Search

CN-122020345-A - Method for detecting and marking flow event in wind tunnel test and database

CN122020345ACN 122020345 ACN122020345 ACN 122020345ACN-122020345-A

Abstract

The invention belongs to the technical field of database construction, and discloses a method for detecting and labeling flow events in a wind tunnel test and a database. The method comprises the steps of obtaining characteristic data corresponding to sampling moments of all measuring points in a wind tunnel test, unifying the characteristic data to the same time axis, carrying out characteristic extraction on the characteristic data unified to the same time axis in a time window with a preset length to obtain characteristic vectors, calculating the mahalanobis distance between the characteristic vectors of adjacent time windows, taking the time window with the mahalanobis distance value larger than a preset threshold value as a flow event, calculating the frequency band energy of the adjacent flow event, setting an event attribute label of the flow event according to the frequency band energy, and establishing an event label of the flow event by utilizing the physical attribute and the event attribute label of the flow event, so that the flow event is marked. By the method and the device, the efficiency of data management and event sample retrieval is remarkably improved.

Inventors

  • LAI WUXING
  • GAO PAN
  • HUANG YONGAN
  • Rao Yuxuan

Assignees

  • 华中科技大学

Dates

Publication Date
20260512
Application Date
20260327

Claims (10)

  1. 1. The method for detecting and marking the flow event in the wind tunnel test is characterized by comprising the following steps: Acquiring characteristic data corresponding to each sampling moment of each measuring point in all events in a wind tunnel test, and unifying the characteristic data to the same time axis; Feature extraction is carried out on the feature data unified to the same time axis in a time window with a preset length to obtain feature vectors; Calculating the Marshall distance between feature vectors of adjacent time windows, wherein the Marshall distance value is larger than a preset threshold value and is used as a flow event in the time windows, and calculating the frequency band energy of the adjacent flow event; and establishing an event identifier of the flow event by using the physical attribute and the event attribute label of the flow event, thereby realizing the labeling of the flow event.
  2. 2. The method for detecting and labeling flow events in a wind tunnel test according to claim 1, wherein the frequency band energy is calculated according to the following formula: Wherein, the For multi-channel weighted power spectral density, For the interval between two adjacent discrete frequencies, To be in time window In the corresponding section of signal, the frequency interval The total band energy in-band, Is the first Power spectral density, weight of individual channels For reflecting the contribution degree of different channels to the target flow phenomenon.
  3. 3. A method for detecting and labeling flow events in a wind tunnel test according to claim 1 or 2, wherein the expression of the event identification is as follows: Wherein, the Is the first A unique identification of the individual event; For the test metadata set, including wind tunnel facility identification, test number and model configuration identification, Is a working condition parameter set comprising Mach number, reynolds number, attack angle, total pressure and total temperature, The starting and ending time of the event; the event attribute is marked and comprises event type, dominant frequency, band energy and intensity index.
  4. 4. A method of detecting and labeling flow events in a wind tunnel test as claimed in claim 3 wherein said eigenvector is formed by a combination of an eigenvalue root mean square, a power spectral density and a sequence of principal mode coefficients extracted from the eigenvalue data.
  5. 5. The method for detecting and marking flow events in a wind tunnel test according to claim 4, wherein said characteristic data comprises pressure, lift and speed in the wind tunnel test.
  6. 6. The multi-source time sequence database for the wind tunnel test flow event is characterized in that a core service layer of the database comprises a multi-source data access and time synchronization module, an event perception storage and index management module, an online flow event detection and labeling module, a flow event sample view generation module and an external data service interface, wherein: the multi-source data access and time synchronization module is used for receiving characteristic data corresponding to each sampling moment of each measuring point, unifying each characteristic data to the same time axis, and storing the unified result in the original measurement time sequence table; the online flow event detection and labeling module detects each flow event and establishes an event identifier, and an event metadata table is established by utilizing the corresponding relation between the flow event and the event identifier; The event sensing storage and index management module is used for setting an index relation between an event metadata table and an event metadata table sub-table, and a plurality of pieces of multi-element information about each flow event in the wind tunnel test are stored in the event metadata table sub-table; the flow event sample view generation module stores and analyzes external tasks, then finds out corresponding sub-tables through event metadata tables according to analysis results, and finds out start-stop and stop lines corresponding to corresponding block files and data in the corresponding sub-tables.
  7. 7. The multi-source timing database of claim 8, further comprising an external data interface in the multi-source timing database for invoking a module in a core business layer of the multi-source timing database.
  8. 8. The multi-source timing database of claim 6 or 7, further comprising a data storage layer comprising an event metadata table, an event metadata table sub-table, an original measurement timing table, and a sample view table for storing external tasks.
  9. 9. The multi-source time series database according to claim 6 or 7, wherein the event metadata table sub-table comprises a channel metadata table, a facility information table, a test information table, a model information table and an original time series index table, wherein the channel metadata table is used for storing multi-source multi-channels involved in wind tunnel tests for unified cataloging and management, recording channel identifications, data source types, channel names, nominal sampling rates, units and measuring point position references, the facility information table is used for basic identifications and capability parameters of wind tunnel facilities, the test information table is used for storing test numbers, affiliated facility identifications, configuration identifications, test start-stop times and preset working condition ranges, the model information table is used for storing tested model configurations, sensor layout references and geometric versions, and the original time series index table is used for storing sampling rates, start and stop times, channel group identifications, event identifications, test information identifications, facility information identifications and storage addresses of original sampled data blocks.
  10. 10. The multi-source time series database according to claim 6 or 7, wherein the expression of the outside person is as follows: Wherein, the For the purpose of modeling the task type, In order to filter the set of events, For the type of data that needs to be used and its channel set, For the length of the time window of a single sample, For the sliding step size of the time window, Is the alignment and transformation rule of the input and output variables.

Description

Method for detecting and marking flow event in wind tunnel test and database Technical Field The invention belongs to the technical field of databases, and particularly relates to a method for detecting and labeling a flow event in a wind tunnel test and a database. Background In the prior engineering practice, the organization mode of wind tunnel test data is mainly a traditional mode of dividing a catalog according to test numbers and dividing files according to equipment or data types, and a part of units are used for constructing a simple relational database or a time sequence database on the basis and recording metadata such as test numbers, working condition parameters, data file paths and the like. The scheme plays a certain role in managing test files and supporting inquiry according to test numbers or time ranges, but from the viewpoint of database modeling, the basic unit of data organization is still limited to a test-channel-time ternary structure, and the abstraction of the physical evolution rule of the flow process is lacking. In wind tunnel tests, researchers have often focused not on the complete time series itself, but on "flow events" with well-defined physical implications therein, such as boundary layer development and separation, stages before and after stall occurs, shock wave formation and movement, oscillation intervals of shock wave-boundary layer interactions, steady-state oscillation segments of deep stall or buffeting, and the like. These flow events are typically manifested as synergistic changes in the multi-source signal, such as abrupt changes in lift and pressure distribution, significant enhancement of energy in a particular frequency band, abrupt changes in flow field structure patterns, and the like. However, in existing systems, the identification and labeling of these flow events often rely on analysis scripts at post-processing stage, and part of the time period of a test is manually labeled by an analyst in an offline environment through threshold decisions, spectral analysis, or empirical rules, and such labeling information is difficult to systematically incorporate into a database data model, and also difficult to multiplex between different tests and different projects. On the other hand, the multi-source data acquisition system of the wind tunnel test is generally composed of data acquisition devices and measurement subsystems of different manufacturers, and the sampling frequency, time reference and trigger mechanism of each system are not completely consistent. In the prior art, a simple timestamp alignment and interpolation resampling mode is adopted, and the rough alignment of multi-source data is realized in the data post-processing. Due to the lack of unified modeling of "multi-source alignment" and "flow event boundaries" from the database level, different researchers often repeatedly implement alignment and segmentation logic in respective scripts, resulting in a great deal of repeated labor, and also being unfavorable for ensuring the consistency and traceability of multi-source alignment results. In addition, as the application of data-driven aerodynamic load prediction, flow field reconstruction, proxy model modeling and other methods in wind tunnel tests increases, researchers increasingly want to construct training samples based on "physical flow events" rather than simple "test number+time period", such as "extract all event fragments where shock oscillations occur and the main frequency of oscillation is in a specified frequency band", "extract all multi-source joint data sets within several seconds before and after stall occurrence", and so on. The existing database system only generally supports inquiry according to test numbers, working condition parameters and time intervals, and lacks indexing and searching capabilities for high-level semantic conditions such as flow event types, event intensity, event spectrum characteristics and the like, so that the sample construction process can only be completed through repeated inquiry and complex script splicing at an application layer, and the efficiency is low, the process is opaque and is difficult to multiplex. In summary, the existing wind tunnel test data management technology has the main defects that firstly, a data organization unit stays on a test-channel-time level, a flow event with a physical meaning is not abstracted into a unified data basic unit, secondly, alignment of multi-source data and event boundary identification mainly depend on upper-layer script post-processing and lack of unified specification and algorithm support in a database writing stage, thirdly, a database index structure and a query interface mainly search according to test numbers and time ranges and cannot effectively support complex search requirements for the flow event, fourthly, sample construction facing a proxy model and deep learning modeling lacks system-level support, manual writing scripts are seriously r