Laser gyroscope performance prediction method considering mission profile

Abstract

The invention discloses a laser gyroscope performance prediction method considering a mission profile, and belongs to the technical field of inertial navigation equipment health management. The prediction method comprises the following steps: according to different task sub-profiles experienced by the laser gyroscope, calculating a task intensity factor for representing the influence degree of each sub-profile on the performance degradation of the laser gyroscope; fusing the task intensity factor into an inference mechanism of a confidence rule base, and constructing a laser gyroscope performance prediction model considering a task profile; then, optimizing parameters of the laser gyroscope performance prediction model by adopting a covariance matrix adaptive evolutionary strategy; and finally, predicting the performance state of the laser gyroscope under a specific task profile by using the optimized model. According to the method, task profile information and expert knowledge can be effectively fused, the performance state prediction precision of the laser gyroscope in a complex task environment is remarkably improved, and a reliable basis is provided for health management, fault prediction and maintenance decision of an inertial navigation system.

Description

Technical Field

[0001] This invention belongs to the field of health management technology for inertial navigation devices, and specifically relates to a method for predicting the performance of laser gyroscopes that takes into account mission profiles. Background Technology

[0002] As a core component of inertial navigation systems, laser gyroscopes play a crucial role in high-end equipment fields such as aerospace, precision measurement, and control due to their high precision, high reliability, and long lifespan. Their performance directly determines the accuracy of the entire system's navigation, measurement, and control. In actual operation, laser gyroscopes often experience complex mission profiles, meaning they are subjected to the combined effects of various environmental stresses, such as temperature and vibration, changing over time at different stages or in different scenarios. This dynamic mission profile can significantly accelerate or alter the performance degradation process of laser gyroscopes, thus posing a severe challenge to predicting their long-term performance. Therefore, achieving accurate prediction of laser gyroscope performance for specific mission profiles is of great significance for improving overall system reliability, implementing predictive maintenance, and optimizing operational strategies.

[0003] Currently, performance prediction methods for such precision devices can be mainly divided into the following three categories:

[0004] One approach is based on mechanistic analysis. This method requires a deep understanding of the internal physicochemical processes of the device and the establishment of a clear degradation model. Theoretically, this method has high interpretability and extrapolation capabilities, but it demands extremely high transparency regarding the internal mechanisms of the research object, and the modeling process is complex. For complex systems like laser gyroscopes that involve multi-physics coupling, their degradation mechanisms are difficult to describe completely with precise mathematical models. Therefore, purely mechanistic methods face significant challenges in practical applications.

[0005] Second, there is the data-driven approach: This approach has been widely used in recent years, utilizing various machine learning algorithms to extract performance evolution patterns from historical data. However, as high-value, long-life precision devices, laser gyroscopes face high costs and long cycles in acquiring their full life-cycle performance data, and frequent testing may cause irreversible damage to their performance. This results in a severe shortage of effective samples available for modeling, making it difficult to support the large data volume requirements of data-driven models.

[0006] Thirdly, there are hybrid-driven approaches that integrate domain expert knowledge, mechanistic models, and measured data to overcome the shortcomings of single methods. Examples include physical information neural networks and belief rule bases (BRBs). BRBs effectively combine limited expert prior knowledge with a small amount of monitoring data, exhibiting relatively low dependence on data volume and providing a potential pathway for performance prediction under small sample conditions. However, when applied to laser gyroscopes, traditional belief rule base models typically treat the operating environment as static or simply segmented, failing to deeply model the differential modulation effect of dynamic task profiles on performance degradation rates. When task profiles switch, the model lacks corresponding adaptive adjustment mechanisms, resulting in insufficient prediction accuracy and robustness under complex, time-varying conditions.

[0007] In summary, existing methods have limitations in accurately predicting the performance of laser gyroscopes under dynamic mission profiles: mechanistic modeling is difficult, data-driven samples are insufficient, and traditional hybrid models cannot effectively characterize the impact of mission profiles. Therefore, there is an urgent need for a method that can closely integrate the dynamic characteristics of mission profiles and achieve reliable predictions under small sample conditions.

[0008] In view of this, this invention is hereby proposed. Summary of the Invention

[0009] The purpose of this invention is to overcome the shortcomings of the prior art and provide a laser gyroscope performance prediction method that considers mission profiles. This addresses the problems of insufficient prediction accuracy, high dependence on data volume, and failure to effectively integrate mission profile information in traditional methods under dynamic mission profiles. This invention quantifies the impact of sub-profiles on laser gyroscope performance degradation, constructs a new confidence rule base (BRB) model incorporating mission intensity factors, and optimizes the model parameters using a constrained covariance matrix adaptive evolution strategy (P-CMA-ES). Ultimately, it achieves high-precision prediction of laser gyroscope performance under small sample conditions, thus providing strong support for laser gyroscope selection and maintenance decisions in subsequent engineering projects.

[0010] The objective of this invention is achieved through the following technical solution:

[0011] This invention provides a method for predicting the performance of a laser gyroscope considering a mission profile, comprising the following steps:

[0012] Step 1: Calculate the task intensity factor, which characterizes the degree of performance degradation of each sub-profile, based on the different task sub-profiles experienced by the laser gyroscope.

[0013] Step 2: Integrate the task intensity factor into the reasoning mechanism of the confidence rule base to construct a laser gyroscope performance prediction model that considers the task profile;

[0014] Step 3: Optimize the parameters of the laser gyroscope performance prediction model using a covariance matrix adaptive evolution strategy;

[0015] Step 4: Based on the optimized laser gyroscope performance prediction model, predict the performance status of the laser gyroscope under a specific task profile.

[0016] Furthermore, in step 1, the different task sub-profiles include storage profiles and operation profiles;

[0017] The work profile is divided into Level I, Level II and Level III work profiles according to the task intensity.

[0018] Further, in step 1, the calculation formula for the task intensity factor is:

[0019]

[0020] In the formula: Sub-section Task intensity factor These correspond to the Level I operation profile, Level II operation profile, Level III operation profile, and storage profile, respectively. For the laser gyroscope in the sub-section Maximum operating time; The attenuation coefficient has a range of values. ; and The laser gyroscope enters and exits the sub-sections respectively. Time cutoff points; and Sub-sections The temperature discount factor and the power discount factor of the laser gyroscope are described below.

[0021] Furthermore, in step 2, the laser gyroscope performance prediction model is a confidence rule base model, whose first... Rules Represented as:

[0022]

[0023] In the formula: These are the performance prediction indicators for the laser gyroscope; This is a reference level for the indicator; Number of rules; The predicted performance value of the gyroscope The corresponding reference level; 1. Confidence level; N is the number of performance reference levels; For the first The rule weight of each rule; Sub-section Task intensity factor; The weights are the indicator weights.

[0024] Furthermore, the performance prediction metrics include zero-order term drift coefficient and first-order term drift coefficient.

[0025] Furthermore, in step 2, integrating the task strength factor into the reasoning mechanism of the confidence rule base specifically includes the rule activation step:

[0026] First, combine the aforementioned task intensity factor Calculate the first Total matching degree of the rules :

[0027]

[0028] Calculate the next Activation weight of the rule :

[0029]

[0030] In the formula: For the first The performance prediction metric in the first The degree of matching in the rule; For the first The weight of each indicator; For the first The rule weight of each rule.

[0031] Furthermore, step 2 also includes a rule reasoning step: using an evidence-based reasoning algorithm to fuse all activated rules, the performance prediction value is calculated using the following formula:

[0032]

[0033] In the formula: This is a predicted value for the performance status of the laser gyroscope. For reference level The utility value; The relative reference level calculated using the evidence reasoning algorithm The overall confidence level; These are intermediate variables in the reasoning process; for the sake of simplification, Actual representative , Actual representative .

[0034] Furthermore, in step 2, the initial parameters of the laser gyroscope performance prediction model are determined based on expert knowledge.

[0035] Furthermore, in step 3, with the goal of minimizing the mean square error between the predicted output and the actual value of the laser gyroscope performance prediction model, the following parameter optimization model is constructed:

[0036]

[0037] In the formula: The parameters to be optimized are, respectively, rule weight, confidence level, indicator weight, and decay coefficient. for Real-time performance status label value; The number of data points.

[0038] Furthermore, in step 3, the parameter optimization model is optimized using a covariance matrix adaptive evolution strategy with projection mechanism, including the following steps:

[0039] Step 3.1: Determine initial values: Set initial parameters Their meanings are respectively: initial parameter vector, initial covariance matrix, initial step size, population size, offspring population size, and optimal solution;

[0040] Step 3.2, Generate the initial population: In the... Generation, generated through normal distribution sampling One candidate solution:

[0041]

[0042] In the formula: It is the first The first generation There are 10 candidate solutions; The representative value is 0 and the covariance is The normal distribution;

[0043] Step 3.3: Project candidate solutions to the constraint hyperplane: Project the candidate solutions... Project onto a hyperplane that satisfies the constraints, ensuring that the solution satisfies the physical constraints:

[0044]

[0045] In the formula: For parameter matrices; Representative at The number of variables in the subspace; T represents transpose, and j represents the j-th group of variables or the j-th subspace;

[0046] Step 3.4: Select the best and update the mean: Select the offspring with the best fitness from the population. Each solution updates the mean to guide the next generation of search centers:

[0047]

[0048] In the formula: These are the weighting coefficients. Representative from the first The selected scheme in the generation;

[0049] Step 3.5: Update the covariance matrix: By dynamically adjusting the covariance matrix, historical information is balanced with the current population distribution.

[0050]

[0051] In the formula: The learning rate; For evolutionary paths;

[0052] Step 3.6, Iteration: Repeat steps 3.2 to 3.5 until the termination condition is met.

[0053] Compared with the prior art, the present invention has the following beneficial effects:

[0054] The prediction method provided by this invention first analyzes the working conditions of the laser gyroscope under various sub-profiles to quantify the task intensity factor affecting the performance degradation of each sub-profile. Then, it constructs an inference mechanism that integrates the task intensity factor into the confidence rule base (BRB), achieving accurate modeling of the dynamic impact of performance degradation under different task profiles. This effectively solves the problem of insufficient prediction accuracy of traditional methods in dynamic environments. The new BRB model that integrates task profile information retains the advantages of strong interpretability based on expert knowledge while enhancing its adaptability to complex dynamic conditions. Simultaneously, the covariance matrix adaptive evolution strategy with projection mechanism (P-CMA-ES) is used to optimize the model parameters, significantly improving the prediction accuracy under small sample conditions. Experimental results show that after improvement by the method of this invention, the prediction mean square error (MSE) is reduced from 1.6 × 10⁻⁶ in the traditional optimized BRB model. -2 Significantly reduced to 4.3×10 -3 The prediction accuracy is improved by nearly an order of magnitude, and the prediction results are basically close to the true values, which fully verifies the reliability and accuracy of the prediction method of this invention. At the same time, under the condition of only 180 sets of data, the method shows excellent adaptability and robustness, providing high-precision and high-reliability technical support for the health management, life prediction and maintenance decision-making of laser gyroscopes under complex dynamic task profiles. Attached Figure Description

[0055] The accompanying drawings are incorporated in and form part of this specification, and together with the description serve to explain the principles of the invention.

[0056] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0057] Figure 1 A flowchart illustrating the laser gyroscope performance prediction method considering the mission profile of this invention;

[0058] Figure 2 This is a schematic diagram of the zero-order term drift coefficient of the laser gyroscope according to an embodiment of the present invention;

[0059] Figure 3 This is a schematic diagram of the first-order drift coefficient of the laser gyroscope according to an embodiment of the present invention;

[0060] Figure 4 This is a comparison chart of the performance state prediction results of the laser gyroscope of the present invention. Detailed Implementation

[0061] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Rather, they are merely examples consistent with some aspects of the invention as detailed in the appended claims.

[0062] To enable those skilled in the art to better understand the technical solutions of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments.

[0063] Please see Figures 1 to 4 The laser gyroscope performance prediction method considering the task profile provided in this embodiment of the invention has the following specific implementation steps:

[0064] Step 1: Calculate the task intensity factor, which characterizes the degree of performance degradation of each sub-profile, based on the different task sub-profiles experienced by the laser gyroscope.

[0065] Specifically, laser gyroscopes undergo different mission phases in practical use, and the combined environmental conditions and operating modes of each phase constitute a mission sub-profile. To quantify the impact of different sub-profiles on performance degradation, it is first necessary to divide the mission sub-profiles.

[0066] In this embodiment, based on the actual working conditions of a certain type of aircraft laser inertial navigation system, the task sub-profile of the laser gyroscope is divided into a storage profile and an operational profile. The operational profile is further divided into Level I operational profile, Level II operational profile, and Level III operational profile according to the task intensity, as detailed below:

[0067] 1. Level I operational profile: This corresponds to high-intensity / emergency response missions, where the aircraft is in a state of high complexity, high risk, and high resource input, typically occurring when transporting medical supplies or wartime resources;

[0068] 2. Level II Operational Profile: Corresponds to routine flight missions, where aircraft carry out routine cargo transport or personnel transport along fixed routes;

[0069] 3. Level III operational profile: This corresponds to the aircraft's "hot standby" status. In this status, the aircraft has completed maintenance and inspection, and may even have been refueled, and is ready to take off at any time.

[0070] 4. Storage profile: The corresponding aircraft reserved by the airline is in a constant temperature and humidity storage environment when the laser gyroscope is not in operation.

[0071] It should be noted that in actual transportation, air transport is the common mode of transport for laser gyroscopes. This mode of transport has the advantage of short transport cycle and can provide sufficient protection for the laser gyroscope. Therefore, the impact of the transport profile on the performance of the laser gyroscope is negligible. Thus, this embodiment does not include the transport profile in the calculation. Based on practical engineering experience, the degree of influence of different sub-profiles on the gyroscope performance degradation rate, from highest to lowest, is as follows: Level I operating profile > Level II operating profile > Level III operating profile > storage profile.

[0072] Based on the above description, this embodiment of the invention focuses on the performance influencing factors of laser gyroscopes under different storage and operational profiles. Through research and communication with relevant aviation departments, it was learned that during the storage phase, the laser gyroscope is in a suitable environment with constant temperature and humidity, and its performance only undergoes natural degradation. However, when it transitions to the operational profile, the laser gyroscope starts working, and factors such as heating and power supply accelerate its performance degradation. Furthermore, the operating conditions differ between different levels of operational profiles, resulting in varying acceleration effects on performance degradation. In addition, operating time, as a key influencing factor on laser gyroscope performance, has a significant impact on its performance degradation process. In summary, considering the differences in operating conditions between the storage and operational profiles, performance degradation characteristics, and the influence of operating time, the following formula for calculating the task intensity factor can be constructed:

[0073]

[0074] The meanings of the parameters in the formula are as follows:

[0075] Sub-section Task intensity factor These correspond to the Level I operation profile, Level II operation profile, Level III operation profile, and storage profile, respectively.

[0076] For this laser gyroscope in sub-section The maximum operating time, in days or years;

[0077] The attenuation coefficient has a range of values. ;

[0078] and These are the entry and exit sub-sections of the laser gyroscope, respectively. Time cutoff points;

[0079] and Sub-sections The temperature discount factor and the power discount factor of the laser gyroscope are calculated.

[0080] Among them, temperature discount factor and power discount factor The parameters need to be determined based on the specific model of the laser gyroscope, the technical manual provided by the manufacturer, and the actual usage environment. In this embodiment, the specific parameter values ​​of the laser gyroscope are shown in Table 1 and Table 2 below:

[0081] Table 1 Temperature discount factor for this laser gyroscope

[0082]

[0083] Table 2. Discount factor for the power-on of this gyroscope

[0084]

[0085] Based on actual records, the maximum operating time of the laser gyroscope in each sub-section in this embodiment is as follows: , , , The initial value of the attenuation coefficient γ is set to 1. Based on the actual working time of each sub-profile (i.e., the time difference between the laser gyroscope's entry and exit; approximately 4 days for Level I and approximately 7 days for Level III, with a storage time of approximately 31 days; the task intensity factor is calculated accurately based on the actual working time), the task intensity factor is calculated as shown in Table 3 below:

[0086] Table 3. Task intensity factors for each sub-section of the laser gyroscope.

[0087]

[0088] The calculation results in Table 3 show that the storage profile has the highest task intensity factor (0.983), indicating that the performance degradation of the laser gyroscope is the slowest under this profile; while the Level I operation profile has the lowest task intensity factor (0.852), confirming that high-intensity / emergency response level tasks have a significant accelerating effect on the performance degradation of the laser gyroscope.

[0089] Step 2: Integrate the task intensity factor into the reasoning mechanism of the confidence rule base to construct a laser gyroscope performance prediction model that considers the task profile.

[0090] As the background art points out, the traditional belief rule base BRB has limitations in handling dynamic task profile problems. This invention proposes to incorporate the task intensity factor in step 1 into the BRB inference mechanism to construct a laser gyroscope performance prediction model that considers the task profile, specifically including the following five aspects:

[0091] First, the model structure is defined: this laser gyroscope performance prediction model is a confidence rule base model, and its first... Rules Represented as:

[0092]

[0093] The meanings of the parameters in the formula are as follows:

[0094] This is the monitoring data for the performance prediction indicators of the laser gyroscope;

[0095] This is a reference level for the indicator;

[0096] Number of rules;

[0097] The predicted performance value of the gyroscope The corresponding reference level;

[0098] Confidence level;

[0099] N represents the number of performance reference levels;

[0100] For the first The rule weight of each rule;

[0101] Sub-section Task intensity factor;

[0102] The weights are the indicator weights.

[0103] Second, the selection and reference value setting of performance prediction indicators: The performance degradation of a laser gyroscope can be reflected by changes in its error coefficients. By performing frequency difference analysis on the angular velocity pulse voltage output by the gyroscope, the zero-order drift coefficient and the first-order drift coefficient (scaling factor) can be obtained. The analysis shows that these two coefficients vary greatly throughout the entire life cycle of the laser gyroscope and can effectively reflect the performance degradation trend, while the error coefficients of other indicators vary less, and the installation error has been compensated for at the initial design of the laser gyroscope. Therefore, this embodiment ultimately determines the zero-order drift coefficient and the first-order drift coefficient as performance prediction indicators.

[0104] After obtaining the task intensity factor of the laser gyroscope, and considering the actual situation described above, five reference levels were set for the zero-order drift coefficient and the first-order drift coefficient, namely "VL, L, M, H, VH". The reference values ​​were determined by experts based on the actual variation range of the data, as shown in Table 4 below. Similarly, the performance states of the laser gyroscope were set to three levels: "L, M, H", with reference values ​​shown in Table 5 below.

[0105] Table 4 Reference values ​​for the performance indicators of this laser gyroscope

[0106]

[0107] Table 5 Reference values ​​for the performance status of this laser gyroscope

[0108]

[0109] In summary, the new BRB model incorporating task intensity factors in this invention consists of a total of 25 rules (5 input levels × 5 input levels), as shown in Table 6 below. The initial parameter settings are as follows: rule weights. All parameters are set to 1. Meanwhile, based on actual conditions, it is known that the zero-order drift coefficient is better able to reflect the performance change trend of the gyroscope than the first-order drift coefficient. Therefore, the index weight corresponding to the zero-order drift coefficient is 1, and the index weight corresponding to the first-order drift coefficient is 0.8. Other initial parameters (such as confidence level) are determined based on expert knowledge.

[0110] Table 6 Parameters of the Initial New BRB

[0111]

[0112] Thirdly, information transformation and rule matching degree calculation:

[0113] In the new BRB model, the acquired performance metric data is transformed using the following formula:

[0114]

[0115] The meanings of the parameters in the formula are as follows:

[0116] For the first After information transformation, the performance prediction index is relative to the first... The degree of matching in the rule;

[0117] and It is the first The performance prediction metric in the first Article and Section The corresponding reference level in the rule and The reference values ​​are provided, and the specific values ​​are determined by experts based on actual value changes and experience.

[0118] Fourthly, rule activation: After obtaining the matching degree of each indicator through the above information transformation, the task intensity factor is combined. First calculate the first Total matching degree of the rules :

[0119]

[0120] This invention uses a task intensity factor The matching degree is directly applied to each indicator in the form of a product. Due to the task intensity factor Within the range of 0-1, when the mission profile strength is high ( When the value is relatively small, the matching degree of all indicators will be weakened proportionally, thus quantifying the impact of the task profile on performance degradation during the rule activation phase.

[0121] Based on the total matching degree Calculate the first Activation weight of the rule :

[0122]

[0123] The meanings of the parameters in the formula are as follows:

[0124] For the first The performance prediction metric in the first The degree of matching in the rule;

[0125] For the first The weight of each indicator;

[0126] For the first The rule weight of each rule.

[0127] Fifth, rule reasoning: After each rule is activated, the Evidential Reasoning (ER) algorithm is used to fuse each activated rule and calculate the performance prediction value. The calculation formula is as follows:

[0128]

[0129] The meanings of the parameters in the formula are as follows:

[0130] The predicted performance status of the laser gyroscope is presented as a score, which is a specific numerical value.

[0131] For reference level The utility value;

[0132] The relative reference level calculated using the evidence reasoning algorithm The overall confidence level;

[0133] These are intermediate variables in the reasoning process;

[0134] To simplify the expression, Actual representative , Actual representative .

[0135] Step 3: Optimize the parameters of the laser gyroscope performance prediction model using the covariance matrix adaptive evolution strategy (P-CMA-ES).

[0136] The laser gyroscope performance prediction model established based on steps 1 and 2 may have biases due to the presence of initial parameters (including rule weights, confidence levels, and index weights) determined by experts and the need to adjust the initial attenuation coefficient according to actual conditions. Therefore, this invention needs to optimize these parameters to improve prediction accuracy.

[0137] Specifically, with the goal of minimizing the mean square error between the predicted output and the actual value of the laser gyroscope performance prediction model, the following parameter optimization model is constructed:

[0138]

[0139] In the formula: The parameters to be optimized are, respectively, rule weight, confidence level, indicator weight, and decay coefficient. for Real-time performance status label value; The number of data points.

[0140] Furthermore, after determining the above-mentioned optimization model, the present invention employs an adaptive evolution strategy for the covariance matrix with a projection mechanism to optimize the parameter optimization model, specifically including the following steps:

[0141] Step 3.1: Determine initial values: Set initial parameters Their meanings are respectively the initial parameter vector, the initial covariance matrix, the initial step size, the population size, the offspring population size, and the optimal solution; where: , representing the initial mean;

[0142] Step 3.2, Generate the initial population: In the... Generation, generated through normal distribution sampling One candidate solution:

[0143]

[0144] In the formula: It is the first The first generation There are 10 candidate solutions; The representative value is 0 and the covariance is The normal distribution;

[0145] Step 3.3: Project candidate solutions to the constraint hyperplane: Project the candidate solutions... Project onto a hyperplane that satisfies the constraints, ensuring that the solution satisfies the physical constraints:

[0146]

[0147] In the formula: For parameter matrices; Representative at The number of variables in the subspace; T represents transpose, and j represents the j-th group of variables or the j-th subspace;

[0148] Step 3.4: Select the best and update the mean: Select the offspring with the best fitness from the population. Each solution updates the mean to guide the next generation of search centers:

[0149]

[0150] In the formula: These are the weighting coefficients. Representative from the first The selected scheme in the generation;

[0151] Step 3.5: Update the covariance matrix: By dynamically adjusting the covariance matrix, historical information is balanced with the current population distribution.

[0152]

[0153] In the formula: The learning rate controls the weights of historical covariance and current information. As the evolutionary path, the covariance matrix is ​​adaptively adjusted to make the search direction gradually align with the steepest descent direction of its objective function;

[0154] Step 3.6, Iteration: Repeat steps 3.2 to 3.5 until the termination condition is met. Generally, the termination condition is reaching the preset maximum number of iterations.

[0155] Step 4: Based on the laser gyroscope performance prediction model optimized in Step 3, predict the performance status of the laser gyroscope under a specific task profile.

[0156] To verify the effectiveness of the method of the present invention, an X-axis laser gyroscope in a laser inertial navigation system of a certain type of aircraft was used as an example. 180 sets of zero-order and first-order drift coefficient data were obtained during the life cycle of the laser gyroscope (which experienced a total of storage profile, Level I operational profile, and Level III operational profile), as follows: Figure 2 , 3 As shown, the 90 sets of data with odd numbers are used as the training set for training the model and optimizing parameters, while the 90 sets of data with even numbers are used as the test set for validating model performance.

[0157] Two models were selected to perform performance predictions on the test set data, as detailed below:

[0158] 1. Traditional Comparison Example: Traditional BRB model optimized by P-CMA-ES algorithm (without introducing factors affecting the task profile);

[0159] 2. Embodiment of the present invention: A new BRB model optimized by the P-CMA-ES algorithm (incorporating task intensity factor).

[0160] The experiment set the optimization algorithm to iterate 300 times, and the prediction results are as follows: Figure 4 As shown; using mean squared error (MSE) as the evaluation metric for model prediction performance, the final evaluation results of the two models are as follows:

[0161] Traditional Comparative Example: The traditional BRB model optimized by P-CMA-ES has an MSE of 1.6 × 10⁻⁶. -2 ;

[0162] Embodiment of the present invention: A new BRB model incorporating a task intensity factor and optimized by P-CMA-ES, with an MSE of 4.3 × 10⁻⁶. -3 .

[0163] Test results show that the new BRB model, which incorporates a task intensity factor and is optimized using the P-CMA-ES algorithm, reduces the mean squared error (MSE) of prediction from 1.6 × 10⁻⁶ compared to the traditional BRB model optimized using P-CMA-ES. -2 Reduced to 4.3×10 -3 The prediction accuracy is improved by nearly an order of magnitude, and the prediction results are basically close to the true values, which fully verifies the reliability and accuracy of the prediction method of the present invention.

[0164] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention.

[0165] It should be understood that the present invention is not limited to the content already described above, and various modifications and changes can be made without departing from its scope. The scope of the present invention is limited only by the appended claims.

Claims

1. A method for predicting the performance of a laser gyroscope considering a mission profile, characterized in that, Includes the following steps: Step 1: Calculate the task intensity factor, which characterizes the degree of performance degradation of each sub-profile, based on the different task sub-profiles experienced by the laser gyroscope. Step 2: Integrate the task intensity factor into the reasoning mechanism of the confidence rule base to construct a laser gyroscope performance prediction model that considers the task profile; Step 3: Optimize the parameters of the laser gyroscope performance prediction model using a covariance matrix adaptive evolution strategy; Step 4: Based on the optimized laser gyroscope performance prediction model, predict the performance status of the laser gyroscope under a specific task profile.

2. The laser gyroscope performance prediction method considering mission profile according to claim 1, characterized in that, In step 1, the different task sub-profiles include storage profiles and operation profiles; The work profile is divided into Level I, Level II and Level III work profiles according to the task intensity.

3. The laser gyroscope performance prediction method considering mission profile according to claim 2, characterized in that, In step 1, the formula for calculating the task intensity factor is: In the formula: Sub-section Task intensity factor These correspond to the Level I operation profile, Level II operation profile, Level III operation profile, and storage profile, respectively. For the laser gyroscope in the sub-section Maximum operating time; The attenuation coefficient has a range of values. ; and The laser gyroscope enters and exits the sub-sections respectively. Time cutoff points; and Sub-sections The temperature discount factor and the power discount factor of the laser gyroscope are described below.

4. The laser gyroscope performance prediction method considering mission profile according to claim 1, characterized in that, In step 2, the laser gyroscope performance prediction model is a confidence rule base model, and its first... Rules Represented as: In the formula: The performance prediction index of the laser gyroscope; This is a reference level for the indicator; The number of rules; The predicted performance value of the gyroscope The corresponding reference level; 1. Confidence level; N is the number of performance reference levels; For the first The rule weight of each rule; Sub-section Task intensity factor; The weights are the indicator weights.

5. The laser gyroscope performance prediction method considering mission profile according to claim 4, characterized in that, The performance prediction metrics include the zero-order term drift coefficient and the first-order term drift coefficient.

6. The laser gyroscope performance prediction method considering mission profile according to claim 4, characterized in that, Step 2, which integrates the task strength factor into the reasoning mechanism of the confidence rule base, specifically includes the rule activation step: First, combine the aforementioned task intensity factor Calculate the first Total matching degree of the rules : Calculate the next Activation weight of the rule : In the formula: For the first The performance prediction metric in the first The degree of matching in the rule; For the first The weight of each indicator; For the first The rule weight of each rule.

7. The laser gyroscope performance prediction method considering mission profile according to claim 6, characterized in that, Step 2 also includes a rule reasoning step: an evidence-based reasoning algorithm is used to fuse all activated rules and calculate the performance prediction value. The calculation formula is as follows: In the formula: This is a predicted value for the performance status of the laser gyroscope. For reference level The utility value; The relative reference level calculated using the evidence reasoning algorithm The overall confidence level; These are intermediate variables in the reasoning process; for the sake of simplification, Actual representative , Actual representative .

8. The laser gyroscope performance prediction method considering mission profile according to claim 1, characterized in that, In step 2, the initial parameters of the laser gyroscope performance prediction model are determined based on expert knowledge.

9. The laser gyroscope performance prediction method considering mission profile according to claim 1, characterized in that, In step 3, with the goal of minimizing the mean square error between the predicted output and the actual value of the laser gyroscope performance prediction model, the following parameter optimization model is constructed: In the formula: The parameters to be optimized are, respectively, rule weight, confidence level, indicator weight, and decay coefficient. for Real-time performance status label value; The number of data points.

10. The laser gyroscope performance prediction method considering mission profile according to claim 9, characterized in that, In step 3, the parameter optimization model is optimized using a covariance matrix adaptive evolution strategy with projection mechanism, including the following steps: Step 3.1: Determine initial values: Set initial parameters Their meanings are respectively: initial parameter vector, initial covariance matrix, initial step size, population size, offspring population size, and optimal solution; Step 3.2, Generate the initial population: In the... Generation, generated through normal distribution sampling One candidate solution: In the formula: It is the first The first generation There are 10 candidate solutions; The representative value is 0 and the covariance is The normal distribution; Step 3.3: Project candidate solutions to the constraint hyperplane: Project the candidate solutions... Project onto a hyperplane that satisfies the constraints, ensuring that the solution satisfies the physical constraints: In the formula: For parameter matrices; Representative at The number of variables in the subspace; T represents transpose, and j represents the j-th group of variables or the j-th subspace; Step 3.4: Select the best and update the mean: Select the offspring with the best fitness from the population. Each solution updates the mean to guide the next generation of search centers: In the formula: These are the weighting coefficients. Representative from the first The selected scheme in the generation; Step 3.5: Update the covariance matrix: By dynamically adjusting the covariance matrix, historical information is balanced with the current population distribution. In the formula: The learning rate; For evolutionary paths; Step 3.6, Iteration: Repeat steps 3.2 to 3.5 until the termination condition is met.

Images