A rotary inertial navigation digital twin system and a construction method thereof

Abstract

The present application relates to a kind of rotary inertial navigation digital twin systems and its construction method, including rotary inertial navigation digital twin physical model layer, rotary inertial navigation digital twin geometric model layer, multi-source information data acquisition layer, decision control layer, diagnosis and prediction analysis layer.Can complete the data information acquisition and transmission of rotary inertial navigation system entity, realize the association and interaction of twin virtual body and entity, can be realized using decision control module Iterative optimization of model parameters, finally in diagnosis and prediction analysis module realizes system overall operation state and the performance prediction and evaluation of digital twin inertial navigation system.

Description

Technical Field

[0001] This invention belongs to the field of inertial navigation technology, specifically relating to a rotating inertial navigation digital twin system and its construction method. Background Technology

[0002] With the rapid development of emerging technologies such as artificial intelligence, big data, and the Internet, the intelligent development of equipment has a reliable foundation. Digital twin technology, as one of the key technologies for the intelligent development of equipment, can realize the synchronous coexistence of digital models and physical entities or processes in the digital space, enabling proactive intervention and dynamic prediction of the performance status of physical entities.

[0003] Currently, digital twins are widely used in the intelligent manufacturing industry, but their application in the field of inertial navigation is relatively limited. For example, the research on the application of digital twin technology in inertial systems published by Ma Yitong et al. [J], Missile and Space Launch Vehicle Technology, 6:37-45; and the inertial navigation big data analysis and application based on digital twin technology [J]. Network Security and Data Governance, 2023, 42(S1):76-81, both propose the application of digital twin technology throughout the entire product lifecycle in inertial systems, but do not provide specific implementation plans for equipment-level digital twins.

[0004] To this end, this invention proposes a method for constructing a rotating inertial navigation digital twin system at the equipment level, in order to optimize, make decisions, control and maintain rotating inertial navigation products, maximize the value of data and models at the equipment level and promote the digital development of rotating inertial navigation equipment. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention proposes a rotating inertial navigation digital twin system and its construction method.

[0006] One of the above-mentioned objectives of the present invention is achieved by the following technical solution:

[0007] A rotating inertial navigation digital twin system includes a rotating inertial navigation digital twin physical model layer, a rotating inertial navigation digital twin geometric model layer, a multi-source information data acquisition layer, a decision control layer, and a diagnosis and predictive analysis layer;

[0008] The rotating inertial navigation digital twin geometric model layer: adopts a data-driven method to realize the dynamic representation of the inertial navigation operation process, ensuring synchronous coexistence with the inertial navigation entity;

[0009] The rotating inertial navigation digital twin physical model layer includes a fiber optic gyroscope, accelerometer, grating, servo system, and rotating inertial navigation solution model.

[0010] Multi-source information data acquisition layer: contains a variety of sensors for real-time acquisition of current, voltage, acceleration, angular velocity, temperature, stress and strain information. At the same time, it provides the solution data inside the rotating inertial navigation system for interaction with the twin model, forming an effective dataset that is fed back to the digital twin geometric and physical models.

[0011] The decision control layer: based on the digital twin physical model, it visualizes and detects the operation of the inertial navigation system by combining real-time multi-source data; based on measured data, it makes real-time predictions of the inertial navigation system by establishing a dynamic mapping between physical entities and physical models; and based on the performance prediction results, it optimizes the relevant parameters of the inertial navigation system.

[0012] Diagnostic and Predictive Analysis Layer: This layer extracts features from the raw data acquired by the multi-source data acquisition layer and performs frequency domain analysis, time domain analysis, statistical analysis, or other analyses related to the operating status of the rotating inertial navigation system. It then compares these analyses with set thresholds or fault characteristics for monitoring and alarm purposes. Simultaneously, it compares the real-time performance prediction results from the decision control layer with the gyroscope temperature compensation data, accelerometer temperature compensation data, and navigation calculation data calculated internally by the rotating inertial navigation system to evaluate the actual performance of the inertial navigation system and provides future performance prediction results for reference.

[0013] Furthermore, the basic models of fiber optic gyroscopes and accelerometers are as follows:

[0014]

[0015] ω out The angular velocity value is the output of the gyroscope, I is the identity matrix, and ΔG is the angular velocity value. ω For gyroscope installation error, δK ω This is the gyroscope scaling factor error. ε is the theoretical value of angular velocity. p The gyroscope has a constant zero bias; a out The acceleration value detected and output by the accelerometer, I is the identity matrix, and ΔG a δK is the accelerometer installation error. g For accelerometer scaling factor error, β is the theoretical value of acceleration. p The accelerometer constant is zero bias.

[0016] Furthermore, in step 1, the temperature compensation models for the scaling factor error, installation error, and zero bias error of the gyroscope and accelerometer are all implemented using a backpropagation (BP) neural network; the relationship between the input and output of the BP neural network is as follows:

[0017] Q i =tansig(W i,1 P i )+bi,1

[0018] O i =purelin(W i,2 Q i )+b i,2

[0019] In the formula: i represents the scaling factor error, installation error, or zero bias error; P i For the system input of scaling factor error, installation error, or zero bias error, i.e., temperature; b i,1 The implicit layer bias is for scaling factor error, installation error, or zero bias error; b i,2 Output layer bias is set for scaling factor error, installation error, or zero bias error; W i,1 W is the hidden layer weight matrix; i,2 Q is the output layer weight matrix; i For hidden layer output; O i The system output represents the scaling factor error, installation error, or zero bias error; tansig is the hyperbolic tangent sigmoid function; purelin is a linear function.

[0020] Furthermore, the servo motor in a servo system typically drives the rotary mechanism directly, and its basic model is as follows:

[0021]

[0022] In the formula, J0 is the moment of inertia of the servo motor shaft, B0 is the rotational damping coefficient of the servo motor, K0 is the rotational stiffness of the servo motor, and θ is the motor rotation angle. The first derivative of the motor's rotation angle, also known as the motor's angular velocity. It is the second derivative of the motor's rotation angle, also known as the motor's angular acceleration.

[0023] Furthermore, the rotating inertial navigation system solution model includes attitude differential equations, velocity differential equations, and position differential equations; among which, the attitude differential equations are:

[0024]

[0025] In the formula, q0, q1, q2, and q3 are the four basic elements of the quaternion equation. The differential of a quaternion, Let be the projection of the angular velocity of the carrier coordinate system b relative to the navigation coordinate system n in three directions of the b system;

[0026] The velocity differential equation is:

[0027]

[0028] In the formula, V nLet the velocity of the vehicle be in the navigation coordinate system. For V n The differential, Let b be the attitude matrix relative to the navigation coordinate system n. Let be the projection of the angular velocity of Earth coordinate system e relative to inertial coordinate system i onto navigation coordinate system n. Let g be the projection of the angular velocity of navigation coordinate system n relative to Earth coordinate system e in navigation coordinate system n. b This is the projection of gravitational acceleration onto the carrier coordinate system b;

[0029] The position differential equation is:

[0030]

[0031] The derivative of the longitude of the carrier's location. Let L be the differential of the latitude of the carrier's location, and h be the altitude of the carrier. R is the differential of the height h of the carrier. M The location of the carrier is at the radius of curvature R along the Earth's meridian. N The carrier's location is at the radius of curvature of the Earth along the circumference of the equinoxes, V. E V N V U These represent the velocity of the carrier in the navigation coordinate system, and its components in the three directions: northeast, south, and north.

[0032] One of the above-mentioned objectives of the present invention is achieved by the following technical solution:

[0033] Step 1: Use 3D modeling software to create a digital twin geometric model of the rotating inertial navigation system;

[0034] Step 2: Establish and simulate a digital twin physical model of the rotating inertial navigation system using digital methods;

[0035] Step 3: In the digital twin platform software, synchronize the geometric model and the physical model to visualize the motion process and input / output results;

[0036] Step 4: Build a data acquisition module to complete data acquisition and transmit it to the digital twin platform for use by the twin model to achieve the association between the virtual twin and the physical entity;

[0037] Step 5: Build a decision control module and use machine learning and deep learning data-driven modeling methods to achieve iterative optimization of model parameters;

[0038] Step 6: Build a diagnostic and predictive analysis module to evaluate the overall operating status of the system and the performance of the digital twin inertial navigation system, and provide a reference.

[0039] Furthermore, in step 1, geometric models of the rotating inertial navigation system, components, and parts are established using 3D modeling software. Assembly and motion relationships are constructed according to the composition of the actual equipment to characterize the relationships between the components. Finite element models of key components of the rotating system, such as motors and bearings, are also established to characterize their stress and deformation.

[0040] Furthermore, step 5 includes:

[0041] Using the graphical display functions of ANSYS Twin Builder and the Unity Manufacturing digital twin platform, the data information collected by the multi-source information acquisition layer is presented in a graphical and visual manner. The input of the digital twin model is determined to be real-time acquired current, voltage, acceleration, angular velocity, temperature, stress and strain information, as well as temperature compensation, shaft angle compensation, and navigation calculation information. Through the digital twin physical model and combined with data-driven modeling methods such as machine learning and deep learning, a twin calculation model is built, and the predicted data of each sensor of the rotating inertial navigation system and the predicted data of navigation calculation performance are output. Based on the prediction results of each sensor and the predicted data of navigation calculation performance, and compared with the subsequently acquired real-time data, the parameters or structure of each digital twin physical model in the rotating inertial navigation system are optimized.

[0042] Furthermore, step 6 involves building the diagnostic and predictive analysis module, which includes: first, model training, preprocessing the raw acquisition information provided by the multi-source data acquisition layer and the characteristic information related to the operating status of the rotating inertial navigation system, including handling missing data and correcting abnormal data, to improve data quality and avoid errors in subsequent data analysis; then, feature extraction is performed on the data, transforming the raw data into representative feature vectors, and using data dimensionality reduction methods such as principal component analysis and linear discriminant analysis to reduce the dimensionality of the high-dimensional original characteristic vectors, thereby improving computational efficiency; finally, the dimensionality-reduced feature vectors are input into the digital twin deep learning model to obtain training and evaluation models.

[0043] Furthermore, in step 6, the collected data is input into the trained deep learning model, and the predicted analysis data is output. This data is then compared with the set thresholds or fault characteristics. If the thresholds are exceeded or fault characteristics are detected, an alarm will be issued in a timely manner. Simultaneously, the real-time performance prediction results performed by the decision control layer are compared with the gyroscope temperature compensation data, accelerometer temperature compensation data, and navigation calculation data calculated internally by the rotating inertial navigation system. Combined with the analysis results of the original collected data, the performance of the digital twin inertial navigation system is evaluated, and the accuracy of the performance prediction results is judged.

[0044] The advantages and positive effects of this invention are as follows:

[0045] This invention provides a rotating inertial navigation digital twin system and its construction method, which can complete the data information acquisition and transmission of the rotating inertial navigation system entity, realize the association and interaction between the virtual twin and the entity, realize the iterative optimization of model parameters through the decision control module, and finally realize the overall system operation status and performance prediction and evaluation of the digital twin inertial navigation system in the diagnosis and prediction analysis module. Attached Figure Description

[0046] Figure 1 This is a schematic diagram of the rotating inertial navigation system based on digital twins according to the present invention;

[0047] Figure 2 This is a schematic diagram of the multi-source data composition of the present invention. Detailed Implementation

[0048] The structure of the present invention will be further described below with reference to the accompanying drawings and embodiments. It should be noted that these embodiments are descriptive and not limiting.

[0049] Please refer to a rotating inertial navigation digital twin system. Figures 1-2 Its inventive features include: a rotating inertial navigation digital twin physical model layer, a rotating inertial navigation digital twin geometric model layer, a multi-source information data acquisition layer, a decision control layer, and a diagnosis and predictive analysis layer. Specific descriptions of each layer are as follows:

[0050] Rotating Inertial Navigation System (INS) Digital Twin Geometric Model Layer: This layer employs a data-driven approach to dynamically represent the INS operation process, ensuring synchronous coexistence with the INS entity. The digital twin geometric model is the foundation of the rotating INS digital twin model.

[0051] Rotating inertial navigation digital twin physical model layer: mainly includes models of fiber optic gyroscopes, accelerometers, gratings, servo systems, and rotating inertial navigation solution.

[0052] The basic models of fiber optic gyroscopes and accelerometers are as follows:

[0053]

[0054] ω out The angular velocity value is the output of the gyroscope, I is the identity matrix, and ΔG is the angular velocity value. ω For gyroscope installation error, δK ω This is the gyroscope scaling factor error. ε is the theoretical value of angular velocity. p The gyroscope has a constant zero bias; a out The acceleration value detected and output by the accelerometer, I is the identity matrix, and ΔG a δK is the accelerometer installation error. g For accelerometer scaling factor error, β is the theoretical value of acceleration.p The accelerometer constant is zero bias;

[0055] Temperature compensation models for scaling factor error, installation error, and zero bias error of gyroscopes and accelerometers all employ backpropagation (BP) neural networks. BP neural networks are an important component of artificial neural networks, containing multiple hidden layers and possessing nonlinear mapping capabilities. They are widely used in approximation, regression, and other application solutions. A BP neural network contains one or more hidden layers and one output layer. Training parameters are sequentially passed from the input node through each hidden layer and the output layer, training the network with samples in a known pattern to generate a precise output-to-output mapping.

[0056] The temperature compensation model based on a BP neural network for scaling factor error, installation error, and zero bias error can be expressed as:

[0057]

[0058] In the formula: i represents the scaling factor error, installation error, or zero bias error; P i For the system input of scaling factor error, installation error, or zero bias error, i.e., temperature; b i,1 The implicit layer bias is for scaling factor error, installation error, or zero bias error; b i,2 Output layer bias is set for scaling factor error, installation error, or zero bias error; W i,1 W is the hidden layer weight matrix; i,2 Q is the output layer weight matrix; i For hidden layer output; O i The system output represents the scaling factor error, installation error, or zero bias error; tansig is the hyperbolic tangent sigmoid function; purelin is a linear function.

[0059] In a servo system, the servo motor typically drives the rotary mechanism directly. Its basic model is as follows:

[0060]

[0061] In the formula, J0 is the moment of inertia of the servo motor shaft, B0 is the rotational damping coefficient of the servo motor, K0 is the rotational stiffness of the servo motor, and θ is the motor rotation angle. The first derivative of the motor's rotation angle, also known as the motor's angular velocity. It is the second derivative of the motor's rotation angle, also known as the motor's angular acceleration.

[0062] The rotating inertial navigation system (INS) solution model includes attitude differential equations, velocity differential equations, and position differential equations. The attitude differential equations are as follows:

[0063]

[0064] In the formula, q0, q1, q2, and q3 are the four basic elements of the quaternion equation. The differential of a quaternion, Let be the projection of the angular velocity of the carrier coordinate system b relative to the navigation coordinate system n in the three directions of the b system.

[0065] The velocity differential equation is:

[0066]

[0067] In the formula, V n Let the velocity of the vehicle be in the navigation coordinate system. For V n The differential, Let b be the attitude matrix relative to the navigation coordinate system n. Let be the projection of the angular velocity of Earth coordinate system e relative to inertial coordinate system i onto navigation coordinate system n. Let g be the projection of the angular velocity of navigation coordinate system n relative to Earth coordinate system e in navigation coordinate system n. b This is the projection of gravitational acceleration onto the carrier coordinate system b.

[0068] The position differential equation is:

[0069]

[0070] The derivative of the longitude of the carrier's location. Let L be the differential of the latitude of the carrier's location, and h be the altitude of the carrier. R is the differential of the height h of the carrier. M The location of the carrier is at the radius of curvature R along the Earth's meridian. N The carrier's location is at the radius of curvature of the Earth along the circumference of the equinoxes, V. E V N V U These represent the velocity of the carrier in the navigation coordinate system, and its components in the three directions: northeast, south, and north.

[0071] Multi-source information data acquisition layer: This layer forms the foundation of the digital twin's data acquisition framework. The rotating inertial navigation system (INS) includes various sensors such as gyroscopes, accelerometers, optical encoders, temperature sensors, current sensors, voltage sensors, and pressure sensors. It can acquire information such as current, voltage, acceleration, angular velocity, temperature, and stress-strain in real time. Simultaneously, it provides the INS with internal calculation data, such as gyroscope temperature compensation data, accelerometer temperature compensation data, and navigation data, for interaction with the twin model, forming an effective dataset that is fed back into the digital twin's geometric and physical models. Data acquisition permeates every aspect of the digital twin INS and is the data foundation of the system.

[0072] Decision and Control Layer: Decision and control is based on the digital twin physical model, combining real-time multi-source data to visualize and monitor the operation of the inertial navigation system (INS). Based on measured data, a dynamic mapping between physical entities and the physical model is established to make real-time predictions of the INS. Based on the performance prediction results, relevant parameters of the INS, such as servo control parameters and inertial component parameters, are optimized. Decision and control evolve in real time with changes in the digital twin model and data acquisition, constituting a complex dynamic process and a key factor in the successful implementation of the digital twin INS.

[0073] Diagnostic and Predictive Analysis Layer: This layer primarily extracts features from the raw data acquired by the multi-source data acquisition layer, such as current, voltage, acceleration, angular velocity, temperature, stress, and strain. It then performs frequency domain analysis, time domain analysis, statistical analysis, or other analyses related to the operating status of the rotating inertial navigation system. These analyses are compared with set thresholds or fault characteristics for monitoring and alarm purposes. Simultaneously, the real-time performance prediction results from the decision control layer are compared with the gyroscope temperature compensation data, accelerometer temperature compensation data, and navigation calculation data calculated internally by the rotating inertial navigation system to evaluate the actual performance of the inertial navigation system and provide future performance predictions for reference.

[0074] The specific implementation process of the above-mentioned rotating inertial navigation digital twin system and its corresponding module layers is as follows:

[0075] 1. Establish digital twin geometric models: Use 3D modeling software such as Solidworks, ProE, and Unity3D to establish geometric models of the rotating inertial navigation system, components (rotating system, inertial components, etc.), and parts (motors, bearings, fiber optic gyroscopes, accelerometers, etc.). Assemble and construct motion relationships according to the actual equipment composition to characterize the relationships between various components. In addition, establish finite element models of key components of the rotating system such as motors and bearings to characterize their stress and deformation.

[0076] 2. Establish a digital twin physical model: The physical model, also known as the mechanism model, is used to establish and simulate the control logic and input-output relationship of the actual system. It can characterize the performance of the rotating inertial navigation entity. The main models include fiber optic gyroscopes, accelerometers, gratings, servo systems, and rotating inertial navigation solutions. The establishment of the physical model is mainly to accurately characterize the performance of the rotating inertial navigation entity and to ensure the real-time synchronous evolution of the physical model and the entity as much as possible.

[0077] 3. In digital twin platforms such as ANSYS Twin Builder and Unity Manufacturing, the integrated environment is used to synchronize the geometric model and the physical model, ensuring the exchange and synchronization of data between different modules. At the same time, the platform can be used to build behavioral models and visualize the motion process and input / output results.

[0078] 4. Set up a data acquisition module: such as Figure 2 As shown, information related to the rotating inertial navigation system (INS) is collected and transmitted to the digital twin platform. The basic components of the INS include a gyroscope, accelerometer, grating encoder, and temperature sensor, with the addition of electrical data acquisition units such as current sensors, voltage sensors, and stress sensors. This allows for real-time acquisition of information such as current, voltage, acceleration, angular velocity, temperature, and stress / strain. It also incorporates internal temperature compensation, axis angle compensation, and navigation calculation information from the INS. This data is transmitted to the digital twin platform via industrial buses such as serial ports, CAN, or OPCUA for interaction between the virtual twin and the physical twin, thus establishing a connection between the virtual twin and the physical twin.

[0079] 5. Build the Decision Control Module: Utilize the graphical display functions of digital twin platforms such as ANSYS Twin Builder and Unity Manufacturing to visually represent the data collected from the multi-source information acquisition layer. The input to the digital twin model is determined to be real-time data collected from gyroscopes, accelerometers, grating encoders, temperature sensors, current sensors, voltage sensors, stress sensors, etc., including current, voltage, acceleration, angular velocity, temperature, stress, and strain information, as well as information on temperature compensation, shaft angle compensation, and navigation calculations. Through the digital twin physical model, combined with data-driven modeling methods such as machine learning and deep learning, a twin calculation model is built, outputting predicted data from each sensor of the rotating inertial navigation system and predicted navigation calculation performance. Based on the predicted results of each sensor and the predicted navigation calculation performance, and compared with subsequently collected real-time data, the parameters or structures of each digital twin physical model in the rotating inertial navigation system, such as the fiber optic gyroscope and accelerometer models, the servo control model, and the rotating inertial navigation calculation model, are optimized.

[0080] 6. Establish a diagnostic and predictive analysis module: First, model training is performed, which involves preprocessing the raw data acquired from the multi-source data acquisition layer and the characteristic information related to the operating status of the rotating inertial navigation system. This includes handling missing data and correcting abnormal data to improve data quality and avoid errors in subsequent data analysis. Feature extraction is performed to transform the raw data into representative feature vectors. Principal component analysis and linear discriminant analysis are used to reduce the dimensionality of the high-dimensional original feature vectors to improve computational efficiency. The dimensionality-reduced feature vectors are then fed into the digital twin deep learning model for training and evaluation. After new data acquisition input, the trained deep learning model outputs predictive analysis data. This data is compared with set thresholds or fault characteristics; if the threshold is exceeded or fault characteristics appear, an alarm is issued promptly. Simultaneously, the real-time performance prediction results from the decision control layer are compared with the gyroscope temperature compensation data, accelerometer temperature compensation data, and navigation calculation data calculated internally by the rotating inertial navigation system. Combined with the analysis results of the raw acquired data, the performance of the digital twin inertial navigation system is evaluated, and the accuracy of the performance prediction results is assessed.

[0081] 7. Integration of Rotating Inertial Navigation Digital Twin System: Through digital twin platforms such as ANSYS Twin Builder and Unity Manufacturing, the geometric model and physical model are synchronized to ensure data exchange and synchronization between different modules. Utilizing the communication between the data acquisition module and the twin platform, entity information is synchronized to the twin geometric and physical models, realizing the association between the twin virtual body and the entity. The decision control module learns from the data to achieve iterative optimization of each model. The diagnosis and predictive analysis module evaluates the overall operating status of the system and the performance of the digital twin inertial navigation system, and provides references.

[0082] Although embodiments and drawings of the present invention have been disclosed for illustrative purposes, those skilled in the art will understand that various substitutions, variations and modifications are possible without departing from the spirit and scope of the present invention and the appended claims. Therefore, the scope of the present invention is not limited to the contents disclosed in the embodiments and drawings.

Claims

1. A rotating inertial navigation digital twin system, characterized in that: It includes a rotating inertial navigation digital twin physical model layer, a rotating inertial navigation digital twin geometric model layer, a multi-source information data acquisition layer, a decision control layer, and a diagnosis and predictive analysis layer; The rotating inertial navigation digital twin geometric model layer: adopts a data-driven method to realize the dynamic representation of the inertial navigation operation process, ensuring synchronous coexistence with the inertial navigation entity; The rotating inertial navigation digital twin physical model layer includes a fiber optic gyroscope, accelerometer, grating, servo system, and rotating inertial navigation solution model. The multi-source information data acquisition layer includes various sensors for real-time acquisition of current, voltage, acceleration, angular velocity, temperature, stress and strain information. At the same time, it provides the solution data inside the rotating inertial navigation system for interaction with the digital twin model, forming an effective dataset that is fed back to the digital twin geometric and physical models. The decision control layer: based on the digital twin physical model, combined with real-time multi-source data, it visualizes and detects the operation process of the inertial navigation system; Based on measured data, the inertial navigation system is predicted in real time by establishing a dynamic mapping between physical entities and physical models; and the relevant parameters of the inertial navigation system are optimized based on the performance prediction results. The diagnostic and predictive analysis layer is used to extract features based on the raw acquisition information provided by the multi-source data acquisition layer, and to perform frequency domain analysis, time domain analysis, and statistical analysis. It also compares the results with set thresholds or fault characteristics for monitoring and alarm. Simultaneously, it compares the real-time performance prediction results from the decision control layer with the gyroscope temperature compensation data, accelerometer temperature compensation data, and navigation calculation data calculated internally by the rotating inertial navigation system to evaluate the actual performance of the inertial navigation system and provide future performance prediction results for reference.

2. The rotating inertial navigation digital twin system according to claim 1, characterized in that: The basic models of fiber optic gyroscopes and accelerometers are as follows: ; ; The angular velocity value detected by the gyroscope. It is the identity matrix. This is due to gyroscope installation error. For gyroscope scaling factor error, This is the theoretical value of angular velocity. The gyroscope has a constant zero bias. To detect the acceleration value output by the accelerometer, It is the identity matrix. Due to accelerometer installation error, For accelerometer scaling factor error, This is the theoretical value of acceleration. The accelerometer constant is zero bias.

3. The rotating inertial navigation digital twin system according to claim 2, characterized in that: The temperature compensation models for the scaling factor error, installation error, and zero bias error of the gyroscope and accelerometer all adopt a backpropagation (BP) neural network approach; the relationship between the input and output of the BP neural network is as follows: ; In the formula: This indicates scaling factor error, installation error, or zero bias error. The system input for scaling factor error, installation error, or zero bias error is temperature; Implicit layer bias for scaling factor error, installation error, or zero bias error; The output layer bias is set to account for scaling factor error, installation error, or zero bias error. The hidden layer weight matrix; This is the output layer weight matrix; For hidden layer output; The system output represents the scaling factor error, installation error, or zero bias error; tansig is the hyperbolic tangent sigmoid function; purelin is a linear function.

4. The rotating inertial navigation digital twin system according to claim 1, characterized in that: In a servo system, the servo motor typically drives the rotary mechanism directly. Its basic model is as follows: ; In the formula, Let be the moment of inertia of the servo motor shaft. servo motor rotational damping coefficient, For the rotational stiffness of the servo motor, The rotation angle of the motor. The first derivative of the motor's rotation angle, also known as the motor's angular velocity. It is the second derivative of the motor's rotation angle, also known as the motor's angular acceleration.

5. The rotating inertial navigation digital twin system according to claim 1, characterized in that: The rotating inertial navigation system (INS) solution model includes attitude differential equations, velocity differential equations, and position differential equations; among which, the attitude differential equations are: ; In the formula, , , , These are the four basic elements of a quaternion equation. , , , The differential of a quaternion, , , Let be the projection of the angular velocity of the carrier coordinate system b relative to the navigation coordinate system n in three directions of the b system; The velocity differential equation is: ; In the formula, Let the velocity of the vehicle be in the navigation coordinate system. for The differential, Let b be the attitude matrix relative to the navigation coordinate system n. Let be the projection of the angular velocity of Earth coordinate system e relative to inertial coordinate system i onto navigation coordinate system n. Let n be the projection of the angular velocity of the navigation coordinate system n relative to the Earth coordinate system e into the navigation coordinate system n. This is the projection of gravitational acceleration onto the carrier coordinate system b; The position differential equation is: ; The derivative of the longitude of the carrier's location. The derivative of the latitude of the carrier's location. The latitude of the location of the carrier. The height of the carrier. The derivative of the height h of the carrier. The carrier is located at the radius of curvature of the Earth's meridian. The carrier is located at the radius of curvature of the Earth along the circumference of the equinoxes. , , These represent the velocity of the carrier in the navigation coordinate system, and its components in the three directions: northeast, south, and north.

6. A method for constructing a rotating inertial navigation digital twin system based on any one of claims 1-5, characterized in that, include: Step 1: Use 3D modeling software to create a digital twin geometric model of the rotating inertial navigation system; Step 2: Establish and simulate a digital twin physical model of the rotating inertial navigation system using digital methods; Step 3: In the digital twin platform software, synchronize the geometric model and the physical model to visualize the motion process and input / output results; Step 4: Build a data acquisition module to complete data acquisition and transmit it to the digital twin platform for use by the twin model to achieve the association between the virtual twin and the physical entity; Step 5: Build a decision control module and use machine learning and deep learning data-driven modeling methods to achieve iterative optimization of model parameters; Step 6: Build a diagnostic and predictive analysis module to evaluate the overall operating status of the system and the performance of the digital twin inertial navigation system, and provide a reference.

7. The method for constructing a rotating inertial navigation digital twin system according to claim 6, characterized in that: In step 1, geometric models of the rotating inertial navigation system, components, and parts are established using 3D modeling software. Assembly and motion relationships are constructed according to the actual equipment composition to characterize the relationships between the components. Finite element models of key components of the motor and bearing rotation system are also established to characterize their stress and deformation.

8. The method for constructing a rotating inertial navigation digital twin system according to claim 6, characterized in that: Step 5 includes: using the graphical display function in ANSYS Twin Builder and the Unity Manufacturing digital twin platform to present the data information collected by the multi-source information acquisition layer in a graphical and visual manner; determining that the input of the digital twin model is the real-time acquired current, voltage, acceleration, angular velocity, temperature, stress and strain information, as well as temperature compensation, shaft angle compensation, and navigation calculation information; building a twin calculation model through the digital twin physical model and combining machine learning and deep learning data-driven modeling methods, and outputting the predicted data of each sensor of the rotating inertial navigation system and the predicted data of navigation calculation performance; and optimizing the parameters or structure of each digital twin physical model in the rotating inertial navigation system based on the predicted results of each sensor and the predicted results of navigation calculation performance, and comparing them with the subsequently acquired real-time data.

9. The method for constructing a rotating inertial navigation digital twin system according to claim 6, characterized in that: Step 6 involves building the diagnostic and predictive analysis module, which includes: first, model training, preprocessing the raw data acquisition information provided by the multi-source data acquisition layer and the characteristic information related to the operating status of the rotating inertial navigation system, including handling missing data and correcting abnormal data to improve data quality and avoid errors in subsequent data analysis; then, feature extraction is performed on the data, transforming the raw data into representative feature vectors, and using principal component analysis and linear discriminant analysis to reduce the dimensionality of the high-dimensional original feature vectors, thereby improving computational efficiency; finally, the dimensionality-reduced feature vectors are fed into the digital twin deep learning model to obtain the training and evaluation model.

10. The method for constructing a rotating inertial navigation digital twin system according to claim 6, characterized in that: In step 6, the collected data is input into the trained deep learning model, and the predicted analysis data is output. The data is then compared with the set thresholds or fault characteristics. If the thresholds are exceeded or fault characteristics are found, an alarm will be issued in a timely manner. At the same time, the real-time performance prediction results performed by the decision control layer are compared with the gyroscope temperature compensation data, accelerometer temperature compensation data, and navigation calculation data calculated internally by the rotating inertial navigation system. Combined with the analysis results of the original collected data, the performance of the digital twin inertial navigation system is evaluated, and the accuracy of the performance prediction results is judged.

Images