A distribution method and system for flow balance lubrication of multiple hydraulic systems

Abstract

The present invention discloses a distribution method and system for flow balance lubrication of multiple hydraulic systems, which relates to the field of hydraulic transmission technology. The method includes obtaining a feature vector library of multiple hydraulic systems; formulating a targeted flow balance lubrication strategy; understanding the interaction between components and the flow demand change trend; building and training a two-way sequence recommendation model; recommending the optimal flow distribution plan for each component; the system includes a feature vector library construction module, a cluster analysis and component portrait construction module, a topic model construction and serialization module, a two-way sequence recommendation model training module, and a flow distribution recommendation module. The present invention more finely understands and optimizes the flow distribution of multiple hydraulic systems, thereby improving system efficiency and performance; and realizes in-depth analysis and intelligent management of multiple hydraulic systems.

Description

Technical Field

[0001] The present invention relates to the field of hydraulic transmission technology, and in particular to a distribution method and system for flow balance lubrication of multiple hydraulic systems. Background Art

[0002] A multi-hydraulic system is a system that uses hydraulic oil or other working media to transfer energy and control multiple hydraulic actuators (such as hydraulic cylinders, hydraulic motors, etc.). In this system, flow and pressure are two key parameters that need to be precisely controlled to ensure efficient and stable operation of the system. In a multi-hydraulic system, by precisely distributing and controlling the flow of liquid to each actuator, it is ensured that they can obtain the appropriate amount of liquid flow at the same time or on demand. This is to prevent the performance of some components from being degraded due to insufficient flow, or energy waste or system damage caused by excessive flow.

[0003] In hydraulic systems, lubrication generally refers to ensuring that there is enough fluid flow between moving parts (such as pump gears, valve sliders, etc.) to reduce friction and wear and improve system efficiency and life. Flow-balanced lubrication places special emphasis on ensuring that all required parts receive proper lubrication while keeping the hydraulic system working properly. Distribution is the ability to dynamically adjust and distribute flow to multiple hydraulic system actuators. This requires taking into account the overall needs of the system, the priority of each component, and how to maintain a balance of flow under different operating conditions.

[0004] In the prior art, for example, Chinese patent 202210481792.3 discloses a method, processor and engineering machinery for hydraulic oil flow distribution, which includes a boom, a hydraulic system, a prime mover and multiple command elements, and determines the distribution flow of hydraulic oil by obtaining the prime mover speed, receiving instructions, judging the required flow and other steps. However, the above method still has the following shortcomings: no detailed component portraits are built for different actuators or mechanical operations, so the specific needs and characteristics of each actuator may not be fully considered when distributing the flow. At the same time, intelligent management is not achieved, but it relies on preset processes and rules.

[0005] Currently, no effective solution has been proposed for the problems in the related technologies. Summary of the invention

[0006] In view of the problems in the related art, the present invention proposes a distribution method and system for flow balanced lubrication of multiple hydraulic systems to overcome the above-mentioned technical problems existing in the existing related art.

[0007] To this end, the specific technical solution adopted by the present invention is as follows:

[0008] According to one aspect of the present invention, a method for distributing flow-balanced lubrication for multiple hydraulic systems is provided, and the method for distributing flow-balanced lubrication for multiple hydraulic systems comprises the following steps:

[0009] S1. Acquire relevant data of multiple hydraulic systems, and use vectorization technology to convert the relevant data of multiple hydraulic systems into vector form to obtain a feature vector library of multiple hydraulic systems.

[0010] S2. Cluster analysis is performed on various data of multiple hydraulic systems, and component portraits are constructed, while targeted flow balance lubrication strategies are formulated.

[0011] S3. Based on component portraits and using the Huffman tree algorithm, a topic model of the hydraulic system is constructed, and the behaviors of the components are serialized to understand the interactions between components and the changing trends of flow demand.

[0012] S4. A bidirectional sequence recommendation model is constructed and trained based on the interaction between components in multi-hydraulic systems, the flow demand change trend, the characteristic vectors of components, and the flow balance lubrication strategy.

[0013] S5. Recommend the optimal flow distribution scheme for each component through the bidirectional sequence recommendation model and according to the current multi-hydraulic system status and the expected flow balance target.

[0014] Furthermore, cluster analysis is performed on various data of multiple hydraulic systems, component portraits are constructed, and targeted flow balance lubrication strategies are formulated, including the following steps:

[0015] S21, uniformly modeling and characterizing the data in the feature vector library to obtain analysis data in a uniform format;

[0016] S22, fusing and classifying the analysis data through cluster analysis, and establishing a correlation relationship diagram between the data source set and the decision requirement set based on distance entropy;

[0017] S23, using the random walk model and maximum likelihood similarity algorithm to analyze the association graph to identify closely related component groups;

[0018] S24, performing in-depth cluster analysis on the data classification results in S22 according to the characteristic data of the multi-hydraulic system, using a k-means algorithm to identify multi-hydraulic behaviors and abnormal patterns, and obtaining cluster analysis results;

[0019] S25, constructing a component profile based on the parameters of each component in a closely related component group;

[0020] S26, refining the cluster analysis results according to the component portraits, more accurately identifying the behaviors and states of the multi-hydraulic systems, and obtaining optimized clustering results;

[0021] S27. Combine the optimized clustering results and component portraits to formulate targeted flow balance lubrication strategies.

[0022] Further, the association graph is analyzed using a random walk model and a maximum likelihood similarity algorithm to determine closely related component groups, including the following steps:

[0023] S231, converting the association relationship graph into a graph model, wherein nodes represent components of the multi-hydraulic system and edges represent relationships between the components;

[0024] S232. Randomly select a starting node on the association graph, where the node represents a component, and initialize a walk probability distribution to indicate the probability of the walker staying at each node, and perform a random walk:

[0025] S233, according to the connection relationship of the association relationship graph, the random walker moves from the current node to the adjacent node, and updates the stay probability of the random walker on each node;

[0026] S234, calculating the similarity between nodes using the maximum likelihood estimation method;

[0027] S235, setting a similarity threshold, wherein the node pairs exceeding the similarity threshold are node pairs with high similarity, and identifying closely related component groups based on the node pairs with high similarity;

[0028] S236, iteratively executing random walk and similarity calculation until the stay probability distribution converges or reaches a preset number of iterations.

[0029] Furthermore, when performing a random walk, a random walk algorithm is executed on the association graph to calculate the probability that each component is correctly linked;

[0030] Iterate the calculation until the probability distribution vector converges;

[0031] Among them, the random walk formula is:

[0032] ;

[0033] In the formula, r is the probability distribution vector;

[0034] r (t+1) represents the updated probability distribution vector, t represents the number of iterations, r (t) Represents the probability distribution vector of the previous iteration;

[0035] c represents the random jump probability;

[0036] M The transition probability matrix represents the association between nodes;

[0037] q A probability distribution vector representing random jumps.

[0038] Furthermore, based on the component portraits and using the Huffman tree algorithm, a topic model of the hydraulic system is constructed, and the behaviors of the components are serialized to understand the interactions between the components and the trend of flow demand changes. The following steps are included:

[0039] S31. Analyze component portraits to determine the relevance and dependency between components;

[0040] S32. Use the Huffman tree algorithm to build a topic model for multi-hydraulic systems and represent the relationships between components as topics;

[0041] S33, training a topic model to learn the topic structure in the multi-hydraulic system and determine the weights of each component under the topic;

[0042] S34. Organize historical data and organize them in time series to analyze how the behaviors of components change over time;

[0043] S35. Use sequence models to model component behaviors and capture the dynamic relationships among them;

[0044] S36. Train the sequence model to learn the patterns and trends of component behavior, including the changing trends of traffic demand and the interactions between components.

[0045] Further, analyzing the component portraits to determine the associations and dependencies between components includes the following steps:

[0046] S311, collecting component operation data;

[0047] S312, analyzing the characteristics of each component and identifying the similarities and differences between the components;

[0048] S313. Determine the correlation and dependency between components to provide a basis for building a topic model.

[0049] Furthermore, the Huffman tree algorithm is used to construct a topic model of the multi-hydraulic system, and the relationship between components is represented as a topic, which includes the following steps:

[0050] S321. According to the correlation and dependency between components, the granularity of each component is calculated by using the principal component analysis method to obtain the granularity value of each component, and the granularity value is used as a weight;

[0051] S322, constructing a topic model using the Huffman tree algorithm, wherein each leaf node represents a component and the internal nodes represent the relationship topics between components;

[0052] S323, merge the two internal nodes with the smallest weights, and create a new internal node, until a complete Huffman tree is constructed;

[0053] S324. Assign a unique code to each node of the Huffman tree to represent the hierarchical relationship between components, and obtain a topic model representing the relationship between multiple hydraulic system components.

[0054] Furthermore, training the topic model to learn the topic structure in the multi-hydraulic system and determining the weights of each component under the topic includes the following steps:

[0055] S331, training a topic model using historical data of multiple hydraulic systems;

[0056] S332. By optimizing the algorithm, the weights of the nodes in the topic model are adjusted so that the topic model better represents the relationship between the components.

[0057] S333. Determine the weight distribution of each component under each topic.

[0058] Furthermore, constructing and training a bidirectional sequence recommendation model based on the interaction between components in the multi-hydraulic system, the flow demand change trend, the characteristic vector of the components and the flow balance lubrication strategy includes the following steps:

[0059] S41, preprocessing the interaction between components in the multi-hydraulic system, the flow demand change trend, the characteristic vector of the component and the flow balance lubrication strategy as input data to obtain preprocessed data;

[0060] S42. Construct a bidirectional sequence recommendation model based on a bidirectional long short-term memory network, and input the preprocessed data into the bidirectional sequence recommendation model for training.

[0061] According to another aspect of the present invention, a distribution system for flow balanced lubrication of multiple hydraulic systems is provided, the distribution system for flow balanced lubrication of multiple hydraulic systems comprising a feature vector library construction module, a cluster analysis and component portrait construction module, a topic model construction and serialization module, a bidirectional sequence recommendation model training module and a flow distribution recommendation module;

[0062] The feature vector library construction module is used to obtain relevant data of multiple hydraulic systems, and convert the relevant data of multiple hydraulic systems into vector form using vectorization technology to obtain a feature vector library of multiple hydraulic systems.

[0063] The cluster analysis and component portrait construction module is used to perform cluster analysis on various data of multiple hydraulic systems, construct component portraits, and formulate targeted flow balance lubrication strategies.

[0064] The topic model building and serialization module is used to build a topic model of the hydraulic system based on component portraits and use the Huffman tree algorithm, and serialize the behavior of the components to understand the interactions between components and the changing trends of flow demand.

[0065] The bidirectional sequence recommendation model training module is used to build and train a bidirectional sequence recommendation model based on the interaction between components in a multi-hydraulic system, the flow demand change trend, the characteristic vector of the components, and the flow balance lubrication strategy.

[0066] The flow distribution recommendation module is used to recommend the optimal flow distribution scheme for each component according to the current multi-hydraulic system status and the expected flow balance target through a bidirectional sequence recommendation model.

[0067] The beneficial effects of the present invention are:

[0068] (1) The present invention provides a method and system for allocating flow-balanced lubrication for multiple hydraulic systems. The data of multiple hydraulic systems are classified through cluster analysis to construct component portraits, thereby formulating customized flow-balanced lubrication strategies for each component or component group. The Huffman tree algorithm is used to construct a topic model for the hydraulic system, and the component behaviors are serialized to understand the interactions between components and changes in flow demand. The trained bidirectional sequence recommendation model is used to recommend the optimal flow distribution plan for each component based on the current system status and flow balance target. The present invention more finely understands and optimizes the flow distribution of multiple hydraulic systems, thereby improving system efficiency and performance. Through the interaction and coordination of each step, in-depth analysis and intelligent management of multiple hydraulic systems can be achieved.

[0069] (2) Through cluster analysis and analysis of association graphs, the association groups of components are identified, and a targeted flow balance lubrication strategy is formulated based on this information. The association graphs between components are analyzed using random walk and maximum likelihood similarity algorithms to determine closely related component groups.

[0070] (3) Based on component portraits and granularity values, a topic model representing the relationship between multi-hydraulic system components is constructed using the Huffman tree algorithm. The topic model is trained to learn the topic structure in the multi-hydraulic system and determine the weight distribution of each component under the topic. A bidirectional long short-term memory network is constructed based on system data and flow balancing strategy, and the model is trained to recommend flow distribution solutions. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0072] Figure 1 is a flow chart of a method for distributing flow-balanced lubrication for multiple hydraulic systems according to an embodiment of the present invention;

[0073] Figure 2 It is a principle block diagram of a distribution system for flow-balanced lubrication of multiple hydraulic systems according to an embodiment of the present invention.

[0074] In the figure:

[0075] 1. Feature vector library construction module; 2. Cluster analysis and component portrait construction module; 3. Topic model construction and serialization module; 4. Bidirectional sequence recommendation model training module; 5. Traffic distribution recommendation module. DETAILED DESCRIPTION

[0076] To further illustrate each embodiment, the present invention provides drawings, which are part of the disclosure of the present invention and are mainly used to illustrate the embodiments and can be used in conjunction with the relevant descriptions in the specification to explain the operating principles of the embodiments. With reference to these contents, ordinary technicians in the field should be able to understand other possible implementations and advantages of the present invention. The components in the figures are not drawn to scale, and similar component symbols are generally used to represent similar components.

[0077] According to an embodiment of the present invention, a distribution method and system for flow-balanced lubrication of multiple hydraulic systems are provided.

[0078] The present invention is further described with reference to the accompanying drawings and specific embodiments. Figure 1 As shown, according to one embodiment of the present invention, a distribution method for flow balanced lubrication of multiple hydraulic systems is provided, and the distribution method for flow balanced lubrication of multiple hydraulic systems comprises the following steps:

[0079] S1. Obtain relevant data of multiple hydraulic systems, including performance parameters, working status, flow requirements, sensor data, operation history, maintenance records, etc.; and use vectorization technology (such as Service2Vector) to convert the relevant data of multiple hydraulic systems into vector form to obtain the feature vector library of multiple hydraulic systems.

[0080] Using vectorization technology to convert multi-hydraulic system related data into vector form includes the following steps:

[0081] Choose an appropriate vectorization technique, such as Service2Vector or other suitable embedding techniques.

[0082] The extracted features are converted into numerical vectors, each of which represents the state of the system at a certain moment.

[0083] Perform necessary dimensionality reduction on the vector to reduce computational complexity and retain important information.

[0084] S2. Cluster analysis is performed on various data of multiple hydraulic systems, and component portraits are constructed, while targeted flow balance lubrication strategies are formulated.

[0085] In a further embodiment, cluster analysis is performed on various data of multiple hydraulic systems, component portraits are constructed, and a targeted flow balance lubrication strategy is formulated, including the following steps:

[0086] S21. Model and characterize the data in the feature vector library in a unified manner to ensure the consistency and comparability of the data during the analysis process and obtain the analysis data in a unified format. This helps to improve the accuracy and reliability of subsequent cluster analysis. The feature vector library of multiple hydraulic systems includes the working parameters of various pumps, valves, and cylinders. These parameters are standardized, such as converting pressure values ​​into standard units (such as bar or Pascal) to ensure that data from different devices can be compared with each other.

[0087] S22. The analysis data is integrated and classified through cluster analysis, and a correlation diagram between the data source set and the decision requirement set (such as flow balance requirement) is established based on the distance entropy. For example, in a multi-hydraulic system, cluster analysis is performed based on parameters such as the working frequency and pressure requirement of the hydraulic cylinder, similar hydraulic cylinders are classified into one category, and a correlation diagram between them and the flow balance requirement is drawn.

[0088] S23. Use the random walk model and maximum likelihood similarity algorithm to analyze the association relationship graph to identify closely related component groups. For example, a hydraulic cylinder group always works together in a specific operation mode. The random walk model identifies this component group and confirms the close association between them through maximum likelihood similarity analysis.

[0089] S24. Perform in-depth cluster analysis on the data classification results in S22 according to the characteristic data of the multi-hydraulic system (such as flow rate and pressure), use the k-means algorithm to identify multi-hydraulic behaviors and abnormal patterns, and obtain cluster analysis results; further subdivide the hydraulic cylinder groups through the k-means algorithm to identify cylinder groups with consistent performance under specific pressure, or find that some cylinder groups are prone to failure under high pressure.

[0090] S25. Build component profiles based on parameters such as the operation mode, workload, and failure history of each component in a closely related component group. For example, create a profile for each hydraulic cylinder, including its usage frequency, maintenance times, failure rate, and other information, which helps to better understand the working characteristics and maintenance requirements of each cylinder.

[0091] S26. Refine the cluster analysis results based on component portraits to more accurately identify the behavior and status of multiple hydraulic systems, and obtain optimized clustering results for improving flow balance lubrication strategies. Some hydraulic cylinders have subtle behavioral differences even in the same category. These differences can be revealed by refining the clusters, providing a basis for formulating more accurate maintenance plans.

[0092] S27. Combine the optimized clustering results and component portraits to formulate a targeted flow balance lubrication strategy. The flow balance lubrication strategy includes:

[0093] Analyze the flow demand and lubrication requirements of each component or component group based on component profiling and clustering results.

[0094] Set flow allocation priorities for different components or groups of components based on the key performance indicators of the multi-hydraulic system and the importance of the components.

[0095] Develop specific flow-balanced lubrication strategies, including flow distribution schemes, dynamic adjustment mechanisms, and emergency response measures.

[0096] For example, a more frequent lubrication and maintenance schedule might be instituted for hydraulic cylinders that experience heavy workloads and high failure rates, whereas a more relaxed lubrication strategy might be adopted for cylinders that experience low and stable workloads.

[0097] It should be noted that by clustering the various data of the multi-hydraulic system and building component portraits, we can have a deeper understanding of the operating status of the multi-hydraulic system and develop a more refined flow balance lubrication strategy for each component, thereby improving the overall performance and reliability of the multi-hydraulic system.

[0098] In a further embodiment, analyzing the association relationship graph using a random walk model and a maximum likelihood similarity algorithm to determine closely related component groups includes the following steps:

[0099] S231. Convert the association relationship graph into a graph model, where nodes represent components of the multi-hydraulic system and edges represent relationships between components (e.g., flow dependency, similarity, or functional correlation).

[0100] S232, randomly select a starting node on the association relationship graph, the node represents a component, and initialize a walk probability distribution, usually a uniform distribution, indicating the probability of the walker staying on each node, and perform random walk: if there are 5 components, the initial probability of each component being the starting node is 1 / 5. Starting from any component, the random walk process is about to start.

[0101] S233. According to the connection relationship of the association relationship graph, the random walker moves from the current node to the adjacent node, and updates the stay probability of the random walker on each node, which is usually achieved through a transition probability matrix.

[0102] S234. The similarity between nodes is calculated using the maximum likelihood estimation method, that is, by comparing the distribution of node stay probabilities, and the similarity calculation uses cosine similarity, Jaccard similarity or other suitable metrics; if two cylinders have similar stay probability distributions during random walks, this may indicate that they have similar behaviors or dependencies at work.

[0103] S235 . Set a similarity threshold. Node pairs exceeding the similarity threshold are considered to have high similarity. Based on the node pairs with high similarity, identify closely related component groups.

[0104] S236, iteratively executing random walk and similarity calculation until the stay probability distribution converges or reaches a preset number of iterations.

[0105] It should be noted that identifying groups of components that are closely related in function or demand in a multi-hydraulic system is very valuable for optimizing the maintenance and operation strategy of the multi-hydraulic system. For example, if two cylinders always act together and the similarity score between them is high, then I consider coordinating their maintenance work to improve efficiency.

[0106] In a further embodiment, when performing the random walk, a random walk algorithm is performed on the association relationship graph to calculate the probability that each component is correctly linked;

[0107] Iterate the calculation until the probability distribution vector converges;

[0108] Among them, the random walk formula is:

[0109] ;

[0110] In the formula, r is a probability distribution vector, indicating the probability of each node being correctly linked.

[0111] r (t+1) represents the updated probability distribution vector, trepresents the number of iterations, r (t) Represents the probability distribution vector of the previous iteration.

[0112] c Represents the random jump probability, which means randomly jumping to other nodes with a certain probability during the walk.

[0113] M The transition probability matrix represents the association between nodes, where the elements M ij Represents a slave node j Transfer to Node i The traffic demand factor is used as the weight between nodes to more accurately reflect the traffic relationship between nodes.

[0114] q The probability distribution vector of random jumps represents the probability of jumping to each node in the absence of links. The feature vector of the component is used as part of the probability distribution vector of random jumps, so that the influence of the component characteristics on the jump behavior is taken into account during the walk.

[0115] S3. Based on component portraits and using the Huffman tree algorithm, a topic model of the hydraulic system is constructed, and the behaviors of the components are serialized to understand the interactions between components and the changing trends of flow demand.

[0116] In a further embodiment, constructing a topic model of a hydraulic system based on component portraits and using a Huffman tree algorithm, and serializing the behaviors of the components to understand the interactions between the components and the flow demand change trend includes the following steps:

[0117] S31. Analyze component portraits to determine the correlation and dependency between components; analyze the detailed characteristics of each component, such as operating data, performance parameters, fault history, etc., to determine the correlation and dependency between components.

[0118] S32. Use the Huffman tree algorithm to build a topic model for multi-hydraulic systems, and represent the relationship between components as topics; encode the relationship between the cylinder and the valve as a topic, and different combinations of cylinders and valves can correspond to different topics.

[0119] S33. Train a topic model to learn the topic structure in multi-hydraulic systems and determine the weights of each component under the topic; for example, the topic model identifies which cylinders and valves usually work together when performing a specific task, thereby assigning higher weights to these component combinations.

[0120] S34. Sort out historical data and organize them in time series to analyze how the behavior of components changes over time; arrange the daily operation records of each component in the hydraulic system in chronological order to analyze the working mode and flow changes of the components in different time periods.

[0121] S35. Use sequence models (such as RNN or LSTM) to model component behaviors and capture the dynamic relationships between them.

[0122] S36. Train the sequence model to learn the patterns and trends of component behaviors, including the changing trends of traffic demand and the interactions between components.

[0123] A topic model and sequence model of a multi-hydraulic system are constructed. The topic model and sequence model help understand the interactions between components and the changing trends of flow demand, thereby providing support for the optimization and maintenance of the multi-hydraulic system.

[0124] In a further embodiment, analyzing the component portraits to determine the association and dependency between components includes the following steps:

[0125] S311. Collect the operating data of the components, including performance parameters, working status, flow requirements, etc.; S312. Analyze the characteristics of each component and identify the similarities and differences between components; S313. Determine the correlation and dependency between components to provide a basis for building a topic model. For example, if it is found that the flow demand of a cylinder is closely related to the opening state of a specific valve, then it can be considered that there is a dependency between the two components.

[0126] In a further embodiment, constructing a topic model of a multi-hydraulic system using a Huffman tree algorithm, and representing the relationships between components as topics includes the following steps:

[0127] S321. According to the association and dependency relationship between the components, the granularity of each component is calculated using the principal component analysis method to indicate the importance or characteristics of the component, and the granularity value of each component is obtained, and the granularity value is used as a weight;

[0128] Among them, the principal component analysis method is used for dimensionality reduction and feature extraction. When using the principal component analysis method to calculate the granularity of each component, the operating data of the component is standardized, the covariance matrix is ​​calculated using the standardized data, and the covariance matrix is ​​eigendecomposed to obtain the eigenvalues ​​and corresponding eigenvectors. According to the size of the eigenvalues, the most important eigenvectors are selected as principal components, and the original data is projected onto the selected principal components to obtain the representation of each component in the new feature space. This new representation is the granularity of the component, where the weight of each principal component is the importance of the component on the principal component.

[0129] S322, constructing a topic model using the Huffman tree algorithm, wherein each leaf node represents a component and the internal nodes represent the relationship topics between components;

[0130] S323, merge the two internal nodes with the smallest weights, and create a new internal node, until a complete Huffman tree is constructed;

[0131] S324. Assign a unique code to each node of the Huffman tree to represent the hierarchical relationship between components, and obtain a topic model representing the relationship between multiple hydraulic system components.

[0132] It should be noted that the Huffman tree algorithm is an optimal binary tree constructed based on weights. By using components as leaf nodes and their dependencies as internal nodes, the Huffman tree can clearly represent the dependencies and associations between components. The tree structure makes it easier to identify the priority and path of traffic distribution. The Huffman tree organizes component relationships in an optimal way, which helps to reduce unnecessary traffic waste and achieve balanced lubrication of traffic, thereby saving energy and maintenance costs.

[0133] In a further embodiment, training a topic model to learn the topic structure in the multi-hydraulic system and determining the weights of each component under the topic includes the following steps:

[0134] S331. Use historical data of multiple hydraulic systems to train the topic model; S332. Adjust the weights of nodes in the topic model through optimization algorithms so that the topic model can better represent the relationship between components; S333. Determine the weight distribution of each component under each topic to reflect its role and influence in the multiple hydraulic systems.

[0135] It should be noted that the purpose of adjusting the weights of nodes in the topic model through an optimization algorithm is to make the topic model more accurately reflect the actual relationship between components. Use an optimization algorithm to adjust the weights. The optimization algorithm can be gradient descent, genetic algorithm, simulated annealing, or other advanced optimization techniques. In each iteration, the optimization algorithm calculates a new weight update based on the objective function and the current weights. This process is repeated until the value of the objective function converges or the predetermined number of iterations is reached.

[0136] S4. A bidirectional sequence recommendation model is constructed and trained based on the interaction between components in multi-hydraulic systems, the flow demand change trend, the characteristic vectors of components, and the flow balance lubrication strategy.

[0137] In a further embodiment, constructing and training a bidirectional sequence recommendation model based on the interaction between components in a multi-hydraulic system, the flow demand change trend, the characteristic vectors of the components, and the flow balance lubrication strategy includes the following steps:

[0138] S41, the interaction between components in the multi-hydraulic system, the flow demand change trend, the characteristic vector of the components and the flow balance lubrication strategy are preprocessed as input data to obtain preprocessed data, including operations such as feature normalization and sequence filling, so as to facilitate model training and use;

[0139] S42. A bidirectional sequence recommendation model is constructed based on a bidirectional long short-term memory network, and the preprocessed data is input into the bidirectional sequence recommendation model for training. During the training process, the model will learn the relationships between components in the system and try to provide the best traffic allocation solution based on the traffic balancing strategy.

[0140] S5. Recommend the optimal flow distribution scheme for each component through the bidirectional sequence recommendation model and according to the current multi-hydraulic system status and the expected flow balance target.

[0141] At the same time, decision visualization tools or explanatory reports are provided to help operators better understand the recommendation logic of the bidirectional sequence recommendation model.

[0142] The present invention describes the behavior characteristics of each component through cluster analysis and component portrait construction. Based on the component portrait, the topic model of the hydraulic system is constructed using the Huffman tree algorithm to learn the patterns and trends of component behavior. At the same time, a recommendation model is constructed based on a bidirectional long short-term memory network and trained. Finally, the trained bidirectional sequence recommendation model is used to recommend the optimal flow distribution plan for each component in combination with the current system status and the expected flow balance target.

[0143] According to another embodiment of the present invention, a distribution system for flow balanced lubrication of multiple hydraulic systems is provided, and the distribution system for flow balanced lubrication of multiple hydraulic systems includes a feature vector library construction module 1, a clustering analysis and component portrait construction module 2, a topic model construction and serialization module 3, a bidirectional sequence recommendation model training module 4 and a flow distribution recommendation module 5.

[0144] The feature vector library construction module 1 is used to obtain relevant data of multiple hydraulic systems, and convert the relevant data of multiple hydraulic systems into vector form using vectorization technology to obtain a feature vector library of multiple hydraulic systems.

[0145] The cluster analysis and component portrait construction module 2 is used to perform cluster analysis on various data of multiple hydraulic systems, construct component portraits, and formulate targeted flow balance lubrication strategies.

[0146] The topic model building and serialization module 3 is used to build a topic model of the hydraulic system based on component portraits and use the Huffman tree algorithm, and serialize the behavior of the components to understand the interaction between the components and the change trend of the flow demand.

[0147] The bidirectional sequence recommendation model training module 4 is used to construct and train a bidirectional sequence recommendation model based on the interaction between components in a multi-hydraulic system, the flow demand change trend, the characteristic vector of the component and the flow balance lubrication strategy.

[0148] The flow distribution recommendation module 5 is used to recommend an optimal flow distribution scheme for each component through a bidirectional sequence recommendation model and according to the current multi-hydraulic system state and the expected flow balance target.

[0149] In summary, the present invention provides a method and system for allocating flow balance lubrication of multiple hydraulic systems. The data of multiple hydraulic systems are classified by cluster analysis, and component portraits are constructed, so as to formulate customized flow balance lubrication strategies for each component or component group. The topic model of the hydraulic system is constructed by using the Huffman tree algorithm, and the component behaviors are serialized to understand the interaction between components and the change of flow demand. The trained bidirectional sequence recommendation model is used to recommend the optimal flow distribution scheme for each component according to the current system state and flow balance target. The present invention understands and optimizes the flow distribution of multiple hydraulic systems more finely, thereby improving the efficiency and performance of the system. Through the interaction and coordination of each step, in-depth analysis and intelligent management of multiple hydraulic systems can be achieved. Through cluster analysis and analysis of association relationship graphs, the association groups of components are identified, and targeted flow balance lubrication strategies are formulated based on this information. The association relationship graphs between components are analyzed using random walk and maximum likelihood similarity algorithms to determine closely related component groups. Based on component portraits and granularity values ​​(weights), a topic model representing the relationship between components of multiple hydraulic systems is constructed using the Huffman tree algorithm. The topic model is trained to learn the topic structure in the multi-hydraulic system and determine the weight distribution of each component under the topic. A bidirectional long short-term memory network is constructed based on system data and flow balancing strategy, and the model is trained to recommend flow distribution solutions.

[0150] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A distribution method for flow balance lubrication of multiple hydraulic systems, characterized in that: The distribution method for flow-balanced lubrication of multiple hydraulic systems comprises the following steps: S1. Acquire relevant data of multiple hydraulic systems, and use vectorization technology to convert the relevant data of multiple hydraulic systems into vector form to obtain a feature vector library of multiple hydraulic systems; S2. Cluster analysis of various data of multiple hydraulic systems, construction of component portraits, and formulation of targeted flow balance lubrication strategies; S3. Based on component portraits and using the Huffman tree algorithm, a topic model of the hydraulic system is constructed, and the behaviors of the components are serialized to understand the interactions between components and the changing trends of flow demand; S4, construct and train a bidirectional sequence recommendation model based on the interaction between components in a multi-hydraulic system, the flow demand change trend, the characteristic vector of the components, and the flow balance lubrication strategy; S5, recommending an optimal flow distribution scheme for each component based on the current multi-hydraulic system state and the expected flow balance target through a bidirectional sequence recommendation model; The cluster analysis of various data of multiple hydraulic systems, the construction of component portraits, and the formulation of targeted flow balance lubrication strategies include the following steps: S21, uniformly modeling and characterizing the data in the feature vector library to obtain analysis data in a uniform format; S22, fusing and classifying the analysis data through cluster analysis, and establishing a correlation relationship diagram between the data source set and the decision requirement set based on distance entropy; S23, using the random walk model and maximum likelihood similarity algorithm to analyze the association graph to identify closely related component groups; S24, performing in-depth cluster analysis on the data classification results in S22 according to the characteristic data of the multi-hydraulic system, using a k-means algorithm to identify multi-hydraulic behaviors and abnormal patterns, and obtaining cluster analysis results; S25, constructing a component profile based on the parameters of each component in a closely related component group; S26, refining the cluster analysis results according to the component portraits, more accurately identifying the behaviors and states of the multi-hydraulic systems, and obtaining optimized clustering results; S27. Combining the optimized clustering results and component portraits, formulate targeted flow balance lubrication strategies; The method of building a topic model of the hydraulic system based on component portraits and using the Huffman tree algorithm, and serializing the behaviors of the components to understand the interactions between the components and the flow demand change trend includes the following steps: S31. Analyze component portraits to determine the relevance and dependency between components; S32. Use the Huffman tree algorithm to build a topic model for multi-hydraulic systems and represent the relationships between components as topics; S33, training a topic model to learn the topic structure in the multi-hydraulic system and determine the weights of each component under the topic; S34. Organize historical data and organize them in time series to analyze how the behaviors of components change over time; S35. Use sequence models to model component behaviors and capture the dynamic relationships among them; S36, training the sequence model to learn patterns and trends of component behaviors, including changing trends of traffic demand and interactions between components; The method of using the Huffman tree algorithm to construct a topic model of a multi-hydraulic system and expressing the relationship between components as topics includes the following steps: S321. According to the correlation and dependency between components, the granularity of each component is calculated by using the principal component analysis method to obtain the granularity value of each component, and the granularity value is used as a weight; S322, constructing a topic model using the Huffman tree algorithm, wherein each leaf node represents a component and the internal nodes represent the relationship topics between components; S323, merge the two internal nodes with the smallest weights, and create a new internal node, until a complete Huffman tree is constructed; S324. Assign a unique code to each node of the Huffman tree to represent the hierarchical relationship between components, and obtain a topic model representing the relationship between multiple hydraulic system components.

2. A distribution method for flow balance lubrication of multiple hydraulic systems according to claim 1, characterized in that: The method of analyzing the association relationship graph using the random walk model and the maximum likelihood similarity algorithm to determine closely related component groups includes the following steps: S231, converting the association relationship graph into a graph model, wherein nodes represent components of the multi-hydraulic system and edges represent relationships between the components; S232. Randomly select a starting node on the association graph, where the node represents a component, and initialize a walk probability distribution to indicate the probability of the walker staying at each node, and perform a random walk: S233, according to the connection relationship of the association relationship graph, the random walker moves from the current node to the adjacent node, and updates the stay probability of the random walker on each node; S234, calculating the similarity between nodes using the maximum likelihood estimation method; S235, setting a similarity threshold, wherein the node pairs exceeding the similarity threshold are node pairs with high similarity, and identifying closely related component groups based on the node pairs with high similarity; S236, iteratively executing random walk and similarity calculation until the stay probability distribution converges or reaches a preset number of iterations.

3. A distribution method for flow balance lubrication of multiple hydraulic systems according to claim 2, characterized in that: When performing the random walk, a random walk algorithm is executed on the association relationship graph to calculate the probability that each component is correctly linked; Iterate the calculation until the probability distribution vector converges; Among them, the random walk formula is: ； In the formula, r is the probability distribution vector; r (t+1) represents the updated probability distribution vector, t represents the number of iterations, r (t) Represents the probability distribution vector of the previous iteration; c represents the random jump probability; M The transition probability matrix represents the association between nodes; q A probability distribution vector representing random jumps.

4. A distribution method for flow balance lubrication of multiple hydraulic systems according to claim 1, characterized in that: Analyzing the component portraits to determine the association and dependency between components includes the following steps: S311, collecting component operation data; S312, analyzing the characteristics of each component and identifying the similarities and differences between the components; S313. Determine the correlation and dependency between components to provide a basis for building a topic model.

5. A distribution method for flow balance lubrication of multiple hydraulic systems according to claim 4, characterized in that: The training of the topic model to learn the topic structure in the multi-hydraulic system and determine the weight of each component under the topic includes the following steps: S331, training a topic model using historical data of multiple hydraulic systems; S332. By optimizing the algorithm, the weights of the nodes in the topic model are adjusted so that the topic model better represents the relationship between the components. S333. Determine the weight distribution of each component under each topic.

6. A distribution method for flow balance lubrication of multiple hydraulic systems according to claim 5, characterized in that: The method of constructing and training a bidirectional sequence recommendation model based on the interaction between components in a multi-hydraulic system, the flow demand change trend, the characteristic vector of the components, and the flow balance lubrication strategy comprises the following steps: S41, preprocessing the interaction between components in the multi-hydraulic system, the flow demand change trend, the characteristic vector of the component and the flow balance lubrication strategy as input data to obtain preprocessed data; S42. Construct a bidirectional sequence recommendation model based on a bidirectional long short-term memory network, and input the preprocessed data into the bidirectional sequence recommendation model for training.

7. A distribution system for flow balance lubrication of multiple hydraulic systems, characterized in that: A method for distributing flow-balanced lubrication for multiple hydraulic systems according to any one of claims 1 to 6, wherein the distribution system for flow-balanced lubrication for multiple hydraulic systems comprises a feature vector library construction module, a cluster analysis and component portrait construction module, a topic model construction and serialization module, a bidirectional sequence recommendation model training module, and a flow distribution recommendation module; The feature vector library construction module is used to obtain relevant data of multiple hydraulic systems and convert the relevant data of multiple hydraulic systems into vector form using vectorization technology to obtain a feature vector library of multiple hydraulic systems; The cluster analysis and component portrait building module is used to perform cluster analysis on various data of multiple hydraulic systems, build component portraits, and formulate targeted flow balance lubrication strategies; The topic model building and serialization module is used to build a topic model of the hydraulic system based on component portraits and use the Huffman tree algorithm, and serialize the behavior of the components to understand the interaction between the components and the change trend of flow demand; The bidirectional sequence recommendation model training module is used to build and train a bidirectional sequence recommendation model based on the interaction between components in a multi-hydraulic system, the flow demand change trend, the characteristic vector of the components, and the flow balance lubrication strategy; The flow distribution recommendation module is used to recommend an optimal flow distribution solution for each component through a bidirectional sequence recommendation model and according to the current multi-hydraulic system state and the expected flow balance target.

Images