CN115828756A - Phased array antenna excitation inversion method based on design and fault knowledge graph - Google Patents
Phased array antenna excitation inversion method based on design and fault knowledge graph Download PDFInfo
- Publication number
- CN115828756A CN115828756A CN202211583008.6A CN202211583008A CN115828756A CN 115828756 A CN115828756 A CN 115828756A CN 202211583008 A CN202211583008 A CN 202211583008A CN 115828756 A CN115828756 A CN 115828756A
- Authority
- CN
- China
- Prior art keywords
- entity
- array element
- fault
- design
- knowledge
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Landscapes
- Variable-Direction Aerials And Aerial Arrays (AREA)
Abstract
The invention discloses a phased array antenna excitation inversion method based on a design and fault knowledge graph, which comprises the following implementation steps: collecting unstructured text data of the phased array antenna in the design and operation and maintenance stages; constructing an ontology of a design and fault knowledge graph; generating a training set and a testing set to obtain a qualified entity extraction model; inputting all text data into a qualified entity extraction model to obtain knowledge triples of a design and fault knowledge graph; constructing a design and fault knowledge graph, and fusing the design and fault knowledge graph with array elements to excite inversion; and inverting the excitation of the array elements of the phased array antenna by using the knowledge graph. The invention realizes the association of phased array antenna design and operation and maintenance phase data, realizes the design of the phased array antenna and the inversion of information in the fault knowledge graph, and increases the reference of the knowledge graph for solving the fault.
Description
Technical Field
The invention belongs to the technical field of physics, and further relates to a phased array antenna excitation inversion method based on a design and fault knowledge map in the technical field of data processing. The invention can be used for phased array antenna excitation inversion requiring fault and design information.
Background
In the process from the design to the operation and maintenance of the phased array antenna, various design information and fault information are scattered and are displayed in a weakening manner, basically, the record is carried out in a text mode, related personnel can only know the content of the design and the operation and maintenance by searching various data information, and the texts cannot be directly related. The phased array antenna fault information analysis relates to participation of operation and maintenance personnel and designers, is a complex dynamic process, and is used for analyzing the phased array antenna fault by using the knowledge graph in order to guarantee normal operation of an equipment system and improve the fault analysis efficiency, so that the relevance between the fault and the design information can be increased, and the designers can more efficiently inquire the fault information, the design information and recommend a solution through the knowledge graph. In the existing fault analysis method based on the knowledge graph, due to the lack of information association or the imperfect recommended solution, the design and fault knowledge graph which has an inversion function and associates the design information and the fault information cannot be obtained.
The university in Zhongshan has proposed a fault analysis method in the patent document "cloud native system fault analysis method based on knowledge graph" (application number 202011554734.6 application publication number CN 112540832A). The method comprises the following implementation steps: the method comprises the steps of firstly, acquiring original data, and constructing a knowledge graph based on the original data to obtain graph data; secondly, carrying out anomaly detection on the graph data through an anomaly detection model to obtain an anomaly node; and thirdly, calculating the similarity of the abnormal node and the replica node corresponding to the abnormal node, and positioning the fault root cause based on the similarity. The method has the defects that the ontology is not constructed before the knowledge graph is constructed, and the method is only limited to the construction of the general knowledge graph with wider knowledge coverage range. Knowledge and data generated in the design, operation and maintenance processes of the phased array antenna are usually specific professional field knowledge, and unstructured data usually lack correlation and need to be subjected to ontology construction first. The method does not specify entity types and relationship types for entity extraction and relationship extraction, and is difficult to apply to the field of phased array antennas.
The Guangzhou Congregation national communication science and technology company discloses a fault handling recommendation method in a patent document applied by Guangzhou Congregation national communication science and technology company, namely a power grid fault handling plan recommendation method and device based on a knowledge graph (application No. 202210328887.1 application publication No. CN 114756686A). The method comprises the following implementation steps: firstly, information is obtained; step two, the fault information is arranged, the format of the fault information is converted, and fault keywords of the fault information are extracted; thirdly, extracting a failure disposal plan in the database according to the failure keyword; and fourthly, matching the extracted fault handling plan, and prompting if the matching fails. The method for recommending fault handling has the disadvantages that the recommended fault handling scheme in the knowledge graph only can take the past fault handling scheme as a reference, can not be inverted according to information to obtain array element excitation for specifically solving the fault, and is not suitable for the field of phased array antennas which need to give out specific array element excitation according to different design information.
Disclosure of Invention
The invention aims to provide a phased array antenna excitation inversion method based on a design and fault knowledge graph aiming at the defects of the prior art, which is used for solving the problems that the design information unstructured data and the fault information unstructured data lack correlation, the conventional fault solution can be used as a reference for recommending the solution, and the array element excitation for specifically solving the fault cannot be obtained according to the specific design information and the fault information.
The method comprises the steps of collecting unstructured text data of a phased array antenna in operation, maintenance and design stages, constructing an ontology of the unstructured text data, enabling the relationship between entities to be more specific through the construction of the ontology, enabling extracted entities to form correlation directly, generating a training set and a testing set to train, obtaining an entity extraction model, extracting the entities in the unstructured text data through the trained entity extraction model, constructing a design and fault knowledge graph, fusing the design and fault knowledge graph into array element excitation inversion, conducting inversion on phased array antenna array element excitation through the knowledge graph according to fault information, and recommending the inverted array element excitation to design information to solve faults.
In order to achieve the purpose, the technical scheme of the invention comprises the following steps:
step 1, collecting unstructured text data generated by a phased array antenna in a design and operation and maintenance stage, wherein the unstructured text data comprises design documents for diagnosis and maintenance records;
step 2, constructing an ontology of a design and fault knowledge graph:
respectively setting 17 entity types and 4 relation types according to unstructured text data generated by a phased array antenna in the design and operation and maintenance stages, constructing the 17 entity types and the 4 relation types into 16 'head entity-relation-tail entity' knowledge triples, and forming a body of a design and fault knowledge map by all the entity types, the relation types and the knowledge triples;
step 3, generating a training set and a test set:
step 3.1, randomly selecting 20% of text data from the unstructured text data, and labeling an entity type label of each selected text data;
step 3.2, generating a training set and a test set according to the marked text data in the proportion of 7;
step 4, obtaining a qualified entity extraction model:
step 4.1, inputting the training set into an entity extraction model based on a bidirectional long-short term memory network, and iteratively updating parameters in the entity extraction model by using a random gradient descent method until a loss function is converged to obtain a trained entity extraction model;
step 4.2, inputting the test set into the trained entity extraction model, outputting entity type labels predicted by each text data in the test set, calculating the accuracy of the entity type labels of all predicted text data output by the entity extraction model, and obtaining a qualified entity extraction model when the extraction accuracy of the extraction model reaches 80%;
step 5, constructing a design and fault knowledge map:
step 5.1, inputting all text data in the unstructured text data into a qualified entity extraction model, and outputting an entity of each text data and a corresponding entity type;
step 5.2, associating the entities in each text data according to the entity types and the knowledge triples in the ontology to obtain the design of the text data and the knowledge triples of the failure knowledge map;
step 5.3, storing all the knowledge triples into a Neo4j graph database, taking an entity of each knowledge triplet as a corresponding node through a generation instruction of the Neo4j graph database, and taking a relation as an edge of a connector entity node and a tail entity node to obtain a design and fault knowledge graph;
step 6, designing and fault knowledge map fusion array element excitation inversion:
step 6.1, extracting entity types from the fault and design knowledge graph as follows: all entities of array element number, array element coordinates, array element excitation, channel fault description and side lobe level allowable error form failed array element excitation by the channel fault description and the array element excitation; forming a group of input parameters of array element excitation inversion by a group of associated array element excitation, channel fault description, side lobe level allowable errors and failed array element excitation;
6.2, selecting a group of input parameters of the unselected array element excitation inversion; substituting array element excitation in the input parameters of the selected array element excitation inversion into a directional diagram function of the antenna array to obtain data of a radiation directional diagram, extracting a side lobe level of each of the left side and the right side of a main lobe of an azimuth plane and a pitch plane from the data, and forming all the extracted data into failure forward directional diagram parameters;
step 6.3, taking the side lobe level and the main lobe beam direction in the parameters of the failure forward directional diagram as targets, substituting the failure array element excitation in the input parameters of the same group of array element excitation inversion into an evolutionary algorithm, iteratively updating the currently iterated array element excitation until an iteration termination condition is met, and taking the output currently iterated array element excitation as the inverted array element excitation;
6.4, correlating the inverted array element excitation with the channel fault description in the input parameters of the same group of array element excitation inversion; taking the 'channel fault description' in each associated inverted array element excitation and channel fault description as a head entity in a 'head entity-relation-tail entity' knowledge triplet and taking the 'inverted array element excitation' as a tail entity to obtain a 'channel fault description-reasoning-inverted array element excitation' triplet, and importing the triplet into a Neo4j database for storage;
6.5, judging whether input parameters of all array element excitation inversion are selected, if so, executing the step 7, otherwise, executing the step 6.2;
and 7, inverting the phased array antenna array element excitation by using a knowledge graph:
step 7.1, collecting text data of the phased array antenna when the phased array antenna fails in the operation and maintenance stage, obtaining entities of the failed text data by adopting the same method as the steps 5.1 and 5.2, searching the relation in the knowledge triple corresponding to the entity type in the body according to the entity type of each entity, associating the entities of the failed data to obtain the knowledge triple of the failed text data, updating the failure information in the design and failure knowledge maps by using the knowledge triple, and obtaining the design and failure knowledge maps after the failure information is updated;
step 7.2, according to the phased array antenna name entity of the fault data, extracting the entity type corresponding to the phased array antenna name entity in the updated design and fault knowledge map as follows: entity of array element excitation, channel fault description, side lobe level allowable error sum; forming a group of input parameters of array element excitation inversion by using the associated array element excitation, channel fault description, side lobe level allowable error and failed array element excitation;
7.3, obtaining inverted array element excitation by adopting the same method as the steps 6.2 and 6.3, and correlating the array element excitation with the extracted channel fault description; taking 'channel fault description' in the associated inverted array element excitation and channel fault description as a head entity in a 'head entity-relation-tail entity' knowledge triple and taking 'inverted array element excitation' as a tail entity to obtain a 'channel fault description-reasoning-inverted array element excitation' triple, importing the triple into a Neo4j database to update related data, and obtaining a design and a fault knowledge map after updating the inverted array element excitation;
and 7.4, searching inverted array element excitation in the updated inverted array element excited design and fault knowledge map by a designer, and modifying phased array antenna array element excitation according to the inverted array element excitation to obtain array design information of the phased array antenna after the fault.
Compared with the prior art, the invention has the following advantages:
firstly, the invention constructs a design and fault knowledge map body of a phased array antenna design and operation and maintenance stage, and overcomes the defect that data is not associated due to the lack of association between unstructured design information and unstructured fault information of the phased array antenna design and operation and maintenance stage when unstructured text data is extracted in the prior art. The method can completely obtain valuable information in the unstructured data, realize the correlation of a large amount of data in the design and operation and maintenance stages of the phased array antenna, enable the operation and maintenance personnel and the designers to inquire the design information in the design stage and the fault information in the operation and maintenance stages, and improve the efficiency of solving the faults of the phased array antenna.
Secondly, the method and the device invert the information in the design and fault knowledge maps of the phased array antenna in the design and operation and maintenance stages and add the array element excitation generated by inversion into the knowledge maps, so that the defect that the fault recommended to be solved in the prior art can only use the past fault measures as references is overcome, the method and the device can invert according to different information to obtain the array element excitation for specifically solving the fault, and the reference of the knowledge maps for solving the fault by designers is increased.
Drawings
FIG. 1 is a flow chart of the present invention;
FIG. 2 is a flow chart of excitation inversion of a design and fault knowledge base fusion array element in the invention.
Detailed Description
The invention is further described below with reference to the figures and examples.
Implementation steps of embodiments of the present invention are further described with reference to fig. 1.
Step 1, collecting unstructured text data generated by a phased array antenna in the design and operation and maintenance stages, wherein the unstructured text data comprises design documents for diagnosis and maintenance records.
500 unstructured text data generated in the phased array antenna design and operation and maintenance phase of a company are used as a source.
And 2, constructing a body of the design and fault knowledge map according to unstructured text data generated by the phased array antenna in the design and operation and maintenance stages.
And setting entity types, relationship types and a knowledge triple of 'head entity-relationship-tail entity'.
The ontology for constructing the text data in the embodiment of the invention comprises 16 entity types and 4 relation types.
The 17 entity types are respectively: phased array antenna name, radio frequency network, power distribution, power module, array unit, T \ R assembly, SMA connecting piece, array design information, array element number, array element coordinate, array element excitation, side lobe level allowable error, working wavelength, channel fault information, channel fault description, fault rate and fault cause.
The 4 relationship types are: build, information, contain, and infer.
The 16 "head entity-relationship-tail entity" knowledge triples are constructed by 17 entity types and 4 relationship types, respectively, as shown in table 1, where each row in table 1 represents a knowledge triplet.
The construction of 17 entity types and 4 relationship types in the ontology into 16 knowledge triples of "head entity-relationship-tail entity" refers to: the method comprises the following steps of combining 17 entity types, construction and information of phased array antenna name, radio frequency network, power distribution, power module, array unit, T \ R component, array design information, array element number, array element coordinate, array element excitation, side lobe level allowable error, working wavelength, channel fault information, channel fault description, fault rate and fault cause, the three relation types of "including" and "reasoning" are constructed as "radio frequency network-phased array antenna name-construction", "power distribution-phased array antenna name-construction", "power module-phased array antenna name-construction", "array unit-phased array antenna name-construction", "T \ R component-phased array antenna name-construction", "SMA connector-phased array antenna name-construction", "array design information-phased array antenna name-information", "array design information-array element number-inclusion", "array design information-array element coordinate-inclusion", "array design information-array element excitation-inclusion", "array design information-side lobe level allowed error-inclusion", "array design information-operating wavelength-inclusion", "channel fault information-inclusion-channel fault description", "channel fault description-inclusion-fault probability", "channel fault description-inclusion-fault cause", "fault cause-reasoning-module name" 16 "head entity-relation-tail entity" knowledge triplets.
Table 1 ontology triple list
And 3, generating a training set and a testing set.
And randomly selecting 100 pieces of text data from the unstructured text data, carrying out entity labeling on the training set and the test set according to the entity types in the body by using a BMEO labeling method, labeling the entity type labels of the text data, and obtaining the labeled text data. For example, if there is a word of type "channel failure description" in some text data, "B-channel failure description" indicates the first character of the word, "M-channel failure description" indicates all the middle characters except the beginning and the end of the word, "E-failure mode" indicates the end character of the word, and the label "O" indicates that the character is not in an entity. 70 pieces of text data in the labeled text data form a training set, and the remaining 30 pieces of text data form a testing set.
And 4, acquiring a qualified entity extraction model.
And 4.1, inputting the training set into an entity extraction model based on the bidirectional long-short term memory network, and iteratively updating parameters in the model by using a random gradient descent method until a loss function is converged to obtain the trained entity extraction model.
The network loss function is as follows:
wherein MSE represents a loss value function between a label predicted by the model and a label labeled by the text data entity, n represents the number of pieces of text data, y represents i A label representing the label of the ith entity in the training set,
representing training set model output predictionThe ith label measured.
The entity extraction model based on the two-way long and short term memory network is a model constructed in the prior art, and the structure and the initial parameter setting of the model are described in the 'knowledge map construction technology and application for aircraft power supply system fault diagnosis' (aeronautical report, 2021, 42) published by Nie Tongsheng et al.
And 4.2, inputting the test set into the trained entity extraction model, outputting the entity type label predicted by each text data in the test set, calculating the accuracy of the entity type labels of all predicted text data output by the entity extraction model, and obtaining a qualified entity extraction model when the extraction accuracy of the extraction model reaches 80%.
The accuracy is calculated as follows:
and F is the number of labels with inconsistent label output by the entity extraction model and label of the test set.
And 5, constructing a design and fault knowledge map.
And 5.1, inputting all text data in the unstructured text data into a qualified entity extraction model, and outputting an entity of each text data and a corresponding entity type.
And 5.2, associating the entities in each text data according to the entity types and the knowledge triples in the ontology to obtain the design of the text data and the knowledge triples of the fault knowledge graph.
For example, from the text data "cause channel 1 failure is caused by SMA connection failure. The entities extracted from the ' are ' channel 1 fault ' and ' SMA connecting piece ', wherein the entity type of the ' channel 1 fault ' is a channel fault description, and the entity type of the ' SMA connecting piece ' is a module. And constructing a channel 1 fault-fault cause-SMA connecting piece triple according to the channel fault description-fault cause-fault position triple in the body.
And 5.3, storing all the knowledge triples into a Neo4j graph database, taking the entity of each knowledge triplet as a corresponding node through a generation instruction of the Neo4j graph database, and taking the relation as one edge of a connector entity node and a tail entity node to obtain a design and fault knowledge graph.
The Neo4j is a high performance open source non-relational database developed by Neo4j corporation.
And 6, designing and fault knowledge map fusion array element excitation inversion.
Step 6.1, extracting entity types from the fault and design knowledge graph as follows: all entities of array element number, array element coordinates, array element excitation, channel fault description and side lobe level allowable error form failed array element excitation by the channel fault description and the array element excitation; and forming a group of input parameters of array element excitation inversion by using a group of associated array element excitation, channel fault description, side lobe level allowable error and failed array element excitation.
6.2, selecting a group of input parameters of the unselected array element excitation inversion; and substituting the array element excitation into a directional diagram function of the antenna array to obtain data of a radiation directional diagram, extracting the left and right first zero positions and a maximum side lobe level of a main lobe of an azimuth plane from the data, simultaneously extracting the left and right first zero positions and a maximum side lobe level of a main lobe of a pitching plane, and forming all the extracted data into parameters of a failure front diagram.
The directional diagram function of the antenna array is as follows:
wherein
As a function of the direction of the field strength of the array elements, theta and
respectively a pitch angle and an azimuth angle, N is the number of array elements, I n Is the amplitude of the array element excitation of the nth array element, e is a natural constant, j is an imaginary number unit, pi is a circumferential rate, lambda is a working wavelength, and x n X-coordinate, y, representing the coordinates of the nth array element n Y-coordinate representing the coordinates of the nth array element, b n The phase of the array element excitation of the nth array element.
And 6.3, setting the population size NP =200 of the genetic algorithm, the iteration number G =200, the cross probability Pc =0.8 and the variation probability Pm =0.05, wherein the length of the chromosome in the population is determined by the number of the array elements. And 4, setting the failed array element excitation obtained in the step 7.1 as the 1 st chromosome in the initial population, and randomly generating other chromosomes.
And 6.4, generating a new population of the current iteration through a roulette selection method, probability intersection with a probability value Pc and probability variation with a probability value Pm. Substituting chromosomes into a directional diagram function of an antenna array to obtain data of a radiation directional diagram, extracting the left and right first zero positions and a maximum side lobe level of an azimuth plane main lobe from the data, simultaneously extracting the left and right first zero positions and a maximum side lobe level of a pitching plane main lobe, forming all the extracted data into parameters of a chromosome corresponding directional diagram, calculating the fitness of the chromosomes, and always keeping the chromosome with the lowest fitness in a population into the 1 st chromosome in the population.
The fitness function is as follows:
fitness＝1/(|FSLL_AZ-FSLL_AZ_set|+|FSLL_EL-FSLL_EL_set|+0.9|NULL_AZ_1-NULL_AZ_set1|+0.9|NULL_AZ_2-NULL_AZ_set2|+0.9|NULL_EL_2-NULL_EL_set2|+0.9|NULL_EL_2-NULL_EL_set2|)
wherein, the fitness of the chromosome is fitness, FSLL _ AZ is the azimuth plane maximum side lobe level in the directional diagram parameter corresponding to the chromosome, FSLL _ AZ _ set is the azimuth plane maximum side lobe level in the failure forward diagram parameter, FSLL _ EL is the pitch plane maximum side lobe level in the directional diagram parameter corresponding to the chromosome, FSLL _ EL _ set is the pitch plane maximum side lobe level in the failure forward diagram parameter, NULL _ AZ _1 is the first zero position at the left of the azimuth plane main lobe in the directional diagram parameter corresponding to the chromosome, and NULL _ AZ _1 is the first zero position at the left of the azimuth plane main lobe in the failure forward diagram parameter, NULL _ AZ _2 is the first right zero position of the azimuth main lobe in the chromosome corresponding directional diagram parameter, NULL _ AZ _ set2 is the first right zero position of the azimuth main lobe in the failure forward diagram parameter, NULL _ EL _1 is the first left zero position of the pitch main lobe in the chromosome corresponding directional diagram parameter, NULL _ EL _ set1 is the first left zero position of the pitch main lobe in the failure forward diagram parameter, NULL _ EL _2 is the first right zero position of the pitch main lobe in the chromosome corresponding directional diagram parameter, and NULL _ EL _ set2 is the first right zero position of the pitch main lobe in the failure forward diagram parameter.
And (5) repeatedly executing the step 6.4 until an iteration termination condition is met, and extracting a first chromosome from the population at which iteration is terminated as the excitation of the inverted array element.
The iteration termination condition is a case where the following two conditions are satisfied simultaneously.
Condition 1: the error between the pitch surface side lobe level in the chromosome corresponding directional diagram parameter with the lowest fitness and the pitch surface side lobe level in the failure front directional diagram parameter is within the side lobe level allowable error.
Condition 2: the chromosome with the lowest fitness corresponds to the error between the azimuth plane side lobe level in the directional diagram parameter and the azimuth plane side lobe level in the failure front directional diagram parameter, and the error is within the side lobe level allowable error.
6.5, correlating the inverted array element excitation with the channel fault description in the input parameters of the same group of array element excitation inversion; taking the 'channel fault description' in each associated inverted array element excitation and fault description as a head entity in a 'head entity-relation-tail entity' knowledge triple and taking the 'inverted array element excitation' as a tail entity to obtain a 'fault description-reasoning-inverted array element excitation' triple, and importing the triple into a Neo4j database for storage, wherein the flow is shown in fig. 2.
6.6, judging whether input parameters of all array element excitation inversion are selected, if so, executing the step 7, otherwise, executing the step 6.2; and obtaining 500 inverted array element excitations from input parameters of all the array element excitation inversion.
And 7, inverting the phased array antenna array element excitation by using the knowledge graph.
And 7.1, collecting text data of the phased array antenna when the phased array antenna fails in the operation and maintenance stage, obtaining entities of the failed text data by adopting the same method as the steps 5.1 and 5.2, searching the relation in the knowledge triple corresponding to the entity type in the body according to the entity type of each entity, associating the entities of the failed data to obtain the knowledge triple of the failed text data, and updating the failure information in the design and failure knowledge maps by using the knowledge triple to obtain the design and failure knowledge maps after the failure information is updated.
Step 7.2, according to the phased array antenna name entity of the fault data, extracting the entity type corresponding to the phased array antenna name entity in the updated design and fault knowledge map as follows: entity of array element excitation, channel fault description, side lobe level allowable error; and forming a group of input parameters of array element excitation inversion by using the associated array element excitation, channel fault description, side lobe level allowable error and failed array element excitation.
7.3, obtaining inverted array element excitation by adopting the same method as the steps 6.2 and 6.3, and correlating the array element excitation with the extracted channel fault description; taking the 'channel fault description' in the associated inverted array element excitation and channel fault description as a head entity in a 'head entity-relation-tail entity' knowledge triple, taking the 'inverted array element excitation' as a tail entity to obtain the 'channel fault description-reasoning-inverted array element excitation' triple, importing the triple into a Neo4j database to update related data, and obtaining the design and fault knowledge map after the updated inverted array element excitation.
And 7.4, searching inverted array element excitation in the updated inverted array element excited design and fault knowledge map by a designer, and modifying phased array antenna array element excitation according to the inverted array element excitation to obtain array design information of the phased array antenna after the fault.
Claims (5)
1. A phased array antenna excitation inversion method based on a design and fault knowledge graph is characterized in that a body of the design and fault knowledge graph is constructed, the design and fault knowledge graph is fused with an array element excitation inversion, and the knowledge graph is used for inverting the phased array antenna array element excitation; the inversion method comprises the following steps:
step 1, collecting unstructured text data generated by a phased array antenna in a design and operation and maintenance stage, wherein the unstructured text data comprises design documents for diagnosis and maintenance records;
step 2, constructing an ontology of a design and fault knowledge map:
respectively setting 17 entity types and 4 relation types according to unstructured text data generated by a phased array antenna in the design and operation and maintenance stages, constructing the 17 entity types and the 4 relation types into 16 knowledge triples of head entity-relation-tail entity, and forming a body of a design and fault knowledge map by all the entity types, the relation types and the knowledge triples;
step 3, generating a training set and a testing set:
step 3.1, randomly selecting 20% of text data from the unstructured text data, and labeling an entity type label of each selected text data;
step 3.2, generating a training set and a test set according to the marked text data in the proportion of 7;
step 4, obtaining a qualified entity extraction model:
step 4.1, inputting the training set into an entity extraction model based on a bidirectional long-short term memory network, and iteratively updating parameters in the entity extraction model by using a random gradient descent method until a loss function is converged to obtain a trained entity extraction model;
step 4.2, inputting the test set into the trained entity extraction model, outputting entity type labels predicted by each text data in the test set, calculating the accuracy of the entity type labels of all predicted text data output by the entity extraction model, and obtaining a qualified entity extraction model when the extraction accuracy of the extraction model reaches 80%;
step 5, constructing a design and fault knowledge map:
step 5.1, inputting all text data in the unstructured text data into a qualified entity extraction model, and outputting an entity of each text data and a corresponding entity type;
step 5.2, associating the entities in each text data according to the entity types and the knowledge triples in the ontology to obtain the design of the text data and the knowledge triples of the failure knowledge map;
step 5.3, storing all knowledge triples into a Neo4j graph database, taking an entity of each knowledge triplet as a corresponding node through a generation instruction of the Neo4j graph database, and taking a relation as an edge of a connector entity node and a tail entity node to obtain a design and fault knowledge graph;
step 6, designing and fault knowledge map fusion array element excitation inversion:
step 6.1, extracting entity types from the fault and design knowledge graph as follows: all entities of array element number, array element coordinates, array element excitation, channel fault description and side lobe level allowable errors form failed array element excitation by the channel fault description and the array element excitation; forming a group of input parameters of array element excitation inversion by a group of associated array element excitation, channel fault description, side lobe level allowable errors and failed array element excitation;
6.2, selecting a group of input parameters of the unselected array element excitation inversion; substituting array element excitation in the input parameters of the selected array element excitation inversion into a directional diagram function of the antenna array to obtain data of a radiation directional diagram, extracting a side lobe level of each of the left side and the right side of a main lobe of an azimuth plane and a pitch plane from the data, and forming all the extracted data into failure forward directional diagram parameters;
step 6.3, taking the side lobe level and the main lobe beam direction in the parameters of the failure forward directional diagram as targets, substituting the failure array element excitation in the input parameters of the same group of array element excitation inversion into an evolutionary algorithm, iteratively updating the currently iterated array element excitation until an iteration termination condition is met, and taking the output currently iterated array element excitation as the inverted array element excitation;
6.4, correlating the inverted array element excitation with the channel fault description in the input parameters of the same group of array element excitation inversion; taking the 'channel fault description' in each associated inverted array element excitation and channel fault description as a head entity in a 'head entity-relation-tail entity' knowledge triple, taking the 'inverted array element excitation' as a tail entity to obtain a 'channel fault description-reasoning-inverted array element excitation' triple, and importing the triple into a Neo4j database for storage;
step 6.5, judging whether input parameters of all array element excitation inversion are selected, if so, executing step 7, otherwise, executing step 6.2;
and 7, inverting the phased array antenna array element excitation by using a knowledge graph:
step 7.1, collecting text data of the phased array antenna when the phased array antenna fails in the operation and maintenance stage, obtaining entities of the failed text data by adopting the same method as the steps 5.1 and 5.2, searching the relation in the knowledge triple corresponding to the entity type in the body according to the entity type of each entity, associating the entities of the failed data to obtain the knowledge triple of the failed text data, updating the failure information in the design and failure knowledge maps by using the knowledge triple, and obtaining the design and failure knowledge maps after the failure information is updated;
step 7.2, according to the phased array antenna name entity of the fault data, extracting the entity type corresponding to the phased array antenna name entity in the updated design and fault knowledge map as follows: entity of array element excitation, channel fault description, side lobe level allowable error sum; forming a group of input parameters of array element excitation inversion by using the associated array element excitation, channel fault description, side lobe level allowable error and failed array element excitation;
7.3, obtaining inverted array element excitation by adopting the same method as the steps 6.2 and 6.3, and correlating the array element excitation with the extracted channel fault description; taking 'channel fault description' in the associated inversion array element excitation and channel fault description as a head entity in a 'head entity-relation-tail entity' knowledge triple, taking 'inversion array element excitation' as a tail entity to obtain a 'channel fault description-reasoning-inversion array element excitation' triple, importing the triple into a Neo4j database to update related data, and obtaining a design and a fault knowledge map after updating the inversion array element excitation;
and 7.4, searching inverted array element excitation in the updated inverted array element excited design and fault knowledge map by a designer, and modifying phased array antenna array element excitation according to the inverted array element excitation to obtain array design information of the phased array antenna after the fault.
2. The design and fault knowledge-graph-based phased array antenna excitation inversion method of claim 1, wherein the constructing 17 entity types and 4 relationship types in the ontology into 16 "head entity-relationship-tail entity" knowledge triples in step 2 is characterized by: the method comprises the following steps of combining 17 entity types, construction and information of phased array antenna names, radio frequency networks, power distribution, power modules, array units, T \ R components, array design information, array element number, array element coordinates, array element excitation, side lobe level allowable errors, working wavelengths, channel fault information, channel fault description, fault rate and fault causes, the three relation types of "including" and "reasoning" are constructed as "radio frequency network-phased array antenna name-construction", "power distribution-phased array antenna name-construction", "power module-phased array antenna name-construction", "array unit-phased array antenna name-construction", "T \ R component-phased array antenna name-construction", "SMA connector-phased array antenna name-construction", "array design information-phased array antenna name-information", "array design information-array element number-inclusion", "array design information-array element coordinate-inclusion", "array design information-array element excitation-inclusion", "array design information-side lobe level allowed error-inclusion", "array design information-operating wavelength-inclusion", "channel fault information-inclusion-channel fault description", "channel fault description-inclusion-fault probability", "channel fault description-inclusion-fault cause", "fault cause-reasoning-module name" 16 "head entity-relation-tail entity" knowledge triplets.
3. The design and fault knowledge-graph based phased array antenna excitation inversion method of claim 1, characterized in that the loss function in step 4.1 is as follows:
wherein MSE represents a loss value function between a label predicted by the model and a label marked by the text data entity, n represents the number of pieces of text data, y represents the number of the text data i A label representing the label of the ith entity in the training set,
the ith label representing the prediction of the training set model output.
4. The design and fault knowledge-graph based phased array antenna excitation inversion method according to claim 1, characterized in that the accuracy in step 4.2 is calculated as follows:
and F is the number of the labels which are output by the entity extraction model and are inconsistent with the label which is output by the entity extraction model and the label which is output by the test set.
5. The design and fault knowledge-graph-based phased array antenna excitation inversion method of claim 1, wherein the iteration termination condition in step 7.3 is a case where the following two conditions are satisfied simultaneously:
the method comprises the following steps that 1, an error between an azimuth plane side lobe level in an inverted array element excitation corresponding directional diagram parameter and an azimuth plane side lobe level in a failure forward directional diagram parameter is within a side lobe level allowable error;
and 2, under the condition that the error between the pitch surface side lobe level in the inverted array element excitation corresponding directional diagram parameter and the pitch surface side lobe level in the failure forward directional diagram parameter is within the side lobe level allowable error.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211583008.6A CN115828756A (en) | 2022-12-09 | 2022-12-09 | Phased array antenna excitation inversion method based on design and fault knowledge graph |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211583008.6A CN115828756A (en) | 2022-12-09 | 2022-12-09 | Phased array antenna excitation inversion method based on design and fault knowledge graph |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115828756A true CN115828756A (en) | 2023-03-21 |
Family
ID=85546171
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211583008.6A Pending CN115828756A (en) | 2022-12-09 | 2022-12-09 | Phased array antenna excitation inversion method based on design and fault knowledge graph |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115828756A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117131445A (en) * | 2023-07-28 | 2023-11-28 | 深圳市财富趋势科技股份有限公司 | Abnormal transaction detection method and system |
-
2022
- 2022-12-09 CN CN202211583008.6A patent/CN115828756A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117131445A (en) * | 2023-07-28 | 2023-11-28 | 深圳市财富趋势科技股份有限公司 | Abnormal transaction detection method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Hong et al. | Energy forecasting: A review and outlook | |
| CN106933728B (en) | Method and device for measuring association between requirement and use case | |
| US20210247246A1 (en) | Edge-cloud collaboration platform for intelligent coking monitoring of cracking furnace tubes and working method thereof | |
| Huang et al. | Technology life cycle analysis: From the dynamic perspective of patent citation networks | |
| CN115828756A (en) | Phased array antenna excitation inversion method based on design and fault knowledge graph | |
| Yu et al. | Parameter identification of photovoltaic models using a sine cosine differential gradient based optimizer | |
| Pascoe et al. | Designing and documenting experiments in CMIP6 | |
| CN108520310A (en) | Wind speed forecasting method based on G-L mixed noise characteristic v- support vector regressions | |
| CN111898316A (en) | Construction method and application of super-surface structure design model | |
| Ponti et al. | A Wasserstein distance based multiobjective evolutionary algorithm for the risk aware optimization of sensor placement | |
| Yang et al. | Collective entity alignment for knowledge fusion of power grid dispatching knowledge graphs | |
| Kumar et al. | Advanced analytics on IV curves and electroluminescence images of photovoltaic modules using machine learning algorithms | |
| Han et al. | DeepRouting: A deep neural network approach for ticket routing in expert network | |
| CN114597970A (en) | Active power distribution network partitioning method based on graph convolution network | |
| CN109857892A (en) | Semi-supervised cross-module state Hash search method based on category transmitting | |
| Wang et al. | Multi‐harmonic sources location based on sparse component analysis and complex independent component analysis | |
| Aoufi et al. | Precise and fast parameter identification of mono-crystalline, poly-crystalline, and mono-facial photovoltaic modules using a new Bat Artificial Bee Colony optimizer | |
| CN115734274A (en) | Cellular network fault diagnosis method based on deep learning and knowledge graph | |
| Zhang et al. | Condition-guided adversarial generative testing for deep learning systems | |
| Bertolotti et al. | CombTransformers: Statement-Wise Transformers for Statement-Wise Representations | |
| Zhang et al. | Demand prediction of emergency supplies under fuzzy and missing partial data | |
| Zhuang et al. | DyS-IENN: a novel multiclass imbalanced learning method for early warning of tardiness in rocket final assembly process | |
| Wadood et al. | Fraction order particle swarm optimization for parameter extraction of triple-diode photovoltaic models | |
| CN112560898B (en) | Load space-time prediction method based on deep learning | |
| CN116562124B (en) | Method and device for predicting influence of high-speed aircraft radome ablation on electromagnetic performance |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |