CN-122015904-A - Knowledge data hybrid-driven rate gyro accurate health state assessment method

CN122015904ACN 122015904 ACN122015904 ACN 122015904ACN-122015904-A

Abstract

The application provides a knowledge data hybrid-driven rate gyroscope accurate health state assessment method which comprises the steps of collecting data of a rate gyroscope, designing key indexes as confidence rule base BRB input according to output information of the rate gyroscope, constructing a confidence rule base, conducting on-line reasoning on a BRB model, conducting training data label conversion, conducting parameter optimization on the BRB model, enabling the model to automatically adjust parameters according to historical data through an optimization process, and accordingly improving accuracy and reliability of the rate gyroscope health state assessment.

Inventors

WANG TAO
CHEN YINCHAO
LIANG ZHAOXIN
CUI XIAOJING

Assignees

中国航空工业集团公司成都飞机设计研究所

Dates

Publication Date: 20260512
Application Date: 20251227

Claims (8)

1. A knowledge data hybrid-driven rate gyroscope accurate health state assessment method, the method comprising: step one, collecting data of a rate gyroscope; Step two, designing key indexes as the input of a confidence rule base BRB according to the output information of the rate gyroscope; step three, constructing a confidence rule base; Step four, BRB model online reasoning; step five, training data label conversion; and step six, optimizing parameters of the BRB model, wherein the model can automatically adjust the parameters according to historical data through an optimization process, so that the accuracy and the reliability of the rate gyroscope health state evaluation are improved.
2. The method of claim 1, wherein the first step comprises: In order to build a health state evaluation model of the rate gyroscope, key performance index data of the rate gyroscope under different working conditions and health states are firstly required to be collected, wherein the key performance index data comprises the following components: Zero-order item drift, which reflects zero-value deviation output by the gyroscope under the static condition, is an important index for measuring the static performance of the gyroscope; And one-time item drift, namely reflecting the change of the scale factors of the gyroscope and influencing the dynamic measurement accuracy of the gyroscope.
3. The method according to claim 1, wherein the second step comprises: By means of output data of the rate gyroscope, two key performance indexes of zero-order item drift and one-time item drift are designed to serve as inputs of a BRB health state evaluation model, the two indexes can effectively reflect performance changes of the rate gyroscope in different health states and are easily affected by external interference and internal aging, and an initial confidence rule base BRB is built by collecting the index data and combining expert knowledge and historical data to provide rule bases for subsequent health state evaluation.
4. The method according to claim 1, wherein the third step comprises: Based on expert knowledge and historical operation data of the rate gyroscope, an initial confidence rule base BRB is constructed, wherein the confidence rule base is used for fusing expert experience and data information, processing data imbalance problems and providing rule bases for subsequent health state evaluation, and based on model input, the kth confidence rule of a health state evaluation model of the rate gyroscope can be defined as: Wherein, the The observation data of the mth index at the time t is represented, A reference value representing the mth front piece in the kth rule; Representing a health status grade; A reference value representing an nth back-piece in a kth rule; Represent the first Confidence rule Weights of (2); Representation of In the rules of The attribute weight below, and initially, Determined by empirical knowledge.
5. The method according to claim 1, wherein the fourth step comprises: wherein t=1, 2, T, T is the total number of sensor data; the (k+1) th and the (k) th reference points, which are the m th sensor output, are determined by an expert, and the current sensor data input of the rate gyro is expressed as Wherein L represents the total number of rules in the BRB model; Representation of For reference values Matching degree of (a) is attribute matching degree, and attribute matching degree The method can be used for calculation by the following method: then, a rule matching degree matrix is obtained by: Wherein, the Rule-based matching degree matrix Activation weight of kth rule The calculation can be made by the following formula: all activated confidence rules will then be fused by the evidence reasoning parsing algorithm, and the fusion process can be described as: Wherein, the And finally, the output of the rate gyroscope health state assessment model can be obtained through the following utility conversion: 。
6. The method according to claim 1, wherein the fifth step comprises: converting each tag data sample into a target confidence; 。
7. The method of claim 6, wherein the step of providing the first layer comprises, The calculation is as follows: Where N is the number of output confidence levels and T is the total number of all data.
8. The method according to claim 1, wherein the step six comprises: The unknown parameters of the BRB model are tuned by means of the historical data, so that the optimization problem of the BRB model can be expressed as: Where n represents the sample size, y i represents the observed value related to the unknown external interference, θ is the unknown parameter vector to be optimized, the objective function is the Huber loss function, which is a robust loss function that appears as a square error when the error is small and as an absolute error when the error is large, so that the method has better robustness when dealing with abnormal values, and the Huber loss function is defined as follows: Wherein, the For controlling the switching of the loss function between a squared error and an absolute error; In the optimization process, the parameter updating strategy of the model comprises updating of confidence coefficient, rule weight, fuzzy set parameter and attribute weight, and the specific updating formula is as follows: Confidence level Is updated by: Rule weights Is updated by: Attribute weight Is updated by: Wherein, the Is the learning rate; Is the gradient to be calculated; gradient calculation is key for parameter updating, and a specific calculation formula is as follows: subsequent confidence gradient It can be calculated as: Wherein: Regular weight gradient It can be calculated as: Wherein: Attribute weight gradient It can be calculated as: Wherein: the calculation formula of (2) is the same as that described above; To preserve the interpretability of the BRB model, after each parameter update, the parameters need to be constrained: 。

Description

Knowledge data hybrid-driven rate gyro accurate health state assessment method Technical Field The application belongs to the technical field of health state evaluation of rate gyroscopes, and particularly relates to a knowledge data hybrid drive accurate health state evaluation method of a rate gyroscope. Background The rate gyroscope is used as a high-precision angular velocity sensor, is widely applied to high-precision systems such as navigation, guidance, attitude control and the like by virtue of excellent measurement precision and stability, and the performance of the rate gyroscope is directly related to the overall performance and reliability of the system. However, rate gyroscopes are susceptible to environmental changes, mechanical wear, electronic component aging, and the like during long-term operation, resulting in performance degradation and even failure. Common faults are manifested as zero-order term drift and abnormal changes in the primary term drift, which can seriously affect the accuracy and stability of the system. Therefore, by designing an accurate and reliable health state evaluation method, the performance change of the rate gyroscope is monitored in real time, and corresponding maintenance measures are taken, so that the method has important practical significance for improving the robustness and reliability of the system. In view of this, in order to improve the traditional rate gyroscope health state assessment method in terms of robustness and accuracy, a new health state assessment scheme based on knowledge data hybrid driving is proposed, and mainly solves the following two problems: 1) And the data volume of different modes is unbalanced. In the health state evaluation of the rate gyroscope, the situation that the data amount is unbalanced often exists in the acquisition of multi-mode data, some sensor data are easy to acquire, and other data are rare, so that the training and generalization capability of an evaluation model can be affected. 2) Accuracy of health status assessment results. The traditional evaluation method relies on simple threshold judgment or statistical analysis, and is difficult to capture the subtle change and complex characteristics of the health state of the rate gyroscope, and particularly in the early failure state or the complex failure mode, the accuracy of the evaluation result needs to be improved. Disclosure of Invention The invention aims to overcome the defects in the prior art and provide a knowledge data hybrid-driven method for evaluating the accurate health state of a rate gyroscope, so as to solve the problems of unbalanced data quantity of different modes and insufficient accuracy of the health state evaluation result and realize accurate and reliable evaluation of the health state of the rate gyroscope. The application provides a knowledge data hybrid-driven rate gyroscope accurate health state assessment method, which comprises the following steps: step one, collecting data of a rate gyroscope; Step two, designing key indexes as the input of a confidence rule base BRB according to the output information of the rate gyroscope; step three, constructing a confidence rule base; Step four, BRB model online reasoning; step five, training data label conversion; and step six, optimizing parameters of the BRB model, wherein the model can automatically adjust the parameters according to historical data through an optimization process, so that the accuracy and the reliability of the rate gyroscope health state evaluation are improved. Preferably, the first step includes: In order to build a health state evaluation model of the rate gyroscope, key performance index data of the rate gyroscope under different working conditions and health states are firstly required to be collected, wherein the key performance index data comprises the following components: Zero-order item drift, which reflects zero-value deviation output by the gyroscope under the static condition, is an important index for measuring the static performance of the gyroscope; And one-time item drift, namely reflecting the change of the scale factors of the gyroscope and influencing the dynamic measurement accuracy of the gyroscope. Preferably, the second step includes: By means of output data of the rate gyroscope, two key performance indexes of zero-order item drift and one-time item drift are designed to serve as inputs of a BRB health state evaluation model, the two indexes can effectively reflect performance changes of the rate gyroscope in different health states and are easily affected by external interference and internal aging, and an initial confidence rule base BRB is built by collecting the index data and combining expert knowledge and historical data to provide rule bases for subsequent health state evaluation. Preferably, the third step includes: Based on expert knowledge and historical operation data of the rate gyroscope, an initial confidence rule base BRB is constru