An inertial platform test data self-supervision deep processing analysis system

Abstract

The application provides a kind of inertial platform test data self-supervision deep processing analysis system, the system is divided into application layer and data processing layer, application layer includes test procedure and result analysis, algorithm verification and precision analysis and inertial information autonomous visualization display unit.The system has the ability of test data self-supervision deep processing analysis, can automatically traverse search and intelligently screen and identify various types of inertial platform test data, and carries out batch processing analysis, completes the template of test procedure and result whole compilation report form;For inertial platform precision test procedure data, automatically call each item system parameter in test data directory and matching modular algorithm, realize the offline verification of inertial platform key algorithm and the calculation and analysis of each precision test;It can adaptively extract the characteristic data of inertial platform key typical inertial information, and complete data visualization display by independently configuring characteristic data visualization strategy.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of inertial platform test data analysis, and particularly relates to an inertial platform test data self-supervised deep processing analysis system. BACKGROUND

[0002] An inertial platform is a stable platform system for measuring carrier attitude information and visual acceleration, which is composed of a gyroscope, a platform body, a frame system and a stable loop. The gyroscope installed on the platform body senses the angular velocity of the platform body relative to the inertial space, and controls it through the stable loop to keep the inertial platform body stable in the inertial space.

[0003] As a key navigation unit in the automatic control system, the new generation of inertial platform has realized many technical breakthroughs such as self-calibration, self-aiming and self-monitoring, and the degree of automation has been significantly improved. With the development of SOC technology, the data acquisition, calculation and transmission of the inertial platform have been greatly accelerated, the data acquisition covers a wide range and has various types, and a large amount of test data is generated in a unit of time. In addition, the inertial platform has many test, ground verification test and flight test tasks, and the continuous power-on time is long (usually more than one month), and a large amount of test data will be generated for each inertial platform. The real performance of the inertial platform needs to be measured by accurate and comprehensive data analysis, so it is very important to accurately and efficiently process the inertial platform test data, accurately analyze the function and performance of the inertial platform, and timely diagnose and locate the faults of the inertial platform. SUMMARY

[0004] The technical problem of the present application is to overcome the shortcomings of the general test data processing method, such as complicated process, long cycle, high labor cost, low accuracy and low professionalism, and to develop an inertial platform test data self-supervised deep processing analysis system according to the technical points and test data characteristics of the new generation of inertial platform. The basic design principles are automation, intelligence, batch processing and professionalism. Through in-depth excavation of the functional requirements of inertial platform test data processing, various data processing modular intelligent algorithms are comprehensively used to significantly improve the deep processing capability of inertial platform test data.

[0005] The technical scheme provided by the present application is as follows:

[0006] An inertial platform test data self-supervised deep processing analysis system, through the comprehensive application of a plurality of key data processing algorithms such as a test data self-traversal module, a self-identification reduction module, a batch processing and compilation report module, an abnormal data self-diagnosis processing module, a modular algorithm self-adaptive matching module and a feature data self-visualization module, realizes the functions of automatic retrieval and batch statistics of inertial platform test process / results, offline verification and precision evaluation of inertial navigation algorithm and autonomous visualization display of inertial information. Specifically, the system comprises:

[0007] Application layer:

[0008] Test flow and result analysis unit, for implementing complete traversal of all test flows and comprehensive retrieval of test data through the test data self-traversal module, and for completing screening and identification of specific test types and data and self-matching and processing analysis through the self-identification reduction module; and for implementing template-based compilation and reporting of analysis data through the batch processing compilation and reporting module;

[0009] Algorithm verification and precision analysis unit, for completing screening of test data through the test data self-traversal module and the self-identification reduction module, automatically matching each system parameter in the test data file and the modular algorithm unit through the modular algorithm adaptive matching module, implementing offline verification of key algorithms of the inertial platform and calculation and analysis of each precision test, and implementing compilation and reporting of analysis data through the batch processing compilation and reporting module;

[0010] Inertial information visualization display unit, for data extraction, mathematical calculation and data display of target inertial information data in the inertial platform test, implementing normalized visualization display of different types of graphical curves through the characteristic data self-visualization module and the modular algorithm adaptive matching module;

[0011] Data processing layer:

[0012] Test data self-traversal module, for implementing complete traversal and retrieval of the inertial platform test flow and data through the breadth-first search algorithm and the depth-first search algorithm;

[0013] Self-identification reduction module, for completing reduction of the test flow through consistent matching of the specific test flow of the inertial platform, and realizing screening and identification of the specific test flow;

[0014] Batch processing compilation and reporting module, for statistics, classification summary and result arrangement of the test flow and data, implementing classification and template-based reporting of different types of data;

[0015] Abnormal data self-diagnosis processing module, for automatically diagnosing abnormal data in the test and performing fault-tolerant processing according to the set standardized data preprocessing criteria;

[0016] Modular algorithm adaptive matching module, for automatically matching each system parameter in the test data and the modular algorithm, implementing offline verification of key algorithms of the inertial platform and calculation and analysis of each precision test;

[0017] Characteristic data self-visualization module, for adaptively extracting inertial information data in the inertial platform test based on the examination requirements of different test flows, and completing self-visualization display of characteristic data of the inertial platform.

[0018] The inertial platform test data self-supervision deep processing analysis system has the following beneficial effects:

[0019] (1) The inertial platform test data self-supervision deep processing analysis system provided by the application has the typical characteristics of comprehensiveness, accuracy, automation and intelligence. Whether the scale of the test data is large or small, the overall, efficient and accurate analysis of the test data of the inertial platform can be automatically realized by specifying the data range once, the data report is standardized, and the data visibility is strong.

[0020] (2) The inertial platform test data self-supervision deep processing analysis system provided by the application realizes comprehensive, accurate, intelligent and efficient batch deep processing analysis of the inertial platform test data.

[0021] (3) The inertial platform test data self-supervision deep processing analysis system provided by the application realizes comprehensive and in-depth processing and analysis of the inertial platform test data with the least labor cost and maximization, greatly shortens the data processing time, and improves the work efficiency by more than 20 times. BRIEF DESCRIPTION OF DRAWINGS

[0022] Figure 1 is the inertial platform test data self-supervision deep processing analysis system architecture;

[0023] Figure 2 is the basic composition of the application layer of the inertial platform test data self-supervision deep processing analysis system;

[0024] Figure 3 is the system application unit function module data flow diagram;

[0025] Figure 4 is the inertial information self-visualization display effect diagram. DETAILED DESCRIPTION

[0026] The characteristics and advantages of the application will become clearer and more explicit with the following detailed description of the application.

[0027] The word "exemplary" is used herein to mean "serving as an example, instance, or illustration." Any implementation described herein as "exemplary" is not necessarily to be construed as preferred or advantageous over other implementations. Although various aspects of embodiments are illustrated in the drawings, the drawings are not necessarily drawn to scale.

[0028] The application provides a kind of inertial platform test data self-supervision deep processing analysis system, for the two-layer architecture system consisting of application layer and data processing layer, wherein application layer is composed of three major application units of test process and result analysis unit 10, algorithm verification and precision analysis unit 20 and inertial information visualization display unit 30, data processing layer is composed of six major algorithm modules of test data self-traversal module 1, self-identification reduction module 2, batch processing compilation report module 3, abnormal data self-diagnosis processing module 4, modular algorithm self-adaptive matching module 5 and feature data autonomous visualization module 6, system architecture as shown in Figure 1 .

[0029] By the application unit of application layer calling each algorithm module of data processing layer, automatically traversing search and intelligently screening and identifying various test data of inertial platform, and carrying out batch processing analysis, template compilation report of test process and result is completed;For inertial platform precision test process data, automatically match and call each item of system parameter and modular algorithm in test data directory, realize offline verification of inertial platform key algorithm and calculation and analysis of each precision test;Adaptive extraction of inertial platform key inertial information and typical self-monitoring analog quantity and other characteristic data, self-selection of feature data visualization strategy completes data visualization display. Basic composition of system as shown in Figure 2 , this system includes test process and result analysis unit 10, algorithm verification and precision analysis unit 20 and inertial information visualization display unit 30 three application units, wherein: the first application unit includes seven function modules of test process statistical module 11, monitoring analog quantity statistical module 12, self-aiming result statistical module 13, self-calibration result statistical module 14, instruction execution time statistical module 15, positioning error calculation module 16 and starlight data statistical module 17;Second application unit includes four function modules of tilt adjustment precision calculation module 21, self-aiming precision calculation module 22, self-calibration precision calculation module 23 and flight navigation precision calculation module 24;Third application unit includes seven function modules of instruction current display module 31, frame angle display module 32, accelerometer output display module 33, torque motor current display module 34, temperature and temperature control display module 35, model flight process display module 36 and telemetry data display module 37.

[0030] The system runs can be compared according to tree structure, data flow diagram as Figure 3The test data is injected into the trunk part, and after injection, each branch (①-⑥) is differentiated and grown, which is a breadth-first tree. Some branches do not meet the conditions and do not grow, only the branches that meet the conditions (①, ③ and ④) continue to grow, which is a depth-first tree. At the end of the depth-first tree branch, various leaves will grow, representing different data files. Different functional modules need to select different leaves as processing objects for feature data extraction and calculation analysis. Finally, the processing and analysis results of each processing object will be compiled and summarized in a report.

[0031] The following describes the implementation process of each functional module in the three application units of the system:

[0032] (1) Inertial platform test process and result analysis unit 10:

[0033] This application unit first realizes complete traversal of all test processes and comprehensive retrieval of test data through the test data self-traversal module 1. Then, through the self-identification reduction module 2, intelligent filtering and identification of specific test types and data are completed, and processing and analysis are performed respectively. Finally, based on the batch processing and report compilation module 3, the analysis data is template-based compiled and reported. This application unit includes seven functional modules, which are as follows:

[0034] 1.1) Test process statistics module 11: The various test data collected by the inertial platform periodically will be stored according to the product number, test date and test process according to the predetermined file path. Different inertial platform product data storage formats are the same.

[0035] When performing test process statistics through this functional module, the test data self-traversal module 1 of the data processing layer is mainly used. Based on the breadth-first search algorithm, all folder directories under the selected path are searched globally, and the search is performed level by level according to the storage form of the current test process data. As shown in Table 1, according to the specified product number (for example: P315 inertial platform) path, first traverse the date path, and all date paths are used as the first level breadth-first tree (for example: three "date paths" in Table 1). Then, traverse the process path under each date path as the second level breadth-first tree. Finally, traverse the data file path under each process number path (traverse the corresponding process number), as the third level breadth-first tree. Through the three-level breadth-first tree, a global search directory is formed, and automatic retrieval and global traversal of all test processes of the selected product are realized. The self-traversal technique is not limited by the number of file directory levels and the number of files, and can automatically filter and traverse to count the test date, test process name and process number in all test processes, and arrange and summarize them in chronological order, as shown in Table 2.

[0036] Table 1 Test process statistics module self-traversal breadth-first tree

[0037]

[0038] Table 2 Test flow summary report

[0039]

[0040] 1.2) Monitor analog statistics module 12: not all test flows need to monitor analog, so when monitoring analog statistics of inertial platform, only the data of the inertial platform system in stable working state in the specific test flow is identified, screened and effective data extraction calculation.

[0041] The monitoring analog statistics module 12 first needs to establish a "monitoring analog flow library", which encapsulates the test flow name that needs to be monitored analog, and continues to use the test data self-traversal module 1 of the data processing layer. Based on the three-level breadth-first tree, further with the help of the depth-first search algorithm, the specific flow in the inertial platform power-on test is set as the backtracking boundary condition based on the "monitoring analog flow library", the folder path that meets the boundary condition is searched optimally level by level, the total number of states is reduced, the flow with monitoring analog file is intelligently screened out, and a five-level depth-first tree is established. For example, table 3, the test data of the specific test flow is traversed and searched.

[0042] Then use the self-identification reduction module 2 of the data processing layer, as shown in table 4, two-level reduction is performed on the data file (.txt) in the fifth-level depth-first tree: the "platform analog test result" file is identified from all data files by using the "monitoring analog file library", which completes the first-level reduction dimensionality reduction. The "monitoring analog file library" encapsulates the keyword for identifying the "platform analog test result" file; then the identified "platform analog test result" data file is matched with the "monitoring analog data format library" to obtain the platform analog test result data processing format in different test flows, so as to complete the second-level reduction. The "monitoring analog data format library" encapsulates the platform analog data processing format of different test flows.

[0043] Finally, the batch processing and report module 3 of the data processing layer is applied, all the test flow monitoring analog data traversed and reduced are processed in series in batches according to the specified processing format, and are sorted and summarized according to the set arrangement rule, forming a template data report for presentation, as shown in table 5.

[0044] Table 3 Self-traversal depth-first tree of monitoring analog statistics module

[0045]

[0046] Table 4 Monitoring analog quantity statistics module self-identification reduction process

[0047]

[0048] Table 5 Monitoring analog quantity batch processing compilation report

[0049]

[0050]

[0051] 1.3) Self-aiming result statistics module 13: Self-aiming test will be conducted during the inertial platform precision test process, and two self-aiming results (α gyroscope and β gyroscope) will be obtained. According to the corresponding folder directory, storage is conducted.

[0052] The self-aiming result statistics module 13 establishes a "self-aiming process library", which encapsulates the test process name that needs to conduct self-aiming result statistics. Through the test data self-traversal module 1, the inertial platform test process is automatically retrieved. In a manner similar to that in the monitoring analog quantity statistics, the processes with self-aiming files are intelligently screened out, forming a five-level depth-first tree, as shown in Table 6. Then, combined with the self-identification reduction module 2, as shown in Table 7, the data files (.txt) in the fifth level of the depth-first tree in Table 6 are reduced by two levels: the "platform attitude acceleration data" file and the "self-aiming result" file are identified from all data files by using the "self-aiming file library", completing the first level of reduction dimension. The "self-aiming file library" encapsulates the keywords for identifying the "platform attitude acceleration data" file and the "self-aiming result" file; the identified data files are matched with the "self-aiming type library" to obtain the self-aiming data processing formats in different test processes, which are respectively used for the subsequent statistics of the α gyroscope and β gyroscope self-aiming results, thereby completing the second level of reduction. The "self-aiming type library" encapsulates the self-aiming data processing formats of different test processes.

[0053] Finally, based on the batch processing compilation report module 3, the self-aiming results of the two gyroscopes are respectively counted: as shown in Table 8, according to the self-aiming type of different processes, the β gyroscope self-aiming results and the gravity acceleration value during self-aiming are counted from the "platform attitude acceleration data" file according to the corresponding data processing format; as shown in Table 9, according to the self-aiming type of different processes, the α gyroscope self-aiming results are extracted from the "self-aiming result" file according to the corresponding data processing format. The self-aiming results of the two gyroscopes in each related test process are arranged in the same table, the compilation report is completed, and is shown in Table 10.

[0054] Table 6 Self-aiming result statistics module self-traversal depth-first tree

[0055]

[0056]

[0057] Table 7 Self-identified reduction process of self-alignment results statistics module

[0058]

[0059] Table 8 Self-alignment results file of beta gyro

[0060] Serial number Time Data type Gravity acceleration value Self-aiming result 0 21:01.6 Normal 9.78452 0.268003 1 21:01.6 Normal 9.78452 0.268003

[0061] Table 9 Self-alignment results file of alpha gyro

[0062]

[0063] Table 10 Self-alignment results compilation report

[0064]

[0065]

[0066] 1.4) Self-calibration results statistics module 14: Only a few processes in the inertial platform precision test process will be tested for self-calibration. The self-calibration results include more than 20 gyro and accelerometer error parameters. All self-calibration error parameters are stored in the form of result files according to the corresponding folder directory.

[0067] The self-calibration results statistics module 14 establishes a "self-calibration process library", which encapsulates the test process names that need to be statistically analyzed for self-calibration results, and automatically retrieves the inertial platform test process through the test data self-traversal module 1 of the data processing layer. The process with self-calibration files is intelligently filtered out to form a five-level depth-first tree, as shown in Table 11.

[0068] Then combined with the self-identified reduction module 2, as shown in Table 12, the data files (.txt) in the fifth level of the depth-first tree in Table 11 are reduced by two levels: the "self-calibration file library" is used to identify the "self-calibration result" file from all data files, completing the first reduction dimension. The "self-calibration file library" encapsulates the keywords for identifying the "self-calibration result" file. The identified "self-calibration result" data file is matched with the "self-calibration type library" to obtain the self-calibration data processing format in different test processes, which is used for subsequent result statistics, thereby completing the second reduction. The "self-calibration type library" encapsulates the self-calibration data processing format of different test processes.

[0069] Finally, the batch processing and report compiling module 3 is applied to statistically summarize the self-calibration results of each data file according to the corresponding data processing format, read the unit data burned when the inertial platform system is shipped, and complete synchronization with the self-calibration results of the current test process, thereby completing the compilation and statistical report of all test process self-calibration results, as shown in Table 13.

[0070] Table 11 Self-calibration result statistical module self-traversal depth-first tree

[0071]

[0072] Table 12 Self-calibration result statistical module self-identification reduction process

[0073]

[0074]

[0075] Table 13 Self-calibration result compilation report

[0076]

[0077]

[0078] 1.5) Instruction execution time statistical module 15: executed after the inertial platform receives a control instruction, and the state is replied after the execution is completed. The instruction time and instruction name sent by the system are stored in the determination file under the folder directory of the corresponding process, and the state replied by the inertial platform is stored in the state information data file under the same folder directory.

[0079] The instruction execution time statistical module 15 is the statistics of the inertial platform instruction and its execution time. First, the "instruction execution time process library" is established, which encapsulates the test process name that needs to perform the inertial platform instruction and execution time statistics; and the test data self-traversal module 1 of the data application layer is continuously called to automatically search the inertial platform test process, intelligently filter out the processes with inertial platform instruction and execution time data files, and form a five-level depth-first tree, as shown in Table 14.

[0080] Then, the data files (.txt) in the fifth level of the depth-first tree in Table 14 are reduced by two levels in combination with the self-identification reduction module 2 as shown in Table 15: three files of "platform control command ①", "platform control command ②" and "platform state test result" are identified from all the data files by using the "instruction execution time file library" to complete the first level of reduction, and the "instruction execution time file library" encapsulates the keywords for identifying the three files; the three identified files are matched with the "instruction execution time type library" to obtain the processing format of the platform instructions and their execution time data in different test processes, which are used for subsequent data processing, so as to complete the second level of reduction, and the "instruction execution time type library" encapsulates the processing format of the platform instructions and their execution time data in different test processes.

[0081] Finally, the batch processing and report compiling module 3 is applied to process the data of the three files of "platform control command ①", "platform control command ②" and "platform state test result" after two levels of reduction according to the corresponding data processing format: the instruction name and the starting time (ms) of the inertial platform received in the test process are obtained from the two files of "platform control command ①" and "platform control command ②", and the ending time (ms) of the corresponding instruction is retrieved from the "platform state test result" file; then, the execution time of the instruction is calculated, converted into seconds, and matched with the template criterion of different instruction execution times to form the instruction execution time statistical result. The instruction execution time statistical results of each data file are summarized to complete the compilation and report of the instruction execution time statistical results of all the test processes as shown in Table 16.

[0082] Table 14 Instruction execution time statistical module self-traverses the depth-first tree

[0083]

[0084] Table 15 Instruction execution time statistical module self-identification reduction process

[0085]

[0086]

[0087] Table 16 Instruction execution time statistical compilation and report

[0088]

[0089] 1.6) Positioning error calculation module 16: The inertial platform has a ground navigation function, and in a specific test process, ground navigation test needs to be performed to evaluate the related use performance of the inertial platform.

[0090] The positioning error calculation module 16 is a statistical calculation of the ground navigation process data; first, a "positioning error calculation flow library" is established, which encapsulates the test flow names that need to be calculated for positioning error, and automatically retrieves the inertial platform test flow through the test data self-traversal module 1 of the data processing layer, intelligently selects the flow that has the platform ground navigation result file, forms a five-level depth-first tree, see Table 17.

[0091] Then combined with the self-identification reduction module 2, as shown in Table 18, the data files (.txt) in the fifth level of the depth-first tree in Table 17 are reduced by two levels: the "platform ground navigation result" file is identified from all data files using the "positioning error calculation file library", which completes the first level of reduction dimensionality, and the "positioning error calculation file library" encapsulates the keywords for identifying the "platform ground navigation result" file; the identified "platform ground navigation result" data file is matched with the "positioning error calculation format library" to obtain the processing format of the ground navigation process data in different test flows, which is used for subsequent data processing, thereby completing the second level of reduction, and the "positioning error calculation format library" encapsulates the positioning data processing format of different test flows.

[0092] Finally, the batch processing and report module 3 is applied to process the "platform ground navigation result" file data after two levels of reduction according to the corresponding data processing format, extract the time, longitude, latitude and altitude at the start of ground navigation and the time, longitude, latitude and altitude at the end of ground navigation, and determine the inertial platform navigation positioning error based on the earth parameter model; the positioning error calculation results of each depth-first tree are summarized, and the report of the navigation positioning error calculation results of all ground navigation test flows is completed, as shown in Table 19.

[0093] Table 17 Positioning error calculation module self-traversal depth-first tree

[0094]

[0095] Table 18 Positioning error calculation module self-identification reduction flow

[0096]

[0097] Table 19 Positioning error calculation module report

[0098]

[0099] 1.7) Starlight data statistics module 17: The inertial platform has a starlight combination navigation function, and through the starlight test information in a dark environment during ground testing, the condition of this part of the function is determined.

[0100] In the process of inertial platform test, starlight test will be carried out. First, a "starlight data flow library" is established, which encapsulates the test flow names that need to be statistically analyzed. The test data self-traversal module 1 of the data processing layer automatically retrieves the inertial platform test flow, intelligently filters the flow with starlight data files, and forms a five-level depth-first tree, taking Table 20 as an example.

[0101] Then, combined with the self-identification reduction module 2, as shown in Table 21, the data files (.txt) in the fifth level of the depth-first tree in Table 20 are reduced by two levels: the "starlight data statistical file library" identifies the "starlight test result ①" and "starlight test result ②" files from all data files, completing the first level of reduction. The "starlight data statistical file library" encapsulates the keywords for identifying the above two files. The two identified files are matched with the "starlight data statistical format library" to obtain the processing format of starlight data in different test flows for subsequent data processing, thereby completing the second level of reduction. The "starlight data statistical format library" encapsulates the starlight data processing format of different test flows.

[0102] Finally, the batch processing and report module 3 is applied to extract the key information such as data frame number, image mean, variance, maximum value, and local variance from the two files "starlight test result ①" and "starlight test result ②" after two levels of reduction. The starlight data statistical results of each depth-first tree are summarized, and the report of all starlight test statistical results is completed, as shown in Table 22.

[0103] Table 20 Starlight data statistical module self-traversal depth-first tree

[0104]

[0105] Table 21 Starlight data statistical module self-identification reduction flow

[0106]

[0107]

[0108] Table 22 Starlight data statistical module report

[0109] Date Test procedure Data frame number Mean value Variance Maximum value Local variance 20210117 Test procedure III 1-10 10 5 13 4 20210118 Test procedure IV 1-10 12 6 15 4 20210120 Test procedure V 1-10 11 5 14 4

[0110] Due to the diversity of inertial platform test types and the variability of test processes, the system often processes data in abnormal states, such as data type mismatch, data exceeding the system's predetermined range, file nonexistence, and human abnormal operation. The presence of abnormal factors will lead to the absence of key information input during system operation, and the system cannot normally implement its functions. Therefore, an abnormal data self-diagnosis processing module 4 is set in the system design, which encapsulates standardized data preprocessing criteria. By identifying specific processes and data key characteristics, the module automatically diagnoses abnormal characteristics in test data, and based on the set processing criteria, it performs data fault-tolerant processing, timely feedbacks non-normal input in the system, and reasonably processes non-critical factors, ensuring the intelligence and robustness of system operation.

[0111] Taking the implementation process of the "monitoring analog quantity statistical module 12" as an example, after completing two-level reduction, if there are abnormal data in some files during data extraction based on the format specifications in the "monitoring analog quantity data format library", the system will perform adaptive processing based on the preset fault-tolerant processing criteria, and will output the problems such as "data format error, test process abnormality, and statistical time mismatch" for classification diagnosis. After completion, it automatically jumps to the subsequent process and does not interrupt the processing due to process and data abnormalities.

[0112] In each functional module of the entire inertial platform test process and result analysis application unit 10, the abnormal data self-diagnosis processing module 4 performs fault-tolerant processing on the inertial platform test data, which is also the main embodiment of the system's self-supervision feature, and conforms to the application feature of "input and do not care".

[0113] (2) Inertial platform algorithm verification and precision analysis unit 20:

[0114] This application unit is the core embodiment of the system's deep processing and analysis capability, and is the concentrated embodiment of the inertial platform's professional data analysis capability. First, the test data self-traversal module 1 and the self-identification reduction module 2 of the data processing layer are used to complete the filtering of test data. Then, based on the modular algorithm adaptive matching module 5 of the data processing layer (including four algorithm submodules: slope adjustment precision calculation algorithm submodule, self-aiming precision calculation algorithm submodule, self-calibration precision calculation algorithm submodule, and flight navigation precision calculation algorithm submodule), each system parameter in the test data file and the modular algorithm unit are automatically matched to realize the offline verification of the inertial platform's key algorithm and the calculation and analysis of each precision test. Finally, the batch processing and report generation module 3 is used to realize the compilation and report generation of analysis data. This unit includes four functional modules, which are as follows:

[0115] 2.1) Inclination leveling accuracy calculation module 21: the inertial platform establishes an inertial reference after inclination leveling, so the inclination leveling accuracy is crucial. During the inclination leveling test, the inertial platform receives control commands (two command values of pitch angle and azimuth angle), and the frame is rotated to the target position based on the two angles. The inclination leveling accuracy calculation mainly includes the comparison of theoretical value and actual value.

[0116] The inclination leveling accuracy calculation module 21 establishes an "inclination leveling accuracy calculation process library", which encapsulates the test process names that need to perform inclination leveling accuracy calculation. First, the test data self-traversal module 1 of the data processing layer automatically retrieves the inertial platform test process, intelligently filters the processes that have performed inclination leveling accuracy test, forms a five-level depth-first tree, and takes table 23 as an example.

[0117] Then, combined with the self-identification reduction module 2, as shown in table 24, the data files (.txt) in the fifth level of the depth-first tree in table 23 are reduced by two levels: the "platform control command ①" file and the "platform attitude acceleration data" file are identified from all data files by using the "inclination leveling accuracy calculation file library", which completes the first level of reduction dimension. The "inclination leveling accuracy calculation file library" encapsulates the keywords for identifying the above two files; by matching the data file with the "inclination leveling accuracy calculation format library", the inclination leveling data processing format in different test processes is obtained, which is used for subsequent inclination leveling accuracy calculation, thereby completing the second level of reduction. The "inclination leveling accuracy calculation format library" encapsulates the inclination leveling accuracy data processing format of different test processes.

[0118] After that, the modular algorithm adaptive matching module 5 is applied. The modular algorithm in the inclination leveling accuracy calculation includes two parts of theoretical value calculation and actual value calculation, wherein the theoretical value calculation includes three sub-parts of self-calibration result reading, control command angle acquisition and command current calculation, and the actual value calculation includes four sub-parts of frame angle calculation, command current calculation, platform attitude solution and acceleration vector fitting. The main functions of each sub-part are shown in table 25. Through the inclination leveling accuracy calculation algorithm sub-module of the modular algorithm adaptive matching module 5, according to the corresponding data processing format, the modular algorithm for inclination leveling accuracy calculation is adaptively matched and loaded for different test processes when performing inclination leveling accuracy calculation, and the inclination leveling accuracy related calculation is completed.

[0119] Finally, through the batch processing and report generation module 3, the batch calculation of the inclination leveling accuracy of multiple processes is realized, and the calculation results are compiled and stored in the report. Each inclination leveling test will generate a compiled report, as shown in table 26.

[0120] Table 23: Inclination leveling accuracy calculation module self-traversal depth-first tree

[0121]

[0122] Table 24 Self-identification reduction process of slope leveling accuracy calculation module

[0123]

[0124] Table 25 Slope leveling accuracy calculation algorithm submodule

[0125]

[0126]

[0127] Table 26 Slope leveling accuracy calculation module compilation report

[0128]

[0129] 2.2) Self-aiming accuracy calculation module 22: In 1.3), the inertial platform self-aiming result statistics is mainly the statistics and arrangement of the self-aiming results, and the self-aiming accuracy calculation focuses on the algorithm verification and accuracy analysis of the process data of the key inertial information.

[0130] The self-aiming accuracy calculation module 22 establishes a "self-aiming accuracy calculation process library", which encapsulates the test process names that need to perform self-aiming accuracy calculation. First, the test data self-traversal module 1 of the data processing layer automatically retrieves the inertial platform test process, intelligently selects the process that has performed self-aiming accuracy test, forms a five-level depth-first tree, and takes Table 27 as an example.

[0131] Then, combined with the self-identification reduction module 2, as shown in Table 28, the data files (.txt) in the fifth level of the depth-first tree in Table 27 are reduced by two levels: the "self-aiming accuracy calculation file library" identifies the "platform attitude acceleration data" file from all data files, completing the first level of reduction dimension, and the "self-aiming accuracy calculation file library" encapsulates the keywords for identifying the "platform attitude acceleration data" file; by matching the data file with the "self-aiming accuracy calculation format library", the self-aiming data processing format in different test processes is obtained, which is used for subsequent self-aiming accuracy calculation, thereby completing the second level of reduction, and the "self-aiming accuracy calculation format library" encapsulates the self-aiming accuracy data processing format in different test processes.

[0132] Then, the self-aiming accuracy calculation algorithm submodule of the self-aiming accuracy calculation modular algorithm adaptive matching module 5 is applied, as shown in Table 29, according to the corresponding data processing format, the specific command current data is adaptively extracted for different test processes, the mean and dispersion of the command current at each position of the two gyroscopes are respectively calculated, and the encapsulated inertial platform control software self-aiming algorithm is adaptively called to complete the offline recalculation verification of the self-aiming results of the a gyroscope and the b gyroscope.

[0133] Finally, the self-calibration accuracy of multiple processes is calculated in batches through the batch processing and report compiling module 3, and the calculation results are compiled and summarized into reports. The self-calibration accuracy compilation table of each process is shown in Table 30.

[0134] Table 27 Self-iteration depth-first tree of self-calibration accuracy calculation module

[0135]

[0136] Table 28 Self-identified reduction process of self-calibration accuracy calculation module

[0137]

[0138] Table 29 Self-calibration accuracy calculation algorithm sub-module

[0139]

[0140] Table 30 Self-calibration accuracy calculation module report compilation

[0141]

[0142] 2.3) Self-calibration accuracy calculation module 23: Inertial platform self-calibration accuracy calculation is the analysis of process data of the statistical results of inertial platform self-calibration in 1.4), and scientific evaluation of self-calibration results is crucial.

[0143] The self-calibration accuracy calculation module 23 establishes a "self-calibration accuracy calculation process library", which encapsulates the test process names that need to perform self-calibration accuracy calculation. First, the test data self-iteration module 1 of the data processing layer automatically retrieves the inertial platform test process, intelligently filters the processes that have performed self-calibration accuracy test, forms a five-level depth-first tree, and takes Table 31 as an example.

[0144] Then, combined with the self-identified reduction module 2, as shown in Table 32, the data files (.txt) in the fifth level depth-first tree in Table 31 are reduced by two levels: the "self-calibration accuracy calculation file library" identifies the "platform attitude acceleration data" file from all data files, completes the first level reduction dimension, and the "self-calibration accuracy calculation file library" encapsulates the keywords for identifying the "platform attitude acceleration data" file; through matching the data file with the "self-calibration accuracy calculation format library", the self-calibration data processing format in different test processes is obtained, which is used for subsequent self-calibration accuracy calculation, thereby completing the second level reduction, and the "self-calibration accuracy calculation format library" encapsulates the self-calibration accuracy data processing format in different test processes.

[0145] After the application of self-calibration precision calculation modular algorithm self-adapting matching module 5, the self-calibration precision calculation algorithm sub-module is applied, as shown in Table 33, according to the corresponding data processing format, the mean and dispersion of the three gyro instruction currents at each position of the self-calibration are calculated by extracting the effective segment instruction current data for the self-calibration related test process.

[0146] Finally, the batch processing and report compiling module 3 is used to realize the batch mathematical statistics of the precision of multiple self-calibration process data, and the calculation results are compiled and reported. The self-calibration precision compilation table of each process is shown in Table 34.

[0147] Table 31 Self-calibration precision calculation module self-traversal depth-first tree

[0148]

[0149] Table 32 Self-calibration precision calculation module self-identification reduction process

[0150]

[0151] Table 33 Self-calibration precision calculation algorithm sub-module

[0152]

[0153] Table 34 Self-calibration precision calculation module compilation and reporting

[0154]

[0155] 2.4) Flight navigation precision calculation module 24: Flight navigation is the most basic function of the inertial platform. During flight navigation, the inertial platform body is stabilized in the inertial space. The flight navigation precision of the inertial platform during the simulation flight stage is the most direct manifestation of the precision of the inertial platform.

[0156] The flight navigation precision calculation module 24 establishes a "flight navigation precision calculation process library", which encapsulates the test process names that need to perform self-aiming precision calculation. First, the test data self-traversal module 1 of the data processing layer automatically retrieves the inertial platform test process, intelligently selects the process that has performed flight navigation precision test, and forms a five-level depth-first tree, as shown in Table 35.

[0157] Then, combined with the self-identification reduction module 2, as shown in Table 36, the data files (.txt) in the fifth level of the depth-first tree in Table 35 are reduced by two levels: the "flight navigation accuracy calculation file library" identifies the "platform attitude acceleration data", "platform tool error ①" and "platform tool error ②" from all data files, completing the first level of reduction dimension, and the "flight navigation accuracy calculation file library" encapsulates the keywords for identifying the above three files; respectively, the three data files are matched with the "flight navigation accuracy calculation format library" to obtain the flight navigation data processing format in different test processes, which is used for subsequent flight navigation accuracy calculation, thereby completing the second level of reduction, and the "flight navigation accuracy calculation format library" encapsulates the flight navigation accuracy data processing format in different test processes.

[0158] Then, the modular algorithm is applied to adaptively match the flight navigation accuracy calculation algorithm submodule of the module 5, as shown in Table 37, for the test process with a simulation flight process, the inertial platform tool error parameters in the process are read, and the frame angle drift rate and acceleration fitting value in the flight navigation [t0, t1] time period are obtained by extracting and calculating the frame angle and accelerometer data in the simulation flight stage.

[0159] Finally, through the batch processing and report compiling module 3, batch data extraction and calculation of multiple sets of flight navigation data accuracy are realized, and the final results are compiled and reported, as shown in Table 38.

[0160] Table 35 Flight navigation accuracy calculation module self-traversal depth-first tree

[0161]

[0162] Table 36 Flight navigation accuracy calculation module self-identification reduction process

[0163]

[0164]

[0165] Table 37 Flight navigation accuracy calculation algorithm submodule

[0166]

[0167] Table 38 Flight navigation accuracy calculation module compilation report

[0168]

[0169] In addition, the abnormal data self-diagnosis processing module 4 is also involved in the operation of each module, and especially for the test process human interruption or fault simulation test data, the fault tolerance processing of the inertial platform test data can be realized through the module, and the stable and smooth operation of the whole system is ensured.

[0170] (3) Inertial information autonomous visualization display unit 30:

[0171] The application unit mainly extracts, calculates and displays the typical and key inertial information data in the inertial platform test, and finally realizes the normalized visualization display of different types of graphical curves through the comprehensive use of the feature data autonomous visualization module 6 and the modular algorithm adaptive matching module 5.

[0172] The feature data autonomous visualization module 6 mainly includes feature files and inertial information data extraction, feature inertial information data calculation, graphical interface configuration, and visualization structure matching sub-module. The main functions are shown in Table 39.

[0173] Table 39 Feature data autonomous visualization module

[0174]

[0175] The inertial information autonomous visualization display unit 30 contains seven functional modules, which are as follows:

[0176] 3.1) Instruction current display module 31: The instruction current is the most direct reflection of the inertial platform leveling locking process and leveling locking accuracy. The module extracts the relevant inertial information of the instruction current through the configured feature data autonomous visualization module 6, and implements the autonomous drawing and curve display of the three gyro instruction currents in the whole test process.

[0177] 3.2) Frame angle display module 32: The frame angle is a direct reflection of the inertial platform attitude. The change curve analysis of the inertial platform frame angle in the test process can accurately identify the rotation of the inertial platform table body or base, and master the working state of the inertial platform. The module extracts the relevant inertial information of the frame angle through the configured feature data autonomous visualization module 6, especially the relevant inertial information of the frame angle in the typical test process involving the rotation of the inertial platform base, and implements the autonomous drawing and joint comparison of the four frame angle curves.

[0178] 3.3) Accelerometer output display module 33: During the test process, the accelerometer output of the inertial platform will change with the change of the table body attitude. The module extracts the accelerometer inertial information through the feature data autonomous visualization module 6, and implements the autonomous drawing and curve display of the three accelerometer outputs in the whole test process.

[0179] 3.4) Moment motor current display module 34: the inertial platform moment motor current is a direct reflection of the shaft end disturbance moment condition, this module effectively extracts the self-monitoring analog quantity information through the characteristic data autonomous visualization module 6, and implements the autonomous drawing and curve display of the four moment motor currents in the whole test process.

[0180] 3.5) Temperature and temperature control display module 35: the inertial platform provides a stable temperature field environment for the inertial instrument through temperature control, this module effectively extracts the self-monitoring analog quantity information through the characteristic data autonomous visualization module 6, and implements the autonomous drawing and curve display of the inertial platform temperature and temperature control power level current in the whole test process.

[0181] 3.6) Mode flight process display module 36: mode flight test is often needed in the inertial platform test process, and the mode flight state and accuracy are judged through the frame angle and accelerometer curve and data. On the one hand, this module applies the characteristic data autonomous visualization module 6 to effectively extract the inertial information, realize the autonomous drawing and curve display of the inertial platform mode flight process attitude and acceleration; on the other hand, it applies the flight navigation accuracy calculation algorithm sub-module in the modular algorithm adaptive matching module 5, and according to the platform tool error related parameters (platform tool error ① and platform tool error ② in table 36), and the corresponding characteristic data are comprehensively matched, to realize the accuracy calculation of the mode flight data.

[0182] 3.7) Telemetry data display module 37: the telemetry data curve mainly includes two parts of inertial measurement information visualization and self-monitoring analog quantity data visualization. This module applies the characteristic data autonomous visualization module 6 to effectively analyze and extract the telemetry inertial information and self-monitoring analog quantity data, and implements the autonomous drawing and curve display of the inertial platform attitude, acceleration and electrical quantity parameters in the telemetry data.

[0183] The inertial information autonomous visualization display unit visualizes the inertial information as shown in Figure 4 .

[0184] The above has been described in detail in combination with the specific embodiments and exemplary examples, but these descriptions cannot be understood as limiting the present application. Those skilled in the art understand that the technical solutions and their embodiments of the present application can be variously replaced, modified or improved without departing from the spirit and scope of the present application, and these all fall within the scope of the present application. The protection scope of the present application is subject to the appended claims.

[0185] The contents not described in detail in the specification of the present application are the known technology of those skilled in the art.

Claims

1. A self-supervised deep processing and analysis system for inertial platform test data, characterized in that, include: Application layer: The test process and result analysis unit (10) is used to perform a complete traversal of all test processes and a comprehensive retrieval of test data through the test data self-traversal module (1); to complete the screening, identification and autonomous matching of specific test types and data through the self-identification and reduction module (2), and to process and analyze them respectively; and to implement the templated compilation of analysis data reports through the batch processing and compilation report module (3). The algorithm verification and accuracy analysis unit (20) is used to complete the screening of test data through the test data self-traversal module (1) and the self-identification and reduction module (2); through the modular algorithm adaptive matching module (5), it automatically matches various system parameters and modular algorithm units in the test data file to implement offline verification of key algorithms of the inertial platform and calculation and analysis of various accuracy tests; and through the batch processing and compilation report module (3), it compiles and reports the analysis data. The inertial information visualization unit (30) is used to extract, perform mathematical calculations and display the target inertial information data in the inertial platform test. Through the feature data autonomous visualization module (6) and the modular algorithm adaptive matching module (5), it implements the normalized visualization display of different types of graphic curves. Data processing layer: The test data self-traversal module (1) is used to perform a complete traversal retrieval of the inertial platform test process and data through the breadth-first search algorithm and the depth-first search algorithm; The self-identification and reduction module (2) is used to reduce the dimensionality of the test process by matching the consistency of the specific test process of the inertial platform, so as to realize the screening and identification of the specific test process. Batch processing and compilation report module (3) is used for statistical analysis, classification and summary of experimental procedures and data and result organization, and implements classification and templated reports for different types of data; The abnormal data self-diagnosis processing module (4) is used to automatically diagnose abnormal data in the experiment according to the set standardized data preprocessing criteria and perform fault-tolerant processing. Modular algorithm adaptive matching module (5) is used to automatically match various system parameters and modular algorithms in the test data, implement offline verification of key algorithms of inertial platform and calculation and analysis of various accuracy tests; The feature data autonomous visualization module (6) is used to adaptively extract inertial information data in inertial platform tests based on the assessment requirements of different test procedures, and to complete the autonomous visualization display of inertial platform feature data.

2. The self-supervised deep processing and analysis system for inertial platform test data according to claim 1, characterized in that, The test process and result analysis unit (10) includes a test process statistics module (11), which calls the test data self-traversal module (1) of the application data processing layer. Based on the breadth-first search algorithm, it performs a global search on all folder directories under the selected path and searches level by level according to the storage format of the current test process data, and implements automatic retrieval and global traversal of all test processes of the selected product.

3. The self-supervised deep processing and analysis system for inertial platform test data according to claim 2, characterized in that, The test process and result analysis unit (10) includes a monitoring simulation quantity statistics module (12), which is used to establish a monitoring simulation quantity process library. The monitoring simulation quantity process library encapsulates the test process names that need to be statistically analyzed. The test data self-traversal module (1) is called. Based on the width search, the depth-first search algorithm is used. Based on the monitoring simulation quantity process library, the specific process in the power-on test of the inertial platform is defined as the backtracking boundary condition, and the process threads with monitoring simulation quantity files are selected. Call the self-identification and reduction module (2) to perform two-level reduction on the data files in the process where there are monitoring simulation files: use the monitoring simulation file library to identify the platform simulation test result file from all data files to complete the first-level reduction and dimensionality reduction. The monitoring simulation file library contains keywords for identifying the platform simulation test result file. Then, the identified platform analog quantity test result data file is matched with the monitoring analog quantity data format library to obtain the platform analog quantity test result data processing format in different test processes, thereby completing the second-level reduction. The monitoring analog quantity data format library encapsulates the platform analog quantity data processing format for different test processes. Call the batch processing and compilation report module (3), serially batch process all the test process monitoring simulation data of the traversal and reduction according to the specified processing format, and sort, summarize and compile according to the set arrangement rules to form a templated data report for presentation.

4. The self-supervised deep processing and analysis system for inertial platform test data according to claim 2, characterized in that, The test process and result analysis unit (10) includes a self-aiming result statistics module (13), which is used to establish a self-aiming process library. The self-aiming process library contains the names of test processes that need to be statistically analyzed for self-aiming results. The test data self-traversal module (1) automatically retrieves the inertial platform test process and filters out the processes that have self-aiming files. The self-identification and reduction module (2) is called to perform two-level reduction on the data files in the process where self-aiming files exist: the self-aiming file library is used to identify the platform attitude acceleration data files and self-aiming result files from all data files to complete the first-level reduction and dimensionality reduction. The self-aiming file library encapsulates the keywords for identifying platform attitude acceleration data files and self-aiming result files. The identified data files are matched with the self-aiming type library to obtain the self-aiming data processing format in different test processes, which is used for subsequent statistics on the self-aiming results of α gyroscope and β gyroscope, thereby completing the second-level reduction. The self-aiming type library encapsulates the self-aiming data processing format for different test processes. Call the batch processing and compilation report module (3) to count the self-aiming results of the two gyroscopes respectively: according to the different self-aiming types of the process, according to the corresponding data processing format, count the self-aiming results of the β gyroscope and the gravitational acceleration value during self-aiming from the platform attitude acceleration data file, extract the self-aiming results of the α gyroscope from the self-aiming result file according to the corresponding data processing format, summarize and arrange the self-aiming results of the two gyroscopes in each relevant test process in the same form, and complete the compilation report.

5. The self-supervised deep processing and analysis system for inertial platform test data according to claim 2, characterized in that, The test process and result analysis unit (10) includes a self-calibration result statistics module (14), which is used to establish a self-calibration process library. The self-calibration process library contains the names of test processes that need to be statistically analyzed for self-calibration results. The test data self-traversal module (1) automatically retrieves the inertial platform test process and filters out the processes that have self-calibration files. The self-identification and reduction module (2) is called to perform two-level reduction on the data files in the process where self-calibration files exist: the self-calibration result files are identified from all data files using the self-calibration file library to complete the first-level reduction and dimensionality reduction. The self-calibration file library contains keywords for identifying self-calibration result files. The identified self-calibration result data files are matched with the self-calibration type library to obtain the self-calibration data processing format in different test processes, thereby completing the second-level reduction. The self-calibration type library contains the self-calibration data processing format for different test processes. Call the batch processing compilation report module (3), and summarize the self-calibration results of each data file according to the corresponding data processing format. At the same time, read the unit data burned by the inertial platform system when it leaves the factory, and synchronize it with the self-calibration results of the current test process to complete the compilation and statistical report of the self-calibration results of all test processes.

6. The self-supervised deep processing and analysis system for inertial platform test data according to claim 2, characterized in that, The test process and result analysis unit (10) includes an instruction execution time statistics module (15), which is used to establish an instruction execution time process library. The instruction execution time process library encapsulates the test process names that need to be statistically analyzed for inertial platform instructions and execution time. It also calls the test data self-traversal module (1) to automatically search for inertial platform test processes and filter out processes that contain inertial platform instruction and execution time data files. The self-identification and reduction module (2) is called to perform two-level reduction on the data files in the process containing inertial platform instructions and execution time data files: the platform control command ①, platform control command ② and platform status test results are identified from all data files using the instruction execution time file library to complete the first-level reduction and dimensionality reduction. The instruction execution time file library contains keywords for identifying the above three files. The three identified files are matched with the instruction execution time type library to obtain the processing format of platform instructions and their execution time data in different test processes, thereby completing the second-level reduction. The instruction execution time type library contains the processing format of platform instructions and execution time data in different test processes. Call the batch processing and compilation report module (3) to process the three files of platform control command ①, platform control command ② and platform status test results after two-level reduction according to the corresponding data processing format: obtain the instruction name and start time received by the inertial platform in the test process from the two files of platform control command ① and platform control command ②, retrieve the end time of the corresponding instruction from the platform status test results file, calculate the instruction execution time, and match the templated criteria of different instruction execution times to form the instruction execution time statistics; summarize the instruction execution time statistics of each data file to complete the compilation report of the instruction execution time statistics of all test processes.

7. The self-supervised deep processing and analysis system for inertial platform test data according to claim 2, characterized in that, The test process and result analysis unit (10) includes a positioning error calculation module (16), which is used to establish a positioning error calculation process library. The positioning error calculation process library encapsulates the test process names that need to be calculated for positioning error. The test data self-traversal module (1) is called to automatically search for inertial platform test processes and filter out processes that have platform ground navigation result files. Call the self-identification and reduction module (2) to perform two-level reduction on the data files in the platform ground navigation result files: use the positioning error calculation file library to identify the platform ground navigation result files from all data files, and complete the first-level reduction and dimensionality reduction. The positioning error calculation file library contains keywords for identifying platform ground navigation result files. The identified platform ground navigation result data file is matched with the positioning error calculation format library to obtain the processing format of ground navigation process data in different test procedures, thereby completing the second-level reduction. The positioning error calculation format library encapsulates the positioning data processing format of different test procedures. Call the batch processing and compilation report module (3) to process the platform ground navigation result file data after two-level reduction according to the corresponding data processing format, extract the time, longitude, latitude and altitude at the start of ground navigation and the time, longitude, latitude and altitude at the end of ground navigation, and determine the inertial platform navigation positioning error based on the Earth parameter model; summarize the positioning error calculation results of each depth priority tree, and complete the compilation report of the navigation positioning error calculation results of all ground navigation test processes.

8. The self-supervised deep processing and analysis system for inertial platform test data according to claim 2, characterized in that, The test process and result analysis unit (10) includes a starlight data statistics module (17), which is used to establish a starlight data process library. The starlight data process library encapsulates the test process names that need to be statistically analyzed by starlight data, and calls the test data self-traversal module (1) to automatically search the inertial platform test process and filter out the processes that have starlight data files. Call the self-identification and reduction module (2) to perform two-level reduction on the data files in the process containing starlight data files: use the starlight data statistics file library to identify the two files, starlight test result ① and starlight test result ②, from all data files to complete the first-level reduction and dimensionality reduction. The starlight data statistics file library contains keywords for identifying the above two files. The two identified files are matched with the starlight data statistical format library to obtain the starlight data processing format in different test procedures, thereby completing the second-level reduction. The starlight data statistical format library encapsulates the starlight data processing format for different test procedures. Call the batch processing and compilation report module (3), extract key information from the two files of starlight test results ① and starlight test results ② of the inertial platform after two-level reduction according to the corresponding data processing format; summarize the starlight data statistics results of each depth-first tree, and complete the compilation report of starlight data statistics results in all starlight tests.

9. The self-supervised deep processing and analysis system for inertial platform test data according to claim 1, characterized in that, The algorithm verification and accuracy analysis unit (20) includes a slant leveling accuracy calculation module (21), which is used to establish a slant leveling accuracy calculation process library. The slant leveling accuracy calculation process library encapsulates the names of the test processes that need to be calculated for slant leveling accuracy. The test data self-traversal module (1) is called to automatically retrieve the inertial platform test process and filter out the process that has undergone oblique leveling accuracy test; The self-identification reduction module (2) is called to perform two-level reduction on the data files in the process of oblique leveling accuracy test: the oblique leveling accuracy calculation file library is used to identify the platform control command ① file and the platform attitude acceleration data file from all data files to complete the first-level reduction and dimensionality reduction. The oblique leveling accuracy calculation file library contains keywords for identifying the above two files. The data file is matched with the oblique leveling accuracy calculation format library to obtain the oblique leveling data processing format in different test processes, which are used for subsequent oblique leveling accuracy calculation to complete the second-level reduction. The oblique leveling accuracy calculation format library contains oblique leveling accuracy data processing formats for different test processes. The modular algorithm adaptive matching module (5) is called. According to the corresponding data processing format, for different experimental procedures, when performing the slant leveling accuracy calculation, the modular algorithm for slant leveling accuracy calculation is adaptively matched and loaded to complete the slant leveling accuracy related calculation. Finally, the batch processing and report compilation module (3) is used to realize the batch calculation of the skew leveling accuracy of multiple processes, and the calculation results are compiled and stored in reports.

10. The self-supervised deep processing and analysis system for inertial platform test data according to claim 1, characterized in that, The algorithm verification and accuracy analysis unit (20) includes a self-aiming accuracy calculation module (22), which is used to establish a self-aiming accuracy calculation process library. The self-aiming accuracy calculation process library encapsulates the names of the test processes that need to be self-aiming accuracy calculated. The test data self-traversal module (1) is called to automatically search the inertial platform test process and filter out the processes that have undergone self-aiming accuracy testing. Combined with the self-identification and reduction module (2), the data files in the process of self-aiming accuracy testing are reduced in two levels: the self-aiming accuracy calculation file library is used to identify the platform attitude acceleration data file from all data files to complete the first level of reduction and dimensionality reduction. The self-aiming accuracy calculation file library contains keywords for identifying platform attitude acceleration data files. The data file is matched with the self-aiming accuracy calculation format library to obtain the self-aiming data processing format in different test processes, which is used for subsequent self-aiming accuracy calculation, thereby completing the second level of reduction. The self-aiming accuracy calculation format library contains the self-aiming accuracy data processing format in different test processes. The modular algorithm adaptive matching module (5) is applied to extract specific command current data according to the corresponding data processing format and different test procedures. The mean and dispersion of the command current at each position of the two gyroscopes are statistically analyzed, and the self-aiming algorithm of the encapsulated inertial platform control software is adaptively called to complete the offline recalculation and verification of the self-aiming results of the α gyroscope and β gyroscope. Finally, the batch processing and report compilation module (3) is used to realize the batch calculation of the self-aiming accuracy of multiple processes and compile and summarize the calculation results into a report.

11. The self-supervised deep processing and analysis system for inertial platform test data according to claim 1, characterized in that, The algorithm verification and accuracy analysis unit (20) includes a self-calibration accuracy calculation module (23), which is used to establish a self-calibration accuracy calculation process library. The self-calibration accuracy calculation process library encapsulates the names of test processes that need to be self-calibrated. First, the test data self-traversal module (1) of the data processing layer automatically retrieves the inertial platform test process and filters out the process that has undergone self-calibration accuracy testing. Then, combined with the self-identification and reduction module (2), the data files in the process of self-calibration accuracy testing are reduced in two levels: the platform attitude acceleration data files are identified from all data files using the self-calibration accuracy calculation file library to complete the first level of reduction and dimensionality reduction. The self-calibration accuracy calculation file library contains keywords for identifying platform attitude acceleration data files. The data files are matched with the self-calibration accuracy calculation format library to obtain the self-calibration data processing format in different test processes, which is used for subsequent self-calibration accuracy calculation, thereby completing the second level of reduction. The self-calibration accuracy calculation format library contains the self-calibration accuracy data processing format in different test processes. The modular algorithm adaptive matching module (5) is used to calculate the mean and dispersion of the three gyroscope command currents at each self-calibration position by extracting the command current data of the effective segment, according to the corresponding data processing format and the test process related to self-calibration. Finally, the batch processing and report compilation module (3) is used to perform batch mathematical statistics on the accuracy of multiple sets of self-calibrated process data and compile reports on the calculation results.

12. The self-supervised deep processing and analysis system for inertial platform test data according to claim 1, characterized in that, The algorithm verification and accuracy analysis unit (20) includes a flight navigation accuracy calculation module (24), which is used to establish a flight navigation accuracy calculation process library. The flight navigation accuracy calculation process library encapsulates the names of test processes that need to perform self-aiming accuracy calculation. First, the test data self-traversal module (1) of the data processing layer automatically retrieves the inertial platform test process and filters out the process that has undergone flight navigation accuracy testing. Then, combined with the self-identification and reduction module (2), the data files in the process of flight navigation accuracy testing are reduced in two levels: the flight navigation accuracy calculation file library is used to identify the platform attitude acceleration data, platform tool error ① and platform tool error ② from all data files to complete the first-level reduction and dimensionality reduction. The flight navigation accuracy calculation file library contains keywords for identifying the above three files. The three data files were matched with the flight navigation accuracy calculation format library to obtain the flight navigation data processing format in different test procedures. This format was then used for subsequent flight navigation accuracy calculations to complete the second-level reduction. The flight navigation accuracy calculation format library encapsulates the flight navigation accuracy data processing format in different test procedures. The modular algorithm adaptive matching module (5) is used to read the error parameters of the inertial platform tool in the test process with simulated flight, and obtain the frame angle drift rate and acceleration fitting value during the flight navigation period by extracting and calculating the frame angle and accelerometer data during the simulated flight phase. Finally, the batch processing and report compilation module (3) is used to extract and calculate the accuracy of multiple sets of flight navigation data in batches, and the final results are compiled into reports.

13. The self-supervised deep processing and analysis system for inertial platform test data according to claim 1, characterized in that, The inertial information visualization display unit (30) includes: Command current display module (31): used to call the feature data autonomous visualization module (6) to extract the inertial information related to the command current, and to autonomously plot and display the command current of the three gyroscopes throughout the entire test process; and / or Frame angle display module (32): Used to call the feature data autonomous visualization module (6) to extract inertial information related to the frame angle, implement autonomous drawing and joint comparison of the four frame angle curves of the inertial platform, and identify the rotation of the inertial platform body or base; and / or Accelerometer Output Display Module (33): Used to call the feature data autonomous visualization module (6) to extract accelerometer inertial information, and to autonomously plot and display curves of the three accelerometer outputs throughout the entire test process; and / or Torque motor current display module (34): used to call the feature data autonomous visualization module (6) to extract the self-monitoring analog quantity information, and to implement autonomous plotting and curve display of the current of the four torque motors throughout the test process; and / or Temperature and temperature control display module (35): Used to call the feature data autonomous visualization module (6) to extract self-monitoring analog quantity information, and to implement autonomous plotting and curve display of inertial platform temperature and temperature control power stage current throughout the test process; and / or The simulated flight process display module (36) is used to call the feature data autonomous visualization module (6) to extract inertial information and implement autonomous plotting and curve display of the attitude and acceleration of the inertial platform during simulated flight; the modular algorithm adaptive matching module (5) calls the modular tool error parameters and performs comprehensive matching with the corresponding feature data to implement the accuracy calculation of the simulated flight data; and / or Telemetry data display module (37): Used to call the feature data autonomous visualization module (6) to analyze and extract telemetry inertial information and self-monitoring analog data, and to autonomously draw and display the attitude, acceleration and electrical parameters of the inertial platform in the telemetry data.

Images