A method for recommending manufacturing capability services of industrial cloud robots
By building an industrial cloud robot manufacturing process information model and manufacturing capability service knowledge graph, combined with a knowledge representation learning model, the problems of industrial cloud robot manufacturing service knowledge management and personalized needs are solved, and efficient recommendations of industrial cloud robot manufacturing capability services are achieved. .
Patent Information
- Application Number
- CN202210488705.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-05-06
- Publication Date
- 2024-11-29
- Estimated Expiration
- 2042-05-06
AI Technical Summary
The existing industrial cloud robot manufacturing capability service recommendation methods are difficult to effectively manage and utilize industrial cloud robot manufacturing service knowledge, and it is difficult to meet personalized needs. It is impossible to quickly and effectively recommend high-quality and reliable services from massive manufacturing services.
Build an industrial cloud robot manufacturing process information model, establish a manufacturing capability service knowledge graph, combine knowledge representation learning model, propose service score optimization methods based on goals such as time, cost, energy consumption and error rate, and generate manufacturing that meets functional and non-functional needs. Recommended capabilities service solutions.
It realizes effective management and personalized recommendations of industrial cloud robot manufacturing capabilities services, improves the effectiveness and accuracy of recommendation results, and can select services according to user needs.
Smart Images

Figure CN114896414B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of industrial cloud robots and manufacturing services, and more specifically, to a method for recommending industrial cloud robot manufacturing capability services. Background Art
[0002] Industrial cloud robots encapsulate the manufacturing capabilities of physical industrial robots into user-accessible services, helping to enhance their manufacturing capabilities. With the widespread adoption of industrial robots in the manufacturing sector, the explosive growth of service information has led to information overload. It is crucial to quickly and effectively recommend high-quality and reliable services from this vast array of manufacturing services.
[0003] Knowledge graphs provide an effective method for organizing, managing, and understanding massive amounts of information. Knowledge representation learning, which represents entities and relationships as dense, low-dimensional vectors, is crucial for knowledge reasoning. Combining knowledge graphs with industrial cloud robots effectively organizes and manages knowledge about industrial cloud robot manufacturing capability services, effectively leveraging high-quality manufacturing capability service knowledge. Existing methods for recommending industrial cloud robot manufacturing capability services struggle to effectively manage and utilize this knowledge and meet personalized needs. Summary of the Invention
[0004] The technical problem to be solved by the present invention is to provide an industrial cloud robot manufacturing capability service recommendation method. On the premise of effectively managing the industrial cloud robot manufacturing service knowledge, the industrial cloud robot manufacturing capability service knowledge is effectively managed for different manufacturing tasks, and high-quality and reliable services are effectively recommended from massive manufacturing capability services.
[0005] The technical solution adopted by the present invention to solve the technical problem is to construct an industrial cloud robot manufacturing capability service recommendation method for recommending industrial cloud robot manufacturing capability services that meet functional requirements and non-functional requirements, including the following steps:
[0006] 1) Construct an industrial cloud robot manufacturing process information model that integrates the industrial cloud robot model, manufacturing task model, manufacturing capability service model, and manufacturing process model;
[0007] 2) Construct an industrial cloud robot manufacturing capability service knowledge graph that provides an information model of the industrial cloud robot manufacturing process;
[0008] 3) Propose a service scoring and optimization method that meets users' personalized needs and preferences, focusing on objectives such as time, cost, energy consumption, and error rate;
[0009] 4) Combined with the knowledge representation learning model for link prediction, a method for generating industrial cloud robot manufacturing capability service recommendation solutions that meet functional and non-functional requirements is proposed.
[0010] Preferably, in step 1), an industrial cloud robot manufacturing process information model is constructed by the following contents:
[0011] Establish an industrial cloud robot model; mainly describe the basic information and capability information of the industrial cloud robot. The basic information includes the static attributes of the industrial cloud robot such as ID, name, type, and geographic location; the capability information includes the technical parameters of the industrial cloud robot (such as maximum angular velocity, working environment temperature, maximum angular velocity) and the manufacturing capability services provided.
[0012] The industrial cloud robot model information is shown in Table 1 below.
[0013] Table 1 Industrial cloud robot model information
[0014]
[0015] Establish a manufacturing task model; mainly including task name, task type, task content, task objectives and other information.
[0016] The manufacturing task information is shown in Table 2 below.
[0017] Table 2 Manufacturing task information
[0018]
[0019] Establish a manufacturing capability service model; it mainly includes three aspects: basic attributes (such as service ID, service name, service type, etc.), functional attributes (i.e., tasks or subtasks that the service can complete) and non-functional attributes such as service time and energy consumption.
[0020] The manufacturing capability service model information is shown in Table 3 below.
[0021] Table 3 Manufacturing capability service model information
[0022]
[0023] Establish a manufacturing process model; describe the currently executed task status, task constraints, and manufacturing process data. The manufacturing process model is represented as follows:
[0024] Task_Process={Task_status,Task_Conditions,Process_data}
[0025] In step 2), the industrial cloud robot manufacturing capability service knowledge graph that provides the industrial cloud robot manufacturing process information model is constructed through the following content:
[0026] Knowledge graphs can formally describe real-world objects and their relationships. Nodes in a knowledge graph represent entities or concepts, while edges represent relationships and attributes between entities / concepts.
[0027] Based on the industrial cloud robot model, manufacturing task model, manufacturing capability service model, and manufacturing process model, combined with historical experience data and expert knowledge, through entity extraction, relationship extraction, and attribute extraction, and using entity disambiguation, we construct an industrial cloud robot manufacturing capability service knowledge graph. The knowledge graph G is represented as follows:
[0028] G=<E,R,S>
[0029] Where E={e1,e2,…,e |E|} is the entity set in the knowledge base, which contains |E| different entities; R = {r1, r2, …, r |R|} is the set of relations in the knowledge base, which contains |R| different relations; Represents the set of triples in the knowledge base.
[0030] In a unified entity and relational space, any triple in the industrial cloud robot manufacturing capability service knowledge graph can be described as:
[0031] S={(h,r,t)|h,t∈E,r∈R}
[0032] Among them, the vector h represents the head entity, t represents the tail entity, and r represents the relationship between the head entity and the tail entity.
[0033] In step 3), the service scoring optimization method that meets the user's personalized needs and preferences is as follows:
[0034] Based on the four non-functional attributes of time, cost, energy consumption, and error rate, combined with user-defined weights, the candidate services are scored and selected to meet user needs and preferences, as described below:
[0035] NFI={Time,Cost,Energy_cost,Failure_rate}
[0036] Among them, Time refers to the time required for the service to execute the subtask, including task execution time and logistics time; Cost refers to the cost of completing the task, including processing cost and logistics cost; Energy_cost refers to the energy consumption of completing the task; Failure_rate refers to the failure rate of the service to execute the task.
[0037] The dimensions of time, cost, energy consumption, and error rate are different, so each indicator needs to be normalized. The normalization formula is as follows:
[0038]
[0039] Among them, p i represents the i-th service in the candidate service set; It represents the jth attribute value of the i-th service in the candidate service set; max(p j ) represents the maximum value of the jth non-functional attribute; min(p j ) represents the minimum value of the jth non-functional attribute. Represents the normalized value of the jth non-functional attribute of the i-th service.
[0040] Calculate the score of each service in the candidate service set and return the service with the highest score to the user as the recommended result. The score calculation formula is as follows:
[0041]
[0042] Among them, ω1, ω2, ω3 and ω4 are the weights of time, cost, energy consumption and error rate given by the user.
[0043] In step 4), the method for generating a recommended solution for industrial cloud robot manufacturing capability service that meets functional and non-functional requirements is as follows:
[0044] Decompose the manufacturing task into its subtask sets;
[0045] The TransE algorithm is used to perform knowledge representation learning on the knowledge graph, and a knowledge representation learning model for the industrial cloud robot manufacturing capability service knowledge graph is acquired;
[0046] To further improve the effectiveness of the recommendation results of the industrial cloud robot manufacturing capability service, the negative sample generation strategy of TransE is optimized, which mainly includes the following steps:
[0047] (1) Determine the entity to be replaced based on Bernoulli distribution;
[0048] (2) Construct candidate entity sets;
[0049] (3) Construct candidate negative triples;
[0050] (4) Screening candidate negative triples;
[0051] (5) Determine whether the candidate negative triplet is in the training set and generate a negative triplet.
[0052] In the process of determining the replaced entity, for each relation r in the training triples, the number of triples containing relation r M and the number of head entities appearing on the left side of relation r N are counted. h , the number of tail entities N that appear on the right side of relation r t. The relation r is divided into four types: "1-1", "1-N", "N-1" and "NN".
[0053] Calculate the average number of tail entities corresponding to each head entity, tph, and the average number of head entities corresponding to each tail entity, hpt, and replace the head entity or tail entity of the triple according to the Bernoulli distribution. The formula is as follows:
[0054]
[0055] When x is equal to 1, the head entity of the triple is replaced; when x=0, the tail entity of the triple is replaced.
[0056] In the process of constructing the candidate entity set, all entities that appear on the same side of the relationship r as the entity to be replaced are found to form the candidate entity set, as shown below:
[0057] E1={e1,e2,...,e n}
[0058] In the process of constructing candidate negative triples, k entities are selected from the candidate entity set E1 to replace the entity to be replaced, forming a candidate negative triple set. When the head entity is replaced, the following is shown:
[0059] S1={(e i ,r,t)|e i ∈E1}
[0060] When the tail entity is replaced, it looks like this:
[0061] S1={(h,r,e1)|e i ∈E1}
[0062] During the screening of candidate negative triples, the score of each negative triple in the candidate negative triple set is calculated, and the one with the lowest score is selected as the candidate negative triple. The score calculation formula is as follows:
[0063]
[0064] Among them, L1 / L2 represents the 1-norm or 2-norm. The lower the score of the triplet, the more real and valid the triplet is. Otherwise, it means that the triplet is a negative triplet.
[0065] In the process of judging whether the candidate negative triple is in the training set, if the candidate negative triple is not in the training triple, return; otherwise, construct the candidate entity set again and repeat the execution.
[0066] Based on the acquired knowledge representation learning model, link prediction is performed to construct a candidate service set;
[0067] For each service in the candidate service set, we score and select them according to the four non-functional attributes of time, cost, energy consumption and error rate, and obtain the recommended industrial cloud robot manufacturing capability service plan that meets the functional and non-functional requirements.
[0068] Since the dimensions of the four non-functional attributes of time, cost, energy consumption, and error rate are different, each indicator needs to be normalized. The normalization formula is as follows:
[0069]
[0070] Among them, p i represents the i-th service in the candidate service set; It represents the jth attribute value of the i-th service in the candidate service set; max(p j ) represents the maximum value of the jth non-functional attribute; min(p j ) represents the minimum value of the jth non-functional attribute. Represents the normalized value of the jth non-functional attribute of the i-th service.
[0071] The service non-functional attributes are weighted and scored as follows:
[0072]
[0073] Where ω1, ω2, ω3, and ω4 are the weights of time, cost, energy consumption, and error rate given by the user. A score is calculated for each service in the candidate service set, and the top-scoring service is returned to the user as the recommended result.
[0074] The implementation of the industrial cloud robot manufacturing capability service recommendation method of the present invention has the following beneficial effects:
[0075] 1. The present invention constructs an industrial cloud robot manufacturing process information model that integrates the industrial cloud robot model, manufacturing task model, manufacturing capability service model and manufacturing process model; it provides a basis for the construction of the industrial cloud robot manufacturing capability service knowledge graph.
[0076] 2. The present invention constructs an industrial cloud robot manufacturing capability service knowledge graph that provides an industrial cloud robot manufacturing process information model; and effectively manages the industrial cloud robot manufacturing process knowledge.
[0077] 3. This invention proposes a service scoring optimization method that meets the personalized needs and preferences of users, focusing on objectives such as time, cost, energy consumption and error rate;
[0078] 4. This paper combines a knowledge representation learning model with link prediction to propose a method for generating recommendations for industrial cloud robot manufacturing capability services that meet both functional and non-functional requirements. It also improves the TransE algorithm to enhance the effectiveness of the recommendation results. BRIEF DESCRIPTION OF THE DRAWINGS
[0079] The present invention will be further described below with reference to the accompanying drawings and embodiments, in which:
[0080] Figure 1 A flowchart illustrating a method for implementing the industrial cloud robot manufacturing capability service recommendation of the present invention;
[0081] Figure 2 This is a schematic diagram of the industrial cloud robot model;
[0082] Figure 3 This is a schematic diagram of the manufacturing task model;
[0083] Figure 4 Schematic diagram of the manufacturing capability service model;
[0084] Figure 5 Schematic diagram of the recommended method for industrial cloud robot manufacturing capability service;
[0085] Figure 6 Schematic diagram of negative sample generation strategy;
[0086] Figure 7 This is a schematic diagram of the components of the idler wheel to be disassembled;
[0087] Figure 8 Primary recommendation diagram;
[0088] Figure 9 Rating preference. DETAILED DESCRIPTION
[0089] In order to have a clearer understanding of the technical features, purposes and effects of the present invention, specific embodiments of the present invention are now described in detail with reference to the accompanying drawings.
[0090] The industrial cloud robot manufacturing capability service recommendation method of the present invention is used to recommend industrial cloud robot manufacturing capability services that meet functional requirements and non-functional requirements, such as Figure 1 As shown, the recommended method mainly includes the following steps:
[0091] 1) Construct an industrial cloud robot manufacturing process information model that integrates the industrial cloud robot model, manufacturing task model, manufacturing capability service model, and manufacturing process model;
[0092] 2) Construct an industrial cloud robot manufacturing capability service knowledge graph that provides an information model of the industrial cloud robot manufacturing process;
[0093] 3) Propose a service scoring and optimization method that meets users' personalized needs and preferences, focusing on objectives such as time, cost, energy consumption, and error rate;
[0094] 4) Combined with the knowledge representation learning model for link prediction, a method for generating industrial cloud robot manufacturing capability service recommendation solutions that meet functional and non-functional requirements is proposed.
[0095] The above steps are steps from disassembly task to obtaining the optimal solution in the embodiment of the present invention.
[0096] Among them, the industrial cloud robot model in step 1) mainly describes the basic information and capability information of the industrial cloud robot. The basic information includes the static attributes of the industrial cloud robot such as ID, name, type, manufacturer, and geographical location; the capability information includes the technical parameters of the industrial cloud robot (such as maximum angular velocity, working environment temperature, maximum angular velocity) and the manufacturing capability service of the industrial cloud robot. The industrial cloud robot model established is as follows Figure 2 shown.
[0097] The manufacturing task model is a model of the manufacturing task, which mainly includes information such as task name, task type, task content, and task goal. Each manufacturing task is a collection of a series of subtasks. Figure 3 shown.
[0098] The industrial cloud robot manufacturing capability service model mainly includes three aspects: basic attributes, functional attributes, and non-functional attributes. Basic attributes describe the basic attributes of the service. They mainly include the name, ID, type, and service status of the service. Functional attributes describe the functions of the industrial cloud robot manufacturing capability service. Non-functional information mainly describes the service performance. Figure 4 shown.
[0099] The manufacturing process model mainly describes the manufacturing process information. The manufacturing process model includes the currently executed task status, task constraints, and manufacturing process data. It is shown below:
[0100] Task_Process={Task_status,Task_Conditions,Process_data}
[0101] In the process of constructing the industrial cloud robot manufacturing capability service knowledge graph given the industrial cloud robot manufacturing process information model in step 2), based on the industrial cloud robot model, manufacturing task model, manufacturing capability service model and manufacturing process model, combined with empirical data and expert knowledge, after entity extraction, relationship extraction and attribute extraction, entity disambiguation is used to construct the industrial cloud robot manufacturing capability service knowledge graph.
[0102] The knowledge graph G is represented as follows:
[0103] G=<E,R,S>
[0104] Where E={e1,e2,…,e |E|} is the entity set in the knowledge base, which contains |E| different entities; R = {r1, r2, …, r |R|} is the set of relations in the knowledge base, which contains |R| different relations; Represents the set of triples in the knowledge base.
[0105] In a unified entity and relational space, any triple in the industrial cloud robot manufacturing capability service knowledge graph can be described as:
[0106] S={(h,r,t)|h,t∈E,r∈R}
[0107] Among them, the vector h represents the head entity, t represents the tail entity, and r represents the relationship between the head entity and the tail entity.
[0108] In step 3), in the process of proposing a service scoring and optimization method that meets the user's personalized needs and preferences, the candidate services that meet the user's needs and preferences are scored and optimized based on the four non-functional attributes of time, cost, energy consumption and error rate, combined with user-defined weights.
[0109] Since the dimensions of these four non-functional attributes are different, each indicator needs to be normalized. The normalization formula is as follows:
[0110]
[0111] Among them, p i represents the i-th service in the candidate service set; It represents the jth attribute value of the i-th service in the candidate service set; max(p j ) represents the maximum value of the jth non-functional attribute; min(p j ) represents the minimum value of the jth non-functional attribute. Represents the normalized value of the jth non-functional attribute of the i-th service.
[0112] The weighted scores are as follows:
[0113]
[0114] Where ω1, ω2, ω3, and ω4 are the weights of time, cost, energy consumption, and error rate given by the user. The score of each service in the candidate service set is calculated according to the above formula, and the service with the highest score is returned to the user as the recommended result.
[0115] In step 4), the method for generating the recommended solution for industrial cloud robot manufacturing capability service is as follows: Figure 5 As shown, it mainly includes the following processes:
[0116] (1) Decompose the manufacturing task into its subtask sets.
[0117] (2) Model the industrial cloud robot manufacturing capability service knowledge, use knowledge representation learning to perform knowledge representation learning on the industrial cloud robot manufacturing capability service knowledge graph, and learn the embedding model of entities and relationships.
[0118] (3) Based on the acquired knowledge representation learning model, link prediction is performed to construct a candidate service set.
[0119] (4) Each service in the candidate service set is scored and optimized to obtain a recommended industrial cloud robot manufacturing capability service solution that meets functional and non-functional requirements.
[0120] When using the knowledge representation learning algorithm to perform knowledge representation learning on the industrial cloud robot manufacturing capability service knowledge graph, the negative sample generation strategy adopted by TransE is improved as follows: Figure 6 The specific process includes the following:
[0121] (1) According to the Bernoulli distribution, determine the entity to be replaced.
[0122] (2) Construct a candidate entity set.
[0123] (3) Construct candidate negative triples.
[0124] (4) Screen candidate negative triples.
[0125] (5) Determine whether the candidate negative triplet is in the training set and generate a negative triplet.
[0126] Example
[0127] A knowledge graph of industrial cloud robot manufacturing capability services is established, and the service recommendation request for the idler wheel disassembly task is used as an example for analysis and explanation.
[0128] According to step 1), an industrial cloud robot manufacturing process information model is constructed that integrates the industrial cloud robot model, manufacturing task model, manufacturing capability service model and manufacturing process model.
[0129] An industrial cloud robot manufacturing capability service knowledge graph is constructed to provide an information model of the industrial cloud robot manufacturing process. The industrial cloud robot manufacturing capability service knowledge graph contains 15,641 entities and 43 relationships.
[0130] The idler wheel to be disassembled is as follows Figure 7 shown.
[0131] The idler pulley disassembly task is to disassemble the idler pulley components (such as the center shaft, bearings, washers, nuts, and sleeves) and place them in the corresponding recycling area. The idler pulley disassembly task can be broken down into five subtasks based on its different components, as shown below:
[0132] task_disassemble={subtask0001,subtask0002,subtask0003,subtask0004,subtask0005}
[0133] Using the improved TransE algorithm, according to Figure 6 The process shown generates negative samples and performs knowledge representation learning on the industrial cloud robot manufacturing capability service knowledge graph.
[0134] Based on the learned knowledge representation learning model, link prediction is performed to construct a candidate service set, such as Figure 8 shown.
[0135] Based on the task requirements and related constraints, the service request information is input and the top three services are recommended for each subtask. The score of each recommended service is calculated based on the user-defined weights of 0.25, 0.3, 0.25, and 0.2.
[0136] The scoring results are as follows Figure 9 shown.
[0137] The final recommended service set to the user is:
[0138] resultService={Service_0022,Service_0070,Service_0201,Service_0211,Service_1072}
[0139] Through the service recommendation request example of a simple disassembly task, the feasibility and effectiveness of the proposed recommendation method for industrial cloud robot manufacturing capability service based on knowledge representation learning were verified.
[0140] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present invention, ordinary technicians in this field can also make many forms without departing from the scope of protection of the present invention and the claims, all of which are protected by the present invention.
Claims
1. A method for recommending industrial cloud robot manufacturing capability services, characterized in that: The following steps are involved: 1) Construct an industrial cloud robot manufacturing process information model that integrates the industrial cloud robot model, manufacturing task model, manufacturing capability service model, and manufacturing process model; 2) Combine the industrial cloud robot manufacturing process information model to build an industrial cloud robot manufacturing capability service knowledge graph; 3) Focusing on the optimization goals of time, cost, energy consumption and error rate, a service scoring optimization method is proposed to meet the personalized needs and preferences of users; 4) Combined with the knowledge representation learning model for link prediction, a method for generating industrial cloud robot manufacturing capability service recommendation solutions that meet both functional and non-functional requirements is proposed; In step 3), a service scoring optimization method that meets the personalized needs and preferences of users is proposed as follows: (1) Based on the four non-functional attributes of time, cost, energy consumption, and error rate, combined with user-defined weights, the candidate services that meet user needs and preferences are scored and selected, as described below: in, Time Refers to the time required for the service to perform a subtask, including task execution time and logistics time; Cost Refers to the cost of completing the task, including processing costs and logistics costs; Energy_cost Refers to the energy consumption to complete the task; Failure_rate Refers to the failure rate of the service to perform tasks; (2) Normalize each indicator. The normalization formula is as follows: in, represents the i-th service in the candidate service set; It represents the jth attribute value of the i-th service in the candidate service set; represents the maximum value of the jth non-functional attribute; It represents the minimum value of the jth non-functional attribute; represents the normalized value of the jth non-functional attribute of the i-th service; (3) Calculate the score of each service in the candidate service set, and return the service with the highest score to the user as the recommended result. The score calculation formula is as follows: in, , , and It is the weight of time, cost, energy consumption and error rate given by the user; In step 4), the method for generating the recommended solution for the industrial cloud robot manufacturing capability service is as follows: (1) Decompose the manufacturing task into its subtask sets; (2) Use the knowledge representation learning algorithm to perform knowledge representation learning on the knowledge graph and acquire the knowledge representation learning model of the industrial cloud robot manufacturing capability service knowledge graph; (3) Based on the acquired knowledge representation learning model, link prediction is performed to construct a candidate service set; (4) Score and select each service in the candidate service set to obtain a recommended industrial cloud robot manufacturing capability service solution that meets both functional and non-functional requirements.
2. The industrial cloud robot manufacturing capability service recommendation method according to claim 1, characterized in that: In the step 1), The industrial cloud robot model describes the basic information and capability information of the industrial cloud robot. The basic information includes the ID, name, type, and geographic location of the industrial cloud robot; the capability information includes the technical parameters of the industrial cloud robot and the manufacturing capability services provided; The manufacturing task model includes task name, task type, task content, and task goal; The manufacturing capability service model includes basic attributes, functional attributes and non-functional attributes. The basic attributes include service ID, service name and service type. The functional attributes include tasks or subtasks that can be completed by the service. The non-functional attributes include the time and energy consumption of the service. The manufacturing process model includes the task status being currently executed, task constraints and manufacturing process data.
3. The industrial cloud robot manufacturing capability service recommendation method according to claim 1, characterized in that: In the above step 2), the method of constructing the industrial cloud robot manufacturing capability service knowledge graph is as follows: (1) Based on the industrial cloud robot model, manufacturing task model, manufacturing capability service model and manufacturing process model, combined with empirical data and expert knowledge, through entity extraction, relationship extraction and attribute extraction, entity disambiguation is used to construct the industrial cloud robot manufacturing capability service knowledge graph; (2) The knowledge graph G is represented as follows: in, is the entity set in the knowledge base, including different entities; is the set of relations in the knowledge base, including different kinds of relationships; Represents the set of triples in the knowledge base; (3) In the unified entity and relational space, any triple in the industrial cloud robot manufacturing capability service knowledge graph is described as: Among them, the vector Represents the head entity, Represents the tail entity, Represents the relationship between the head entity and the tail entity.
Citation Information
Patent Citations
Knowledge graph construction method for man-machine cooperation dismounting task
CN111191047A
Intelligent heat supply service recommendation method based on knowledge graph
CN111274413A
