the sequence of the first eigendecomposition pre-aligned feed crushing state associated eigenvector are cascaded to obtain the first principal component pre-aligned feed crushing state associated eigenvector.

10. A distributed automatic batching control method, characterized in that: include: Acquire feed crushing monitoring video collected by a feed monitoring module in a distributed system, crushed raw material status data collected by a raw material information extraction module in a distributed system, and speed values ​​of a crusher at multiple predetermined time points collected by a crusher speed extraction module in a distributed system; Extracting a feed crushing state associated feature vector and a grinder speed feature vector from the feed crushing monitoring video collected by the feed monitoring module in the distributed system, the crushing raw material state data collected by the raw material information extraction module in the distributed system, and the grinder speed values ​​at multiple predetermined time points collected by the grinder speed extraction module in the distributed system; Based on the feed crushing state associated feature vector and the grinder speed feature vector, it is determined whether the grinder speed needs to be adjusted.

Images