Material data vectorization coding and index matching method and system

Abstract

The invention provides a material data vectorization coding and index matching method and system, and the method comprises the steps: coding elements contained in each material into a plurality of integer fields through employing the characteristic that the number of chemical elements is limited, and constructing a B + tree index for each field in a database; meanwhile, a target element set in the retrieval stage is coded into integer vectors, materials containing specific element sets are quickly retrieved in a multi-integer-field matching mode, multi-dimensional feature coding representation is conducted on material data, multi-dimensional retrieval of the material data is met, material information is fully utilized to establish an all-around efficient retrieval mechanism, and the retrieval efficiency is improved. The matched material data can be accurately and quickly retrieved, and the accuracy of material data identification and the consistency of coding are improved.

Description

Technical Field

[0001] The present invention relates to the field of data processing, and in particular to a method and system for vectorized encoding and index matching of material data. Background Art

[0002] Materials data is used to characterize a material's composition, structure, and physical properties, and is characterized by large amounts of data and complex structures. To meet the needs of materials research in various scenarios, large amounts of materials data are integrated into a database, making it easier for researchers to query and obtain the required materials data.

[0003] In the prior art, patent document CN117171587B discloses a material information encoding method, platform, electronic device and storage medium. This technology extracts a first keyword representing a key technical feature from the material information and recommends a code matching in combination with a coding library. If the first keyword is inconsistent with the standard keyword, a supplementary keyword is introduced for adjustment to generate a unique code. This encoding method realizes the automatic conversion of materials to standardized codes through steps such as keyword standardization, matching evaluation, and unit of measurement analysis, solving the problems of the existing national standard coding system being complicated, the description method being diversified, and the coding being ambiguous, and significantly improving the accuracy of material information identification and the consistency of coding. However, the above-mentioned prior art can only encode a single keyword of the material information, resulting in the coding being related only to the technical features of a single aspect of the material information, and being unable to perform multi-dimensional encoding representation of the material information, and being unable to meet the multi-dimensional retrieval of the material information. It is difficult to fully utilize the material information to establish a comprehensive and efficient retrieval mechanism, and it is impossible to accurately and quickly retrieve matching material data. Summary of the Invention

[0004] The purpose of the present invention is to provide a material data vectorized encoding and index matching method and system. Taking advantage of the limited number of chemical elements, the elements contained in each material are encoded into multiple integer fields, and then a B+ tree index is constructed for each field in the database; at the same time, the target element set in the retrieval stage is encoded into an integer vector, and the materials containing a specific element set are quickly retrieved through a multi-integer field matching method, and the material data is characterized by multi-dimensional feature coding to meet the multi-dimensional retrieval of material data. The material information is fully utilized to establish a comprehensive and efficient retrieval mechanism, and the matching material data is accurately and quickly retrieved, thereby improving the accuracy of material data recognition and the consistency of encoding.

[0005] The present invention is achieved through the following technical solutions:

[0006] Material data vectorized encoding and index matching methods, including:

[0007] Preprocessing material data on a database, and synchronously changing a B+ tree index of the database according to the preprocessing result;

[0008] Extracting material components from the material data to be added, performing vectorization processing and encoding processing on the material components to obtain encoding information of the material components; and updating the B+ tree index according to the encoding information;

[0009] A target element set is extracted from the query request, and the target element set is vectorized and encoded to obtain query encoding information; the query encoding information is matched and compared with the B+ tree index, and target material data is retrieved from the database.

[0010] Optionally, performing material data preprocessing on the database and synchronously changing the B+ tree index of the database according to the preprocessing result includes:

[0011] Compare all material data in the database to determine the similarity between any two material data; wherein the similarity includes the similarity between any two material data in terms of composition, crystal structure and physical properties;

[0012] According to the similarity, all material data in the database are pre-processed by removing duplication and merging;

[0013] According to the deduplication and merging preprocessing results, index information synchronization deduplication and merging change processing is performed on the initial B+ tree index of the database.

[0014] Optionally, extracting material components from the material data to be added, performing vectorization processing and encoding processing on the material components to obtain encoding information of the material components; and updating the B+ tree index according to the encoding information, including:

[0015] Extracting material components from the material data to be added, and obtaining a set of elements included in the material components;

[0016] Generate an element vector of the material component according to the element set included in the material component and a preset element set;

[0017] Dividing the element vector into a plurality of sub-segments, performing integer coding processing on the sub-segments, and obtaining coding information of the material components;

[0018] Generate B+ tree index information corresponding to the encoding information, and store and update the B+ tree index information in the B+ tree index.

[0019] Optionally, extracting a target element set from a query request, performing vectorization processing and encoding processing on the target element set to obtain query encoding information; matching and comparing the query encoding information with the B+ tree index, and retrieving target material data from the database, includes:

[0020] Perform semantic analysis on the query request to obtain the material element fields contained in the query request; perform word extraction and integration on the material element fields to obtain a target element set;

[0021] Generate a query element vector according to the target element set and a preset element set;

[0022] Dividing the query element vector into a number of sub-segments, performing integer encoding processing on the sub-segments to obtain query encoding information;

[0023] The query coding information and the B+ tree index are matched and compared to identify coding information in the B+ tree index that has completely consistent sub-segments with the query coding information; and target material data is retrieved from the database based on the coding information that has completely consistent sub-segments.

[0024] Optionally, the query element vector is divided into several sub-segments, including:

[0025] Extract the maximum allowed query response time;

[0026] Real-time monitoring of the single query response time of each element vector query;

[0027] Performing ratio processing on the single query response time of each element vector query and the maximum query response time to obtain a response time ratio corresponding to the element vector query;

[0028] Obtaining a response time variation factor by querying a corresponding response time ratio using the element vector;

[0029] The response time variation factor is obtained by the following formula:

[0030]

[0031] Where H represents the response time variation factor; n represents the total number of element vector queries; P ti represents the response time ratio corresponding to the i-th element vector query; P tp and P tz Represents the average and median response time ratios corresponding to n element vector queries;

[0032] comparing the response time variation factor with a preset factor threshold;

[0033] When the response time variation factor does not exceed a preset factor threshold, the element vector is maintained with a preset number of initially divided sub-segments;

[0034] When the response time variation factor exceeds a preset factor threshold, the number of sub-segments into which the element vector is divided is adjusted;

[0035] The adjusted number of sub-segments is obtained by the following formula:

[0036]

[0037] Where W represents the number of sub-segments after adjustment. If W is a decimal, the value is rounded up. W0 represents the number of sub-segments for the preset initial division. H represents the response time change factor. H y Indicates the preset factor threshold.

[0038] Material data vectorized encoding and index matching system, including:

[0039] Data preprocessing module, used for preprocessing material data of the database;

[0040] An index synchronization change module, configured to synchronously change the B+ tree index of the database according to the preprocessing result;

[0041] A coding information generation module is used to extract material components from the material data to be added, perform vectorization processing and coding processing on the material components, and obtain coding information of the material components;

[0042] An index updating module, configured to update the B+ tree index according to the encoding information;

[0043] A query code generation module is used to extract a target element set from a query request, perform vectorization processing and encoding processing on the target element set, and obtain query code information;

[0044] A retrieval module is used to match and compare the query coding information with the B+ tree index to retrieve target material data from the database.

[0045] Optionally, the data preprocessing module is used to preprocess material data on the database, including:

[0046] Compare all material data in the database to determine the similarity between any two material data; wherein the similarity includes the similarity between any two material data in terms of composition, crystal structure and physical properties;

[0047] According to the similarity, all material data in the database are pre-processed by removing duplication and merging;

[0048] The index synchronization change module is used to synchronously change the B+ tree index of the database according to the preprocessing result, including:

[0049] According to the deduplication and merging preprocessing results, index information synchronization deduplication and merging change processing is performed on the initial B+ tree index of the database.

[0050] Optionally, the coding information generation module is used to extract material components from the material data to be added, perform vectorization processing and coding processing on the material components, and obtain coding information of the material components, including:

[0051] Extracting material components from the material data to be added, and obtaining a set of elements included in the material components;

[0052] Generate an element vector of the material component according to the element set included in the material component and a preset element set;

[0053] Dividing the element vector into a plurality of sub-segments, performing integer coding processing on the sub-segments, and obtaining coding information of the material components;

[0054] The index updating module is used to update the B+ tree index according to the coding information, including:

[0055] Generate B+ tree index information corresponding to the encoding information, and store and update the B+ tree index information in the B+ tree index.

[0056] Optionally, the query code generation module is configured to extract a target element set from a query request, perform vectorization processing and encoding processing on the target element set, and obtain query code information, including:

[0057] Perform semantic analysis on the query request to obtain the material element fields contained in the query request; perform word extraction and integration on the material element fields to obtain a target element set;

[0058] Generate a query element vector according to the target element set and a preset element set;

[0059] Dividing the query element vector into a number of sub-segments, performing integer encoding processing on the sub-segments to obtain query encoding information;

[0060] The retrieval module is used to match and compare the query code information with the B+ tree index to retrieve target material data from the database, including:

[0061] The query coding information and the B+ tree index are matched and compared to identify coding information in the B+ tree index that has completely consistent sub-segments with the query coding information; and target material data is retrieved from the database based on the coding information that has completely consistent sub-segments.

[0062] Optionally, the query code generation module divides the query element vector into several sub-segments, including:

[0063] Extract the maximum allowed query response time;

[0064] Real-time monitoring of the single query response time of each element vector query;

[0065] Performing ratio processing on the single query response time of each element vector query and the maximum query response time to obtain a response time ratio corresponding to the element vector query;

[0066] Obtaining a response time variation factor by querying a corresponding response time ratio using the element vector;

[0067] The response time variation factor is obtained by the following formula:

[0068]

[0069] Where H represents the response time variation factor; n represents the total number of element vector queries; P ti represents the response time ratio corresponding to the i-th element vector query; P tp and P tz Represents the average and median response time ratios corresponding to n element vector queries;

[0070] comparing the response time variation factor with a preset factor threshold;

[0071] When the response time variation factor does not exceed a preset factor threshold, the element vector is maintained with a preset number of initially divided sub-segments;

[0072] When the response time variation factor exceeds a preset factor threshold, the number of sub-segments into which the element vector is divided is adjusted;

[0073] The adjusted number of sub-segments is obtained by the following formula:

[0074]

[0075] Where W represents the number of sub-segments after adjustment. If W is a decimal, the value is rounded up. W0 represents the number of sub-segments for the preset initial division. H represents the response time change factor. H y Indicates the preset factor threshold.

[0076] Compared with the prior art, the present invention has the following beneficial effects:

[0077] The material data vectorized encoding and index matching method and system provided in this application utilize the limited number of chemical elements to encode the elements contained in each material into multiple integer fields, and then construct a B+ tree index for each field in the database; at the same time, the target element set in the retrieval stage is encoded as an integer vector, and the materials containing a specific element set are quickly retrieved through multi-integer field matching. The material data is characterized by multi-dimensional feature encoding to meet the multi-dimensional retrieval of material data, and fully utilizes material information to establish a comprehensive and efficient retrieval mechanism, accurately and quickly retrieve matching material data, and improve the accuracy of material data identification and the consistency of encoding. BRIEF DESCRIPTION OF THE DRAWINGS

[0078] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative work. Among them:

[0079] Figure 1 A flow chart of the material data vector encoding and index matching method provided by the present invention.

[0080] Figure 2 This is a structural diagram of the material data vector encoding and index matching system provided by the present invention. DETAILED DESCRIPTION

[0081] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are described in detail below in conjunction with the accompanying drawings. It will be understood that the specific embodiments described herein are only used to explain the present application, rather than to limit the present application. It should also be noted that, for ease of description, only some, rather than all, structures related to the present application are shown in the accompanying drawings. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0082] As used herein, the terms "comprise," "comprising," and "having," and any variations thereof, are intended to cover non-exclusive inclusions. For example, a process, method, system, product, or apparatus comprising a series of steps or elements is not limited to the listed steps or elements but may optionally include steps or elements not listed, or may optionally include other steps or elements inherent to the process, method, product, or apparatus.

[0083] References to "embodiments" herein mean that a particular feature, structure, or characteristic described in connection with the embodiments may be included in at least one embodiment of the present application. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor does it constitute an independent or alternative embodiment that is mutually exclusive of other embodiments. It is understood, both explicitly and implicitly, by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0084] See also Figure 1 As shown, an embodiment of the present application provides a material data vectorized encoding and index matching method. The material data vectorized encoding and index matching method includes:

[0085] Preprocess the material data of the database and synchronously change the B+ tree index of the database according to the preprocessing results;

[0086] Extract material components from the material data to be added, perform vectorization and encoding on the material components to obtain encoding information of the material components; update the B+ tree index based on the encoding information;

[0087] The target element set is extracted from the query request, and the target element set is vectorized and encoded to obtain query encoding information; the query encoding information is matched and compared with the B+ tree index, and the target material data is retrieved from the database.

[0088] The beneficial effects of the above embodiments are as follows: the material data vectorized encoding and index matching method utilizes the limited number of chemical elements to encode the elements contained in each material into multiple integer fields, and then constructs a B+ tree index for each field in the database; at the same time, the target element set in the retrieval stage is encoded into an integer vector, and the materials containing a specific element set are quickly retrieved through multi-integer field matching, and the material data is characterized by multi-dimensional feature encoding to meet the multi-dimensional retrieval of material data, and fully utilizes material information to establish a comprehensive and efficient retrieval mechanism, accurately and quickly retrieve matching material data, and improve the accuracy of material data recognition and the consistency of encoding.

[0089] In another embodiment, preprocessing material data on a database and synchronously changing a B+ tree index of the database according to the preprocessing result include:

[0090] Compare all material data in the database to determine the similarity between any two material data; the similarity includes the similarity between any two material data in terms of composition, crystal structure and physical properties;

[0091] Based on the similarity, all material data in the database are pre-processed by removing duplicates and merging;

[0092] Based on the deduplication and merging preprocessing results, the initial B+ tree index of the database is subjected to index information synchronization deduplication and merging change processing.

[0093] In order to achieve the maximum integrated query of material data, a shared database can be constructed. Different users can upload material data to the database after authentication. Taking into account the different editing modes of material data by different users, there are homogeneous heterogeneous data in the database, that is, different users upload two different forms of material data to the database. In fact, these two different forms of material data represent the same data content. The above homogeneous heterogeneous data situation leads to serious data redundancy in the database and wastes storage space. It will also lead to the retrieval of multiple results under the same query request, and the uniqueness and accuracy of the retrieval results cannot be guaranteed. For this reason, it is necessary to perform deduplication and merging preprocessing on all data in the database to achieve the simplicity and uniqueness of the data in the database. Specifically, all material data in the database are compared to determine the similarity between any two material data in terms of chemical element composition, crystal structure and physical properties. Among them, the similarity between two material data in terms of chemical element composition refers to the similarity between the two material data at the level of chemical element type and the number of atoms of each type of chemical element; the similarity between two material data in terms of crystal structure refers to the similarity between the two material data at the level of primitive unit cell basis vectors and atomic coordinates; the similarity between two material data in terms of physical properties refers to the similarity between the two material data at the level of continuous physical properties such as formation energy, density, melting point, boiling point, thermal conductivity, thermal expansion coefficient, and discrete physical properties such as band gap type (direct or indirect) and magnetic order type (paramagnetic, ferromagnetic, etc.). By calculating the similarity between two material data at the level of chemical element composition, crystal structure, and physical property characteristics, the similarity between the two material data can be calculated from different dimensions, thereby improving the reliability of the identification of the similarity between the two material data. A weighted average calculation is performed based on the similarities of the two material data at the three levels of chemical element composition, crystal structure, and physical property characteristics to obtain a comprehensive similarity value for the two material data. The above comprehensive similarity value is compared with a preset similarity threshold. If the above comprehensive similarity value exceeds the preset similarity threshold, the two material data are determined to be highly similar. In this case, the two material data are pre-processed with deduplication and merging of the chemical element composition sub-data, crystal structure sub-data, and physical property sub-data. During the database establishment process, a corresponding initial B+ tree index is synchronously formed. When the material data in the database undergoes deduplication and merging pre-processing, the relevance of the material data in the database also changes. In order to ensure that the B+ tree index truly and correctly reflects the relevance of the material data in the database, the initial B+ tree index of the database is subjected to index information synchronization deduplication and merging modification processing based on the deduplication and merging pre-processing results to ensure that the relevance between the modified B+ tree index and the material data in the database remains consistent. Among them, constructing the database B+ tree index and performing the index information synchronization deduplication and merging modification processing on the initial B+ tree index are both conventional methods in the field and are not described in detail here.

[0094] In another embodiment, material components are extracted from the material data to be added, and the material components are vectorized and encoded to obtain encoding information of the material components; and a B+ tree index is updated according to the encoding information, including:

[0095] Extracting material components from the material data to be added, and obtaining a set of elements contained in the material components;

[0096] Generate an element vector of the material component according to the element set contained in the material component and the pre-set element set;

[0097] Divide the element vector into several sub-segments, perform integer coding on the sub-segments, and obtain the coding information of the material components;

[0098] Generate B+ tree index information corresponding to the encoding information, and store and update the B+ tree index information in the B+ tree index.

[0099] When uploading new material data to the database, in order to ensure that the B+ tree index will form index information consistent with the new material data after the new material data is added to the database, the new material data needs to be vectorized and encoded. Specifically, there are currently 118 known chemical elements, and the set of elements is pre-set as E = {e1, e2, ..., e k}, where k = 118. The element vector of the material components extracted from the new material data is defined as follows:

[0100]

[0101] in, It is the set of elements corresponding to the material components extracted from the new material data.

[0102] The element vector Divided into consecutive 32-bit subsegments as follows:

[0103] where j = 1, 2, 3, 4

[0104] If the sub-segment is less than 32 bits, the high bits will be automatically filled with 0.

[0105] The integer representation of each sub-segment is as follows:

[0106]

[0107] Among them, d j is the integer representation of the j-th sub-segment;

[0108] Finally, the final storage form D of the encoded information of the new material data is as follows:

[0109]

[0110] In the above formula, Represents the complete set of tuples of 4 integers.

[0111] In the database, encoding information of the new material data is established for the above four integer sub-segments, and B+ tree index information corresponding to the encoding information is generated. The B+ tree index information is stored and updated in the B+ tree index. In essence, all chemical elements are mapped into a fixed-length Boolean vector space and divided into multiple small segments with a fixed bit width (for example, 32 bits) in sequence. Each Boolean sub-vector is further converted into an integer segment to form encoding information in the form of a material data vector, thereby realizing the uniqueness calibration of the material data index information.

[0112] In another embodiment, a target element set is extracted from a query request, the target element set is vectorized and encoded to obtain query encoding information; the query encoding information is matched and compared with a B+ tree index, and target material data is retrieved from a database, including:

[0113] Perform semantic analysis on the query request to obtain the material element fields contained in the query request; perform word extraction and integration on the material element fields to obtain the target element set;

[0114] Generate a query element vector based on the target element set and the pre-set element set;

[0115] Divide the query element vector into several sub-segments, perform integer encoding processing on the sub-segments, and obtain query encoding information;

[0116] The query coding information and the B+ tree index are matched and compared to identify the coding information in the B+ tree index that has completely consistent sub-segments with the query coding information; based on the coding information with completely consistent sub-segments, the target material data is retrieved from the database.

[0117] When searching for target material data in a database, the query request is first subjected to text semantic parsing to obtain all material element fields contained in the query request. These material element fields may include, but are not limited to, fields containing material element characters. Key material element-related terms are then extracted and integrated from all material element fields to obtain a target element set. This target element set may include, but is not limited to, a set of elements corresponding to the material components to be retrieved in the query request. Following the same process described above, the target element set is then matched against a pre-defined set of elements to generate a query element vector. The query element vector is then divided into several sub-segments, and the sub-segments are integer-encoded to obtain query encoding information.

[0118] The above-mentioned query coding information and the index information of the B+ tree index of the database have the same four integer format, so that the above-mentioned query coding information can be directly compared with the B+ tree index without any conversion processing. To this end, the query coding information and the B+ tree index are matched and compared, and the coding information in the B+ tree index that has a completely consistent sub-segment with the query coding information is identified, so that the material data corresponding to the above-mentioned coding information with a completely consistent sub-segment in the database is used as the target material data retrieved by the query request corresponding to the above-mentioned query, ensuring that the matching material data is retrieved accurately and quickly, and improving the accuracy of material data identification and the consistency of coding.

[0119] In another embodiment, the query element vector is divided into several sub-segments, including:

[0120] Extract the maximum allowed query response time;

[0121] Real-time monitoring of the single query response time of each element vector query;

[0122] The response time ratio of the element vector query is obtained by performing ratio processing on the single query response time of each element vector query and the maximum query response time;

[0123] Obtain the response time variation factor by querying the corresponding response time ratio using the element vector;

[0124] The response time variation factor is obtained by the following formula:

[0125]

[0126] Where H represents the response time variation factor; n represents the total number of element vector queries; P ti represents the response time ratio corresponding to the i-th element vector query; P tp and P tz Represents the average and median response time ratios corresponding to n element vector queries;

[0127] comparing the response time variation factor with a preset factor threshold;

[0128] When the response time variation factor does not exceed a preset factor threshold, the element vector is maintained with a preset number of initially divided sub-segments;

[0129] When the response time variation factor exceeds a preset factor threshold, the number of sub-segments into which the element vector is divided is adjusted;

[0130] The adjusted number of sub-segments is obtained by the following formula:

[0131]

[0132] Where W represents the number of sub-segments after adjustment. If W is a decimal, the value is rounded up. W0 represents the number of sub-segments for the preset initial division. H represents the response time change factor. H y Indicates the preset factor threshold.

[0133] By extracting the maximum allowed query response time and monitoring the response time of each element vector query in real time, the response time ratio and response time variation factor are calculated. When the response time variation factor exceeds a preset threshold, the number of sub-segments used to divide the element vector is adjusted. This allows for dynamic optimization of the element vector division based on actual query response time variations, potentially reducing the processing time of each query, improving query efficiency, and ensuring that query response times are closer to or meet system requirements. Properly adjusting the number of sub-segments used to divide the element vector enables the system to more efficiently utilize system resources when processing query requests. For example, when query response times are long, increasing the number of sub-segments may break down the query task into smaller pieces, thereby fully utilizing the system's parallel processing capabilities and avoiding resource waste. Conversely, when query response times are short, reducing the number of sub-segments can reduce system management overhead and improve resource utilization. By dynamically adjusting the number of sub-segments used to divide the element vector, the system can adaptively adjust its processing strategy based on changes in query response time. This helps prevent system anomalies or performance fluctuations caused by excessively long or short query response times, thereby enhancing system stability and reliability, and improving overall system performance and service quality.

[0134] See also Figure 2 As shown, an embodiment of the present application provides a material data vectorized encoding and index matching system. The material data vectorized encoding and index matching system includes:

[0135] Data preprocessing module, used for preprocessing material data of the database;

[0136] Index synchronization change module, used to synchronize the change of the database's B+ tree index according to the preprocessing results;

[0137] A coding information generation module is used to extract material components from the material data to be added, perform vectorization processing and coding processing on the material components, and obtain coding information of the material components;

[0138] Index update module, used to update the B+ tree index according to the encoding information;

[0139] A query code generation module is used to extract a target element set from a query request, vectorize and encode the target element set, and obtain query code information;

[0140] The retrieval module is used to match and compare the query coding information with the B+ tree index and retrieve the target material data from the database.

[0141] The beneficial effects of the above embodiments are as follows: the material data vectorized encoding and index matching system utilizes the limited number of chemical elements to encode the elements contained in each material into multiple integer fields, and then constructs a B+ tree index for each field in the database; at the same time, the target element set in the retrieval stage is encoded into an integer vector, and the materials containing a specific element set are quickly retrieved through multi-integer field matching, and the material data is characterized by multi-dimensional feature encoding to meet the multi-dimensional retrieval of material data, and fully utilizes material information to establish a comprehensive and efficient retrieval mechanism, accurately and quickly retrieve matching material data, and improve the accuracy of material data recognition and the consistency of encoding.

[0142] In another embodiment, the data preprocessing module is used to preprocess the material data of the database, including:

[0143] Compare all material data in the database to determine the similarity between any two material data; the similarity includes the similarity between any two material data in terms of composition, crystal structure and physical properties;

[0144] Based on the similarity, all material data in the database are pre-processed by removing duplicates and merging;

[0145] The index synchronization change module is used to synchronize the changes to the database's B+ tree index based on the preprocessing results, including:

[0146] Based on the deduplication and merging preprocessing results, the initial B+ tree index of the database is subjected to index information synchronization deduplication and merging change processing.

[0147] In order to achieve the maximum integrated query of material data, a shared database can be constructed. Different users can upload material data to the database after authentication. Taking into account the different editing modes of material data by different users, there are homogeneous heterogeneous data in the database, that is, different users upload two different forms of material data to the database. In fact, these two different forms of material data represent the same data content. The above homogeneous heterogeneous data situation leads to serious data redundancy in the database and wastes storage space. It will also lead to the retrieval of multiple results under the same query request, and the uniqueness and accuracy of the retrieval results cannot be guaranteed. For this reason, it is necessary to perform deduplication and merging preprocessing on all data in the database to achieve the simplicity and uniqueness of the data in the database. Specifically, all material data in the database are compared to determine the similarity between any two material data in terms of chemical element composition, crystal structure and physical properties. Among them, the similarity between two material data in terms of chemical element composition refers to the similarity between the two material data at the level of chemical element type and the number of atoms of each type of chemical element; the similarity between two material data in terms of crystal structure refers to the similarity between the two material data at the level of primitive unit cell basis vectors and atomic coordinates; the similarity between two material data in terms of physical properties refers to the similarity between the two material data at the level of continuous physical properties such as formation energy, density, melting point, boiling point, thermal conductivity, thermal expansion coefficient, and discrete physical properties such as band gap type (direct or indirect) and magnetic order type (paramagnetic, ferromagnetic, etc.). By calculating the similarity between two material data at the level of chemical element composition, crystal structure, and physical property characteristics, the similarity between the two material data can be calculated from different dimensions, thereby improving the reliability of the identification of the similarity between the two material data. A weighted average calculation is performed based on the similarities of the two material data at the three levels of chemical element composition, crystal structure, and physical property characteristics to obtain a comprehensive similarity value for the two material data. The above comprehensive similarity value is compared with a preset similarity threshold. If the above comprehensive similarity value exceeds the preset similarity threshold, the two material data are determined to be highly similar. In this case, the two material data are pre-processed with deduplication and merging of the chemical element composition sub-data, crystal structure sub-data, and physical property sub-data. During the database establishment process, a corresponding initial B+ tree index is synchronously formed. When the material data in the database undergoes deduplication and merging pre-processing, the relevance of the material data in the database also changes. In order to ensure that the B+ tree index truly and correctly reflects the relevance of the material data in the database, the initial B+ tree index of the database is subjected to index information synchronization deduplication and merging modification processing based on the deduplication and merging pre-processing results to ensure that the relevance between the modified B+ tree index and the material data in the database remains consistent. Among them, constructing the database B+ tree index and performing the index information synchronization deduplication and merging modification processing on the initial B+ tree index are both conventional methods in the field and are not described in detail here.

[0148] In another embodiment, the coding information generation module is used to extract material components from the material data to be added, perform vectorization processing and coding processing on the material components, and obtain coding information of the material components, including:

[0149] Extracting material components from the material data to be added, and obtaining a set of elements contained in the material components;

[0150] Generate an element vector of the material component according to the element set contained in the material component and the pre-set element set;

[0151] Divide the element vector into several sub-segments, perform integer coding on the sub-segments, and obtain the coding information of the material components;

[0152] The index update module is used to update the B+ tree index based on the encoding information, including:

[0153] Generate B+ tree index information corresponding to the encoding information, and store and update the B+ tree index information in the B+ tree index.

[0154] When uploading new material data to the database, in order to ensure that the B+ tree index will form index information consistent with the new material data after the new material data is added to the database, the new material data needs to be vectorized and encoded. Specifically, there are currently 118 known chemical elements, and the set of elements is pre-set as E = {e1, e2, ..., e k}, where k = 118. The element vector of the material components extracted from the new material data is defined as follows:

[0155]

[0156] in, It is the set of elements corresponding to the material components extracted from the new material data.

[0157] The element vector Divided into consecutive 32-bit subsegments as follows:

[0158] where j = 1, 2, 3, 4

[0159] If the sub-segment is less than 32 bits, the high bits will be automatically filled with 0.

[0160] The integer representation of each sub-segment is as follows:

[0161]

[0162] Among them, d j is the integer representation of the j-th sub-segment;

[0163] Finally, the final storage form D of the encoded information of the new material data is as follows:

[0164]

[0165] In the above formula, Represents the complete set of tuples of 4 integers.

[0166] In the database, encoding information of the new material data is established for the above four integer sub-segments, and B+ tree index information corresponding to the encoding information is generated. The B+ tree index information is stored and updated in the B+ tree index. In essence, all chemical elements are mapped into a fixed-length Boolean vector space and divided into multiple small segments with a fixed bit width (for example, 32 bits) in sequence. Each Boolean sub-vector is further converted into an integer segment to form encoding information in the form of a material data vector, thereby realizing the uniqueness calibration of the material data index information.

[0167] In another embodiment, the query code generation module is configured to extract a target element set from a query request, perform vectorization processing and encoding processing on the target element set, and obtain query code information, including:

[0168] Perform semantic analysis on the query request to obtain the material element fields contained in the query request; perform word extraction and integration on the material element fields to obtain the target element set;

[0169] Generate a query element vector based on the target element set and the pre-set element set;

[0170] Divide the query element vector into several sub-segments, perform integer encoding processing on the sub-segments, and obtain query encoding information;

[0171] The retrieval module is used to match the query code information with the B+ tree index and retrieve the target material data from the database, including:

[0172] The query coding information and the B+ tree index are matched and compared to identify the coding information in the B+ tree index that has completely consistent sub-segments with the query coding information; based on the coding information with completely consistent sub-segments, the target material data is retrieved from the database.

[0173] When searching for target material data in a database, the query request is first subjected to text semantic parsing to obtain all material element fields contained in the query request. These material element fields may include, but are not limited to, fields containing material element characters. Key material element-related terms are then extracted and integrated from all material element fields to obtain a target element set. This target element set may include, but is not limited to, a set of elements corresponding to the material components to be retrieved in the query request. Following the same process described above, the target element set is then matched against a pre-defined set of elements to generate a query element vector. The query element vector is then divided into several sub-segments, and the sub-segments are integer-encoded to obtain query encoding information.

[0174] The above-mentioned query coding information and the index information of the B+ tree index of the database have the same four integer format, so that the above-mentioned query coding information can be directly compared with the B+ tree index without any conversion processing. To this end, the query coding information and the B+ tree index are matched and compared, and the coding information in the B+ tree index that has a completely consistent sub-segment with the query coding information is identified, so that the material data corresponding to the above-mentioned coding information with a completely consistent sub-segment in the database is used as the target material data retrieved by the query request corresponding to the above-mentioned query, ensuring that the matching material data is retrieved accurately and quickly, and improving the accuracy of material data identification and the consistency of coding.

[0175] In another embodiment, the query code generation module divides the query element vector into several sub-segments, including:

[0176] Extract the maximum allowed query response time;

[0177] Real-time monitoring of the single query response time of each element vector query;

[0178] The response time ratio of the element vector query is obtained by performing ratio processing on the single query response time of each element vector query and the maximum query response time;

[0179] Obtain the response time variation factor by querying the corresponding response time ratio using the element vector;

[0180] The response time variation factor is obtained by the following formula:

[0181]

[0182] Where H represents the response time variation factor; n represents the total number of element vector queries; P ti represents the response time ratio corresponding to the i-th element vector query; P tp and P tz Represents the average and median response time ratios corresponding to n element vector queries;

[0183] comparing the response time variation factor with a preset factor threshold;

[0184] When the response time variation factor does not exceed a preset factor threshold, the element vector is maintained with a preset number of initially divided sub-segments;

[0185] When the response time variation factor exceeds a preset factor threshold, the number of sub-segments into which the element vector is divided is adjusted;

[0186] The adjusted number of sub-segments is obtained by the following formula:

[0187]

[0188] Where W represents the number of sub-segments after adjustment. If W is a decimal, the value is rounded up. W0 represents the number of sub-segments for the preset initial division. H represents the response time change factor. H y Indicates the preset factor threshold.

[0189] By extracting the maximum allowed query response time and monitoring the response time of each element vector query in real time, the response time ratio and response time variation factor are calculated. When the response time variation factor exceeds a preset threshold, the number of sub-segments used to divide the element vector is adjusted. This allows for dynamic optimization of the element vector division based on actual query response time variations, potentially reducing the processing time of each query, improving query efficiency, and ensuring that query response times are closer to or meet system requirements. Properly adjusting the number of sub-segments used to divide the element vector enables the system to more efficiently utilize system resources when processing query requests. For example, when query response times are long, increasing the number of sub-segments may break down the query task into smaller pieces, thereby fully utilizing the system's parallel processing capabilities and avoiding resource waste. Conversely, when query response times are short, reducing the number of sub-segments can reduce system management overhead and improve resource utilization. By dynamically adjusting the number of sub-segments used to divide the element vector, the system can adaptively adjust its processing strategy based on changes in query response time. This helps prevent system anomalies or performance fluctuations caused by excessively long or short query response times, thereby enhancing system stability and reliability, and improving overall system performance and service quality.

[0190] In general, the material data vectorized encoding and index matching method and system utilizes the limited number of chemical elements to encode the elements contained in each material into multiple integer fields, and then constructs a B+ tree index for each field in the database; at the same time, the target element set in the retrieval stage is encoded as an integer vector, and the materials containing a specific element set are quickly retrieved through multi-integer field matching. The material data is characterized by multi-dimensional feature encoding to meet the multi-dimensional retrieval of material data, and the material information is fully utilized to establish a comprehensive and efficient retrieval mechanism, so as to accurately and quickly retrieve matching material data, thereby improving the accuracy of material data identification and the consistency of encoding.

[0191] The above is only a specific embodiment of the present invention, and any other improvements made based on the concept of the present invention are considered to be within the scope of protection of the present invention.

Claims

1. Material data vectorized encoding and index matching method, characterized in that: include: Preprocessing material data on a database, and synchronously changing a B+ tree index of the database according to the preprocessing result; Extracting material components from the material data to be added, performing vectorization processing and encoding processing on the material components to obtain encoding information of the material components; and updating the B+ tree index according to the encoding information; A target element set is extracted from the query request, and the target element set is vectorized and encoded to obtain query encoding information; the query encoding information is matched and compared with the B+ tree index, and target material data is retrieved from the database.

2. The material data vectorized encoding and index matching method according to claim 1, characterized in that: Preprocessing material data on a database, and synchronously changing a B+ tree index of the database according to the preprocessing result, including: Compare all material data in the database to determine the similarity between any two material data; wherein the similarity includes the similarity between any two material data in terms of composition, crystal structure and physical properties; Based on the similarity, all material data in the database are pre-processed for deduplication and merging; based on the results of the pre-processing for deduplication and merging, the initial B+ tree index of the database is subjected to index information synchronization deduplication and merging change processing.

3. The material data vectorized encoding and index matching method according to claim 1, characterized in that: Extracting material components from the material data to be added, performing vectorization processing and encoding processing on the material components, and obtaining encoding information of the material components; Updating the B+ tree index according to the encoding information includes: Extracting material components from the material data to be added to obtain a set of elements included in the material components; generating an element vector of the material components based on the set of elements included in the material components and a pre-set complete set of elements; Dividing the element vector into a plurality of sub-segments, performing integer coding processing on the sub-segments, and obtaining coding information of the material components; Generate B+ tree index information corresponding to the encoding information, and store and update the B+ tree index information in the B+ tree index.

4. The material data vectorized encoding and index matching method according to claim 1, wherein: Extracting a target element set from a query request, performing vectorization processing and encoding processing on the target element set to obtain query encoding information; Matching and comparing the query code information with the B+ tree index, and retrieving target material data from the database, including: Perform semantic analysis on the query request to obtain the material element fields contained in the query request; perform word extraction and integration on the material element fields to obtain a target element set; Generate a query element vector according to the target element set and a preset element set; Dividing the query element vector into a number of sub-segments, performing integer encoding processing on the sub-segments to obtain query encoding information; The query coding information and the B+ tree index are matched and compared to identify coding information in the B+ tree index that has completely consistent sub-segments with the query coding information; and target material data is retrieved from the database based on the coding information that has completely consistent sub-segments.

5. The material data vectorized encoding and index matching method according to claim 4, characterized in that: Divide the query element vector into several sub-segments, including: Extract the maximum allowed query response time; Real-time monitoring of the single query response time of each element vector query; Performing ratio processing on the single query response time of each element vector query and the maximum query response time to obtain a response time ratio corresponding to the element vector query; Obtaining a response time variation factor by querying a corresponding response time ratio using the element vector; The response time variation factor is obtained by the following formula: Where H represents the response time variation factor; n represents the total number of element vector queries; P ti represents the response time ratio corresponding to the i-th element vector query; P tp and P tz Represents the average and median response time ratios corresponding to n element vector queries; comparing the response time variation factor with a preset factor threshold; When the response time variation factor does not exceed a preset factor threshold, the element vector is maintained with a preset number of initially divided sub-segments; When the response time variation factor exceeds a preset factor threshold, the number of sub-segments into which the element vector is divided is adjusted; The adjusted number of sub-segments is obtained by the following formula: Where W represents the number of sub-segments after adjustment. If W is a decimal, the value is rounded up. W0 represents the number of sub-segments for the preset initial division. H represents the response time change factor. H y Indicates the preset factor threshold.

6. Material data vector coding and index matching system, characterized by: include: Data preprocessing module, used for preprocessing material data of the database; An index synchronization change module, configured to synchronously change the B+ tree index of the database according to the preprocessing result; A coding information generation module is used to extract material components from the material data to be added, perform vectorization processing and coding processing on the material components, and obtain coding information of the material components; An index updating module, configured to update the B+ tree index according to the encoding information; A query code generation module is used to extract a target element set from a query request, perform vectorization processing and encoding processing on the target element set, and obtain query code information; A retrieval module is used to match and compare the query coding information with the B+ tree index to retrieve target material data from the database.

7. The material data vectorized encoding and index matching system according to claim 6, characterized in that: The data preprocessing module is used to preprocess the material data of the database, including: Compare all material data in the database to determine the similarity between any two material data; wherein the similarity includes the similarity between any two material data in terms of composition, crystal structure and physical properties; Based on the similarity, all material data in the database are pre-processed by deduplication and merging; the index synchronization change module is used to synchronously change the B+ tree index of the database according to the pre-processing result, including: According to the deduplication and merging preprocessing results, index information synchronization deduplication and merging change processing is performed on the initial B+ tree index of the database.

8. The material data vectorized encoding and index matching system according to claim 6, characterized in that: The coding information generation module is used to extract material components from the material data to be added, perform vectorization processing and coding processing on the material components, and obtain coding information of the material components, including: Extracting material components from the material data to be added to obtain a set of elements included in the material components; generating an element vector of the material components based on the set of elements included in the material components and a pre-set complete set of elements; Dividing the element vector into a plurality of sub-segments, performing integer coding processing on the sub-segments, and obtaining coding information of the material components; The index updating module is used to update the B+ tree index according to the coding information, including: Generate B+ tree index information corresponding to the encoding information, and store and update the B+ tree index information in the B+ tree index.

9. The material data vectorized encoding and index matching system according to claim 6, characterized in that: The query code generation module is used to extract a target element set from a query request, perform vectorization processing and encoding processing on the target element set, and obtain query code information, including: Perform semantic analysis on the query request to obtain the material element fields contained in the query request; perform word extraction and integration on the material element fields to obtain a target element set; Generate a query element vector according to the target element set and a preset element set; Dividing the query element vector into a number of sub-segments, performing integer encoding processing on the sub-segments to obtain query encoding information; The retrieval module is used to match and compare the query code information with the B+ tree index to retrieve target material data from the database, including: The query coding information and the B+ tree index are matched and compared to identify coding information in the B+ tree index that has completely consistent sub-segments with the query coding information; and target material data is retrieved from the database based on the coding information that has completely consistent sub-segments.

10. The material data vectorized encoding and index matching system according to claim 9, characterized in that: The query code generation module divides the query element vector into several sub-segments, including: Extract the maximum allowed query response time; Real-time monitoring of the single query response time of each element vector query; Performing ratio processing on the single query response time of each element vector query and the maximum query response time to obtain a response time ratio corresponding to the element vector query; Obtaining a response time variation factor by querying a corresponding response time ratio using the element vector; The response time variation factor is obtained by the following formula: Where H represents the response time variation factor; n represents the total number of element vector queries; P ti represents the response time ratio corresponding to the i-th element vector query; P tp and P tz Represents the average and median response time ratios corresponding to n element vector queries; comparing the response time variation factor with a preset factor threshold; When the response time variation factor does not exceed a preset factor threshold, the element vector is maintained with a preset number of initially divided sub-segments; When the response time variation factor exceeds a preset factor threshold, the number of sub-segments into which the element vector is divided is adjusted; The adjusted number of sub-segments is obtained by the following formula: Where W represents the number of sub-segments after adjustment. If W is a decimal, the value is rounded up. W0 represents the number of sub-segments for the preset initial division. H represents the response time change factor. H y Indicates the preset factor threshold.

Images