Parameter identification method that transmits tm release residual velocity estimation deviation <20%

Abstract

The present application relates to the technical field of data processing, and especially relates to a parameter identification method for releasing residual velocity estimation deviation of less than 20% of TM. First, in the global range, the values of each group of to-be-identified parameters in the inertia sensor test quality release residual velocity estimation model are randomly initialized as the initial individual position in the population; and a plurality of dominant individuals are selected therefrom as the learning starting point; then, the learning ability is updated to make each individual generate a new solution around the dominant individual in different ranges, so as to take into account the global search ability and the local optimization ability. The method provided by the present application reduces the dependence on a single starting point, improves the stability and applicability of the algorithm, and ensures that the test quality release residual velocity estimation error is less than 20%.

Description

Technical Field

[0001] This invention belongs to the field of data processing technology, and in particular relates to a parameter identification method for inertial telemetry (IT) release residual velocity estimation deviation <20%. Background Technology

[0002] In conventional inertial sensor testing, the identification of mass release estimation model parameters is often obtained by technicians through multiple adjustments and attempts based on actual conditions and experience. This process is inefficient and has a low success rate in parameter identification.

[0003] With the continuous development of machine learning, those skilled in the art have begun to combine swarm intelligence optimization algorithms with parameter identification in inertial sensor test mass release estimation models, hoping to improve the success rate and efficiency of parameter identification. Among them, Particle Swarm Optimization (PSO) has been widely used in engineering optimization problems due to its simple structure, few parameters, and ease of implementation. This algorithm searches by simulating the cooperative behavior of the swarm, achieving a certain balance between global search and local optimization. However, most existing swarm optimization algorithms adopt a fixed update strategy, and the individual search behavior is relatively simple. Especially in complex nonlinear problems, they are easily affected by the initial population distribution and local extrema, causing the search process to get stuck in local optima. At the same time, traditional algorithms usually guide the population update only through a single optimal individual, lacking a multi-dominant individual cooperative search mechanism, making it difficult to balance global search capability and local fine search capability. In addition, in terms of search range control, existing methods mostly use fixed parameters and cannot dynamically adjust the search scale according to the optimization process, thus limiting the convergence efficiency and stability of the algorithm in complex parameter identification problems.

[0004] Therefore, this paper proposes a population optimization algorithm that can simulate the dynamic changes in learning ability, has multi-advantage individual collaborative search, and an adaptive search range adjustment mechanism. This algorithm is of great significance for improving the test quality of inertial sensors and releasing the accuracy of parameter identification of estimation models. Summary of the Invention

[0005] In view of this, the present invention aims to provide a parameter identification method with an inertial telemetry (IT) release residual velocity estimation deviation of <20%. First, an initial population is randomly generated globally, and multiple dominant individuals are selected as learning starting points. Then, the model is updated through learning ability so that each individual generates new solutions around the dominant individuals in different ranges, thereby taking into account both global search ability and local optimization ability.

[0006] To achieve the above objectives, the technical solution created by this invention is implemented as follows: A parameter identification method for inertial telemetry (ITT) release residual velocity estimation error <20%, comprising: S1: On a global scale, the values ​​of each set of parameters to be identified in the residual velocity estimation model of the test mass release of the inertial sensor are randomly initialized as the initial individual positions in the population; S2: Based on the current individual position, construct a learning ability update model and calculate the fitness of each individual position under the initial value; S3: Based on the fitness obtained in step S2, sort the positions of individuals in the population, select multiple dominant individual positions as learning starting points, and record the current global best individual position; S4: Using the dominant individual position obtained in step S3 as the local search center, update the model using the learning ability constructed in step S2 to generate a new generation of individual positions of dominant individuals within the current search range. S5: Calculate the fitness value corresponding to the position of the new generation of individuals in step S4, and adjust the search range corresponding to the current position of each dominant individual. S6: Based on the fitness updated in step S5, repeat steps S2 to S3 until the number of iterations exceeds the set maximum iteration value, and output the global optimal individual position and value at this time.

[0007] Furthermore, in step S2: the fitness is calculated as the variance between the estimated displacement of the released pin and the true value curve based on the identification value; the learning ability update model is: ; Where C represents the current learning ability, C max The threshold for learning ability is represented by t, and the learning age at an individual's location is represented by t. max σ represents the peak age of learning ability at an individual location, and σ represents the distribution parameter of learning ability at an individual location.

[0008] Furthermore, in step S3, all individual positions are sorted from best to worst fitness, and the positions of several top-ranked dominant individuals are selected as the starting point for learning; each dominant individual position corresponds to a local search center.

[0009] Furthermore, in step S4, the location of the next generation of individuals is generated using the following formula: X k =X0+C×rand(-L,L); Among them, X k Let X0 represent the position of the kth new generation individual, C represent the current learning ability, X0 represent the current dominant individual position, L represent the current search range boundary, and rand represent the function for generating random numbers.

[0010] Furthermore, the process of adjusting the search range corresponding to the current advantageous individual positions in step S5 includes: when the global optimal individual position has not changed for a preset number of consecutive times, the current search range boundary is reduced: if the current search range boundary is greater than a preset threshold, it is reduced by a preset first step length; if the current search range boundary is less than a preset threshold, it is reduced by a preset second step length.

[0011] Furthermore, the first step is longer than the second step.

[0012] Compared with the prior art, the present invention can achieve the following beneficial effects: The parameter identification method for achieving a release residual velocity estimation error of <20% in the inertial telemetry (ITT) system, as described in this invention, firstly randomly generates an initial population globally and selects multiple dominant individuals as learning starting points. Subsequently, the learning ability update model enables each individual to generate new solutions around the dominant individuals within different ranges, thus balancing global search capability and local optimization capability. During the iteration process, this invention adaptively controls the search range parameter by monitoring the continuous changes in the global optimal solution. When no better solution is obtained after multiple iterations, the search range is dynamically narrowed to improve local fine-grained search capability and enhance the ability to escape local extrema. Compared to traditional population optimization algorithms, this invention employs a multi-dominant individual collaborative search mechanism, reducing dependence on a single starting point, improving algorithm stability and applicability, and ensuring that the test quality release residual velocity estimation error is below 20%. Attached Figure Description

[0013] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments and descriptions of the invention are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings: Figure 1 A flowchart illustrating the parameter identification method for inertial telemetry release residual velocity estimation deviation <20% as described in an embodiment of the present invention; Figure 2 The flowchart illustrates the parameter identification method for inertial telemetry (ITT) release residual velocity estimation deviation <20% as described in the embodiments of the present invention. Detailed Implementation

[0014] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and do not constitute a limitation thereof. It should be noted that, unless otherwise specified, the embodiments and features described herein can be combined with each other.

[0015] In the description of this invention, it should be understood that the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Therefore, a feature defined with "first," "second," etc., may explicitly or implicitly include one or more of that feature. In the description of this invention, unless otherwise stated, "a plurality of" means two or more.

[0016] The invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0017] like Figure 1 and Figure 2 As shown in the embodiment of the present invention, the parameter identification method for inertial telemetry release residual velocity estimation deviation <20% includes: S1: On a global scale, the values ​​of each set of parameters to be identified in the inertial sensor test mass release residual velocity estimation model are randomly initialized as the initial individual positions in the population.

[0018] S2: Based on the current individual position, construct a learning ability update model and calculate the fitness of each individual position under the initial value.

[0019] In some embodiments, the fitness in step S2 is: calculating the variance between the estimated displacement of the released pin and the true value curve based on the identification value; the learning ability of individuals in the population is described by a Gaussian function, that is, the learning ability update model in step S2 is: ; Where C represents the current learning ability, C max The threshold for learning ability is represented by t, and the learning age at an individual's location is represented by t. max σ represents the peak age of learning ability at an individual's location, and σ represents the distribution parameter of learning ability at that individual's location. In this embodiment of the invention, the distribution parameter σ of learning ability is generated randomly within a preset range to simulate the differences in learning ability under different parameter dimensions.

[0020] S3: Based on the fitness obtained in step S2, sort the positions of individuals in the population, select multiple dominant individual positions as learning starting points, and record the current global best individual position.

[0021] In some embodiments, all individual positions are sorted from best to worst fitness, and several top-ranked dominant individual positions are selected as learning starting points. Each dominant individual position corresponds to a local search center. Local search centers are used to enhance the ability to escape local extreme regions. Specifically, there are three dominant individual positions, each determined according to fitness ranking and corresponding to a different search range.

[0022] S4: Using the dominant individual position obtained in step S3 as the local search center, update the model with the learning ability constructed in step S2 to generate a new generation of individual positions of the dominant individual positions within the current search range.

[0023] In some embodiments, the next generation of individual locations is generated in step S3 using the following formula: X k =X0+C×rand(-L,L); Among them, X k Let X0 represent the position of the k-th new generation individual, X0 represent the current dominant individual position, L represent the current search range boundary, and rand represent the function for generating random numbers. In this embodiment of the invention, each dominant individual corresponds to a different search range, for example, the first dominant individual corresponds to 0.7L, and the second dominant individual corresponds to 1.5L, to form a multi-scale search mechanism.

[0024] S5: Calculate the fitness value corresponding to the position of the new generation of individuals in step S4, and adjust the search range corresponding to the current dominant individual positions.

[0025] In some embodiments, the boundary of the current search range is adaptively adjusted by monitoring the change status of the global optimal individual position. Specifically, when the global optimal individual position has not changed for a preset number of consecutive times, the boundary of the current search range is reduced to improve the local search accuracy; if the boundary of the current search range is greater than a preset threshold, it is reduced by a preset first step size; if the boundary of the current search range is less than the preset threshold, it is reduced by a preset second step size.

[0026] In some embodiments, the first step is longer than the second step.

[0027] In a specific embodiment of the present invention, the preset threshold is specifically set to 0.25, the first step length is specifically set to 0.01, and the second step length is specifically set to 0.0001. The current search range boundary is reduced using the following formula: ; Where L represents the current search range boundary; L' represents the reduced search range boundary. This approach maintains a large search range in the early stages of the algorithm's search to enhance global search capabilities; in the later stages, the search range is gradually reduced, concentrating candidate individuals near dominant individuals for refined local searches, thereby improving parameter identification accuracy and algorithm stability.

[0028] S6: Based on the fitness updated in step S5, repeat steps S2~S3 iteratively until the number of iterations exceeds the set maximum iteration value, and output the global optimal individual position and value at this time. The individual position obtained at this time (i.e., the value of the parameter to be identified in the inertial sensor test mass release residual velocity estimation model obtained at this time) can make the inertial sensor test mass release residual velocity estimation deviation less than 20%.

[0029] This invention simulates the evolution of learning ability, combines a multi-advantageous individual collaborative search strategy with an adaptive search range adjustment mechanism, dynamically adjusts the search range during the iteration process, and guides the generation of a new generation of candidate solutions through multiple advantageous individuals, thereby achieving a balance between global search and local fine optimization, and effectively improving the accuracy and stability of parameter identification.

[0030] Furthermore, to demonstrate that the parameters determined by the method provided by this invention can reduce the estimation deviation of the residual velocity of the inertial test mass release to less than 20%, the following verification experiment was conducted.

[0031] First, to ensure the relative deviation of the residual velocity estimation is below 20% in order to guarantee the quality of the test, the relative deviation of the displacement curve of the release pin must be strictly controlled within 10%. This constraint is determined by the release dynamics mechanism, the error propagation law, and the system modeling characteristics, and is a necessary accuracy condition for achieving high-precision residual velocity estimation. A detailed explanation is given below.

[0032] As can be seen from the test mass release dynamic model, the motion displacement of the release pin is the core state variable describing the dynamic response of the mechanism, and its motion acceleration is obtained by the second derivative of the displacement curve with respect to time: ; Among them, a RT (t) represents the acceleration of the ejector pin at time t when it is released, X RT (t) represents the displacement of the ejector pin at time t after release; The force acting on the test mass is directly related to the acceleration of the release pin, while the residual velocity of the released test mass is obtained by integrating the acceleration of the test mass over the release time period. ; Among them, a TM (t) represents the acceleration of the test mass at time t, and the acceleration a of the released pin. RT (t) is proportional, where T represents the total duration of the measured displacement time series data of the released ejector pin, v TM (t) represents the residual velocity at time t. Therefore, the accuracy of the residual velocity estimation depends entirely on the accuracy of the acceleration estimation, which in turn depends entirely on the accuracy of the displacement curve of the released pin.

[0033] Define the relative deviation of displacement estimation as: ; in, This indicates the relative deviation of the displacement estimate. This represents the estimated displacement of the released ejector pin at time t. Since acceleration is the second derivative of displacement, under the conditions of small deviation and linearization approximation, the relative deviation of acceleration satisfies: ; in, Indicates the relative deviation of acceleration. This represents the estimated value of the acceleration corresponding to the release of the ejector pin at time t; That is, the displacement deviation is approximately amplified to twice its original value after second-order differentiation. The residual velocity is obtained by integrating the acceleration; the integration operation does not change the order of magnitude of the relative deviation, therefore the relative velocity deviation is approximately equal to the relative acceleration deviation. ; in, Indicates the relative deviation of speed. The estimated value of the residual velocity at time t; Combining the above equations, we can obtain the core transmission relationship between displacement deviation and velocity deviation: ; Constraints Substituting into the above equation, we get Further solving yields In summary, to ensure test quality, the relative deviation of the estimated residual velocity must be less than 20%, and the relative deviation of the estimated displacement curves of the release end and the release pin must be less than 10%.

[0034] In this embodiment of the invention, 100 sets of measured displacement timing data of the release pin and step drive voltage signal are randomly generated as true values ​​to be identified. At the same time, an algorithm is used for identification. Substituting the identified value and the true value into the model, two output values ​​can be obtained. Using the RMSE between the two as the evaluation function, the maximum relative deviation of the true displacement value of the release pin can be calculated to be 1.6% through the 100 sets of identification results. Therefore, this method can ensure that the estimation error of the residual velocity of the release pin is less than 20%.

[0035] It should be understood that the various forms of processes shown above can be used to reorder, add, or delete steps. For example, the steps described in this invention disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this invention can be achieved, and this is not limited herein.

[0036] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A parameter identification method for inertial telemetry (ITT) release residual velocity estimation deviation <20%, characterized in that, include: S1: On a global scale, the values ​​of each set of parameters to be identified in the residual velocity estimation model of the test mass release of the inertial sensor are randomly initialized as the initial individual positions in the population; S2: Based on the current individual position, construct a learning ability update model and calculate the fitness of each individual position under the initial value; S3: Based on the fitness obtained in step S2, sort the positions of individuals in the population, select multiple dominant individual positions as learning starting points, and record the current global best individual position; S4: Using the dominant individual position obtained in step S3 as the local search center, update the model using the learning ability constructed in step S2 to generate a new generation of individual positions of the dominant individual positions within the current search range. S5: Calculate the fitness value corresponding to the position of the new generation of individuals in step S4, and adjust the search range corresponding to the current position of each dominant individual. S6: Based on the fitness updated in step S5, repeat steps S2 to S3 until the number of iterations exceeds the set maximum iteration value, and output the global optimal individual position and value at this time.

2. The parameter identification method for inertial telemetry (ITT) release residual velocity estimation deviation <20% according to claim 1, characterized in that, In step S2: The fitness is the variance between the estimated displacement of the release pin calculated based on the identification value and the true value curve. The learning ability update model is as follows: ； Where C represents the current learning ability, C max The threshold for learning ability is represented by t, and the learning age at an individual's location is represented by t. max σ represents the peak age of learning ability at an individual location, and σ represents the distribution parameter of learning ability at an individual location.

3. The parameter identification method for inertial telemetry (ITT) release residual velocity estimation deviation <20% according to claim 1, characterized in that, In step S3, all individual positions are sorted from best to worst fitness, and the positions of several top-ranked dominant individuals are selected as the starting point for learning; each dominant individual position corresponds to a local search center.

4. The parameter identification method for inertial telemetry (ITT) release residual velocity estimation deviation <20% according to claim 1, characterized in that, In step S4, the location of the next generation of individuals is generated using the following formula: X k =X0+C×rand(-L,L); Among them, X k Let X0 represent the position of the kth new generation individual, C represent the current learning ability, X0 represent the current dominant individual position, L represent the current search range boundary, and rand represent the function for generating random numbers.

5. The parameter identification method for inertial telemetry (ITT) release residual velocity estimation deviation <20% according to claim 1, characterized in that, The process of adjusting the search range corresponding to the current positions of each dominant individual in step S5 includes: When the global optimal individual position remains unchanged for a preset number of consecutive iterations, the boundary of the current search range is reduced: If the current search range boundary is greater than the preset threshold, then decrease it by the preset first step length; If the current search range boundary is less than the preset threshold, then decrease it by the preset second step size.

6. The parameter identification method for inertial telemetry (ITT) release residual velocity estimation deviation <20% according to claim 5, characterized in that, The first step is longer than the second step.

Images