Search

CN-122020531-A - Reference construction method and device for multi-source transient signal data set

CN122020531ACN 122020531 ACN122020531 ACN 122020531ACN-122020531-A

Abstract

The application discloses a reference construction method and device for a multisource transient signal data set, and relates to the field of sensors. The system characterizes dynamic behaviors of transient signals under different working conditions through unified scene design and multi-source sensor configuration, constructs a multi-dimensional dynamic index system covering core elements such as sampling precision, transient response time, dynamic range, frequency domain characteristic recovery, time sequence consistency and the like, realizes quantitative evaluation of multi-source transient signal data quality, introduces a pareto multi-objective optimization and reinforcement learning mechanism, realizes multi-objective balanced evaluation and dynamic update of a data set reference, supports migration and use among different tasks and different working conditions, and forms a set of repeatable, expandable and engineering floor multi-source transient signal data set reference system through standardized index definition, calculation mode and evaluation flow, thereby providing a unified data basis for development and evaluation of a follow-up dynamic compensation algorithm and an intelligent diagnosis model.

Inventors

  • XU BO
  • XU DONGYANG
  • TANG HAO
  • YOU JIA
  • WANG SHUO

Assignees

  • 海南大学

Dates

Publication Date
20260512
Application Date
20260126

Claims (10)

  1. 1. A method of reference construction of a multisource transient signal dataset, the method comprising: Acquiring a multi-source transient signal under a single scene transient working condition; constructing a multi-dimensional dynamic index system, mapping the multi-source transient signals to a quality space under the multi-dimensional dynamic index system, and obtaining a heterogeneous index set; Constructing a pareto multi-target optimization model, acquiring a candidate data set of the multi-target optimization model according to the heterogeneous index set, and acquiring a non-dominant solution set through a non-dominant sequencing genetic algorithm; and dynamically updating the reference standard through reinforcement learning.
  2. 2. The method of claim 1, wherein the multi-dimensional dynamic metric system comprises metrics characterizing sampling accuracy, transient response time, dynamic range, frequency domain characteristic recovery, and timing consistency.
  3. 3. The method of claim 1, wherein the obtaining a candidate dataset of multi-objective optimization problems from the heterogeneous index set comprises: And carrying out dimensionless treatment on the heterogeneous index set according to the polarity of the index in the multidimensional dynamic index system to obtain the candidate data set, wherein the polarity of the index comprises a benefit index and a cost index.
  4. 4. The method of claim 1, wherein said synthetically scoring said non-dominant solution set comprises: And constructing a comprehensive scoring function by taking geometric average as a decision function, and acquiring the scoring result of the non-dominant solution set by using the comprehensive scoring function.
  5. 5. The method of claim 1, wherein dynamically updating the reference by reinforcement learning comprises constructing a state space and an action space; the state space consists of the heterogeneous index set and task preference weights, and the action space comprises actions for intervening in the multi-source transient signal acquisition flow and the candidate data set acquisition flow.
  6. 6. The method of claim 5, wherein dynamically updating the reference by reinforcement learning further comprises constructing a reward function; The reward function is constructed according to the comprehensive scoring function, the dynamic scoring baseline and the cost function, and the cost function is constructed according to resource consumption caused by executing actions.
  7. 7. A reference construction apparatus for a multi-source transient signal data set, the apparatus comprising: The system comprises an offline training module, a multi-dimensional dynamic index system, a pareto multi-target optimization model, a non-dominant solution set, a comprehensive scoring module, a reference standard, a non-dominant ranking genetic algorithm, a multi-target optimization model and a multi-target optimization model, wherein the offline training module is used for acquiring multi-source transient signals under a transient condition of a single scene; And the online fine adjustment module is used for dynamically updating the reference standard through reinforcement learning.
  8. 8. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor implements the steps of the method of any of claims 1 to 6 when the computer program is executed.
  9. 9. A computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the steps of the method of any of claims 1 to 6.
  10. 10. A computer program product comprising a computer program, characterized in that the computer program, when being executed by a processor, implements the steps of the method of any of claims 1 to 6.

Description

Reference construction method and device for multi-source transient signal data set Technical Field The application relates to the technical field of sensors, in particular to a reference construction method and device for a multisource transient signal data set. Background With the continuous promotion of new technological revolution and industrial revolution, the high-technology fields of aerospace, intelligent automobiles, the Internet of things and the like are vigorously developed, the requirements for high-precision testing and monitoring are explosive growth, in the fields, the accurate capturing of transient signals becomes a key core for detecting rapid changes in a dynamic environment, and the technology is widely applied to various important aspects such as high-frequency vibration testing, dynamic fault diagnosis, complex system performance evaluation and the like. In typical aerospace tasks such as launch of a carrier rocket, ground thermal test of an engine, attitude maneuver of a rocket flight segment and the like, structural vibration, impact load, cabin pressure fluctuation and multiple physical field coupling effects are rapidly changed in millisecond orders, and extremely high requirements are put forward on dynamic response performance of a sensor and consistency of a test system. Whether the transient test data is accurate or not can directly influence the reliability of flight state interpretation, fault positioning and subsequent model improvement design. The firing and explosion processes of the weapon release large dynamic range signals in a very short time, which puts stringent requirements on the sensitivity, response speed and anti-interference capability of the sensor, and is directly related to the safety and the effective exertion of functions of a weapon system. In the prior engineering practice, transient test in the aerospace field is mostly dependent on a large number of distributed heterogeneous sensors such as pressure, acceleration, strain, temperature and the like. These sensors have significant differences in dynamic range, frequency response, noise characteristics, and environmental suitability, resulting in the multi-source transient signals exhibiting significant non-uniformity under the same operating conditions. On the other hand, due to the complex environment of high temperature, high humidity and high sound pressure level of the transmitting field and the complex working conditions of high overload, high vibration and high coupling during flight, transient events have the characteristics of short-time burst, strong nonlinearity and strong coupling, and the test data are extremely easy to be distorted, drifting and lost in the time-frequency domain direction. At present, although a few impact and vibration open source data sets exist in the aerospace field, most of the data sets are history records of specific test tasks, access is limited, working condition labeling is incomplete, meanwhile, the existing data sets are constructed to pay more attention to a single sensor or a single scene, and the requirements of uniform compensation of heterogeneous sensors and evaluation of a cross-scene algorithm are difficult to meet. In order to support flight test and ground verification tasks, a plurality of transient signal data sets facing specific tasks, such as a carrier rocket engine length Cheng Shiche data set, a structural vibration test data set and the like, are gradually accumulated in engineering practice. In addition, there is also internationally a published data set of aerospace impact signals represented by NASA (National Aeronautics and Space Administration, american aerospace agency) impact test data, which is widely used for structural dynamics analysis and fault diagnosis algorithm verification. However, most of data sets are data collected under specific models and specific working conditions, scene coverage is limited, and the data sets are difficult to directly migrate to other tasks or new configurations, most of the data sets are concentrated in single-type sensors, a multi-source and multi-physical-quantity combined collection data system is lacked, data set construction is focused on data acquisition, and systematic and quantifiable reference definition is lacked on indexes such as sampling precision, transient response, dynamic range, frequency domain recovery, cross-sensor time sequence consistency and the like. Under the background of lack of open source data, researchers often rely on self-built data sets and use methods of generating countermeasure networks, transfer learning and the like to enhance data so as to alleviate the problems of insufficient samples and unbalanced distribution. However, the method commonly multiplexes the existing data base, and is difficult to fundamentally solve the pain points such as deviation between the synthesized data and the real working condition, lack of uniform evaluation standards for data quality ev