Space radiation synergistic effect ground simulation and electronic device performance prediction method

Abstract

The invention discloses a space radiation synergistic effect ground simulation and electronic device performance prediction method, and the method comprises the steps: achieving device-level high-fidelity digital twinning through the cooperation of SolidWorks / COMSOL, Unity 3D and Socket, achieving device-level mechanism deep twinning through the collaborative analysis mechanism of Sensor TCAD and SPICE, achieving data-driven intelligent prediction through the cooperation of MySQL and CNN-LSTM hybrid neural networks, and achieving device-level high-fidelity digital twinning through the cooperation of the Socket and the CNN-LSTM hybrid neural networks. The three parts are tightly coupled through a standardized data stream to form a complete, closed-loop and efficient space radiation effect evaluation and prediction solution.

Description

Technical Field

[0001] This invention belongs to the field of reliability engineering and radiation effect assessment technology of aerospace electronic components, specifically involving a method for ground simulation of space radiation synergy effects and prediction of electronic device performance. Background Technology

[0002] With the development of aerospace technology, the space radiation environment faced by spacecraft electronic systems is becoming increasingly severe. High-energy charged particles such as protons and heavy ions can cause sudden failures of semiconductor devices, such as single-event upsets (SEU) and sudden-escape failures (SEB), as well as total dose effect (TID), i.e., long-term degradation failures, seriously threatening the on-orbit safety and long-term operation of spacecraft. Therefore, it is crucial to conduct sufficient and accurate radiation effect simulation and evaluation on the ground. However, existing technologies for conducting sufficient and accurate radiation effect simulation and evaluation on the ground have at least the following limitations: First, there is a disconnect between irradiation device simulation and device effect evaluation. Ground simulation experiments mainly focus on the physical realization of the high-energy particle irradiation device itself and the beam parameter control, while there is a lack of direct digital and model-based correlation between the radiation field generated by the device and the complex physical damage mechanism inside the device. This leads to a large degree of blindness in experimental design, making it difficult to achieve accurate mapping and in-depth analysis from environmental simulation to failure mechanism. Second, failure mode analysis is isolated. Most studies conduct separate experiments and simulations for sudden failures, such as single-event upsets, and degradation failures, such as total dose effects. However, in real-world space environments, devices continue to be impacted by single-event events even after long-term performance degradation, resulting in a complex synergistic effect between the two. Current technologies lack effective means to quantitatively simulate and evaluate this crucial synergistic effect; thirdly, evaluation is inefficient, costly, and lacks predictive capability. The traditional "design-manufacturing-irradiation testing-analysis" process heavily relies on physical prototypes and field measurements, which are characterized by long cycles, high costs, and significant destructiveness. Limited experimental data is insufficient to accurately predict the performance evolution of devices throughout their entire lifecycle under complex and variable radiation environments, failing to meet the demands of rapidly iterating aerospace electronic design. Summary of the Invention

[0003] In view of the shortcomings of the prior art, the first objective of the present invention is to provide a method for ground simulation of space radiation synergy and prediction of electronic device performance.

[0004] To achieve the above-mentioned technical objectives, the present invention adopts the following technical solution:

[0005] This invention provides a method for ground-based simulation of space radiation synergistic effects and prediction of electronic device performance, comprising the following steps:

[0006] S1. Construct a digital twin of the ground simulation device and realize bidirectional real-time interaction between the digital twin and the physical device control system to obtain and output standardized digital radiation field parameters.

[0007] S2. Using the digital radiation field parameters as input, construct a physical model of the target device based on semiconductor process and device simulation tools, simulate the total dose degradation effect and single-event burst failure, analyze the synergistic effect of the two on the multi-bit flip sensitivity, and output quantitative evaluation results.

[0008] S3. Construct a relational database containing environmental parameters, device structural parameters, and simulation and experimental results. Based on this database, use a hybrid model of convolutional neural network and long short-term memory network to predict the performance evolution of the device under radiation.

[0009] In a preferred embodiment, step S1, constructing the digital twin specifically includes: using 3D design software to perform detailed 3D geometric modeling of the key subsystems of the ground simulation device; importing the geometric model into a virtual reality engine for lightweight processing and building an immersive virtual scene; and integrating physical field distribution data based on finite element simulation.

[0010] In the preferred embodiment, in step S1, the bidirectional real-time interaction is achieved through a hybrid scheme combining a custom Socket communication protocol with fiber optic transmission. Specifically, this includes: the sensor data of the physical device is converted into optical signals by an optical transceiver and transmitted to the data aggregation server via fiber optic cable; the communication client in the virtual reality engine connects to the server via the Socket protocol, receives real-time data to update the state of the digital twin and generate the digital radiation field parameters, and simultaneously sends control commands from the digital twin to the physical device for execution.

[0011] In the preferred embodiment, step S2, analyzing the influence mechanism of synergistic effect on multi-site flip sensitivity specifically includes: establishing a pre-damaged device model after total dose irradiation; comparing the single-particle transient response of the novel device model and the pre-damaged device model under the same heavy ion incident conditions, and analyzing the mechanisms of critical charge reduction, charge sharing enhancement, and equivalent shortening of sensitive node spacing.

[0012] In a preferred embodiment, step S2 further includes: importing the degradation amount of electrical parameters corresponding to the pre-damaged device model and the multi-bit flip transient current injection waveform obtained from the synergistic effect analysis into the circuit simulation environment, and performing transient fault injection simulation in the complete circuit netlist to evaluate the circuit-level error rate and functional failure risk.

[0013] In a preferred embodiment, in step S3, the relational database includes at least a logically related environmental parameter table, a device structure parameter table, a simulation result table, a physical experiment result table, and a synergistic effect analysis result table.

[0014] In a preferred embodiment, in step S3, the convolutional neural network is used to extract spatial correlation patterns from the input features, and the long short-term memory network is used to learn the temporal dependencies of performance parameters as they evolve over time or with cumulative injection volume.

[0015] In a preferred embodiment, the specific architecture of the hybrid model is as follows: the input layer integrates numerical parameters with preprocessed time-series data; the hidden layer adopts a structure of first a convolutional neural network layer and then a long short-term memory network layer; the output layer is configured as a regression-type or categorical output according to the prediction task.

[0016] In a preferred embodiment, the training of the hybrid model is based on historical data in the aforementioned relational database, employs the Adam optimizer, and introduces a Dropout layer and an early stopping strategy to prevent overfitting.

[0017] Beneficial effects

[0018] 1. It has achieved deep vertical integration of the evaluation process, seamlessly connecting the three levels of technology: device-level analog digital twin, device-level physical mechanism (TCAD / SPICE) and system-level intelligent prediction (AI), forming an integrated solution of "environmental simulation-mechanism revelation-performance prediction".

[0019] 2. This method innovatively quantifies the synergistic failure effect, particularly its impact on multi-particle flips: Through a serial simulation process of "degradation first, then sudden change," a "TID pre-damage device model" is constructed. Long-term degradation effects are introduced as the initial state into single-event effect analysis, achieving for the first time a quantitative simulation assessment of how the total dose effect modulates the sensitivity to single-event multi-particle flips at the physical mechanism level. This method can accurately analyze key mechanisms such as critical charge reduction and enhanced charge sharing, thus more realistically reflecting the circuit-level error rate under complex space radiation environments, and the assessment conclusions are more consistent with actual operating conditions.

[0020] 3. Significantly improves evaluation efficiency and prediction capabilities. By constructing a database that integrates mechanism simulation and artificial intelligence, it can extrapolate a large number of unknown operating conditions using limited physical test data, quickly predict device performance evolution and failure risks, and provide a high-performance digital tool for the selection, hardening design and on-orbit life prediction of spacecraft components, significantly reducing R&D costs and time. Attached Figure Description

[0021] Figure 1 Overall flowchart.

[0022] Figure 2 Flowchart for building a digital twin.

[0023] Figure 3 Flowchart of twin analysis of device radiation failure mechanism.

[0024] Figure 4 Radiation effect data platform and intelligent prediction process.

[0025] Figure 5 3D modeling display.

[0026] Figure 6 Twin page monitoring chart. Detailed Implementation

[0027] This invention provides a specific and implementable method for ground-based simulation of space radiation effects and prediction of device performance. The core of this method lies in constructing a vertically integrated technology chain from macroscopic devices to microscopic mechanisms, and then to intelligent prediction. Its specific implementation is as follows:

[0028] I. Construction and Real-time Interaction of a High-Fidelity Digital Twin for a Low-Energy Particle Ground Simulation Device

[0029] This phase aims to establish a digital twin that corresponds one-to-one with the physical device and can interact in real time, providing an accurate digital model for subsequent analysis.

[0030] 1.1 High-precision modeling of three-dimensional geometry and physical fields

[0031] Geometric Modeling: Using SolidWorks 3D design software, a 1:1 detailed 3D geometric model was created for the key subsystems of the ground simulation device (such as the electron cyclotron resonance (ECR) ion source, radio frequency antenna, plasma cavity, accelerating tube, and target chamber). The modeling process strictly followed the engineering drawings to ensure the accuracy of dimensions and assembly relationships.

[0032] Model Lightweighting and Scene Building: Import the SolidWorks (SW) model into the Unity3D virtual reality engine in an intermediate format (such as STEP). Lightweighting of the model is then performed within the engine, and an immersive virtual scene including devices, auxiliary equipment, and a virtual monitoring interface is built to provide operators with an intuitive interactive environment.

[0033] Integrated physical field simulation: Finite element physical field simulations were performed using COMSOL for the electromagnetic and plasma distributions inside the ECR source, and the electrostatic / electromagnetic fields in the accelerator tube. The field distribution data obtained from the simulations (such as magnetic field intensity contour maps and electric potential distributions) were then compiled into a radiation database.

[0034] 1.2 Implementation of Virtual-Real Interaction

[0035] A virtual-physical interaction middleware was developed to enable bidirectional communication between the digital twin and the physical device control system. The data interface primarily utilizes local communication protocols based on industry standards. To achieve stable and efficient real-time data interaction between the digital twin and the experimental data acquisition system, this phase employs a hybrid approach combining a custom Socket communication protocol with fiber optic transmission as the core means of virtual-physical interaction. This method is particularly suitable for experimental devices operating in environments with strong electromagnetic interference or requiring reliable long-distance transmission, ensuring high-fidelity, low-latency synchronization of critical data.

[0036] The implementation process is as follows: On the physical device side, key radiation field parameters and equipment status signals (such as beam intensity, vacuum level, scanning magnetic field current, etc.) are first acquired by the corresponding sensing units or controllers. These signals are not directly accessed through traditional industrial networks, but rather through dedicated optical transceivers to convert electrical signals into optical signals, which are then transmitted via single-mode or multimode optical fibers. Near the digital twin's operating server, optical transceivers are deployed to restore the optical signals to their original digital or analog electrical signals, which are ultimately output to a dedicated data aggregation server.

[0037] In virtual reality engines (such as Unity3D), a dedicated network communication client module is developed. This module establishes a stable point-to-point connection with the server program running on the aforementioned data aggregation server through a custom TCP / IP Socket protocol. The protocol message format is customized according to the characteristics of the transmitted data (such as real-time waveforms, scalar parameters, and status words) to ensure efficient and accurate data parsing.

[0038] During interaction, the parameter settings or control commands that the operator makes to the 3D model in the virtual scene are converted into predefined command messages by the communication module and sent to the data aggregation server via a Socket connection. After necessary interface conversion, the messages are finally delivered to the control system of the physical device for execution. Simultaneously, the device operation and radiation field monitoring data transmitted in real time via the fiber optic link are parsed by the server and continuously pushed to the digital twin through the same Socket channel. This uplink data stream not only drives the real-time and realistic updates of the model status, instrument readings, and radiation field distribution visualization in the virtual scene, but more importantly, it is integrated and calibrated to form a standardized "digital radiation field parameter report" (containing key information such as particle energy, spatial distribution, and flux rate), providing accurate and reliable input conditions for subsequent device-level physical mechanism simulations. The entire communication architecture has heartbeat detection, disconnection reconnection, and data verification mechanisms to ensure the reliability, real-time performance, and data integrity of the virtual-physical linkage process.

[0039] II. Device Radiation Failure Mechanism Twin Based on TCAD / SPICE and Co-analysis Model

[0040] This stage uses the standardized "digital radiation field parameters" output from the first stage as precise input to perform high-fidelity simulation at the device physical level. The core objective is to reveal the synergistic mechanism between degradation failure and sudden failure.

[0041] 2.1 Particle Incident Physics Mechanism and TCAD High-Fidelity Modeling

[0042] Based on semiconductor process and device simulation tools (such as Synopsys Sentaurus TCAD), a refined 3D physical model of the target memory cell (such as SRAM) is constructed. This model accurately reproduces the transistor's geometry, doping distribution, dielectric layer properties, and the layout of adjacent cells. The physical basis of the simulation describes the microscopic processes of interaction between high-energy particles and semiconductor materials: heavy ions or protons cause lattice displacement damage through non-ionization energy loss, and excite high-density electron-hole pairs on the incident track through ionization energy loss. The TCAD model quantitatively simulates the transient transport and final collection processes of these charge carriers by self-consistently solving the drift-diffusion equation, Poisson's equation, and defect dynamics equations.

[0043] 2.2 Simulation Modeling of Degradation Effects and Sudden Failure

[0044] Modeling of Total Dose Degradation Effect: Simulating Performance Degradation Caused by Long-Term Irradiation. The cumulative proton / gamma-ray dose parameters provided by the digital twin are input into the TCAD model to calculate the evolution of oxide trap charge and interface states with dose. Degradation data for key transistor parameters (such as threshold voltage Vth, subthreshold swing, and leakage current Ioff) are extracted through static electrical simulation.

[0045] Single-event burst failure modeling: Simulating transient disturbances caused by single-event injection. In transient simulations, heavy ions are simulated to inject a plasma track with a specific linear energy transfer value at a selected location. Through transient current analysis, transient current pulses collected at the drain, source, and substrate of the irradiated transistor and its neighboring transistors are obtained to evaluate the initial conditions for single-event flip and charge sharing.

[0046] 2.3 Mechanism Analysis of the Synergistic Effect on MBU Sensitivity

[0047] 2.3.1 Establishing a pre-damage model

[0048] The degradation of device electrical parameters (such as ΔVth) after a specific total dose of irradiation and the distribution of micro-defects are used as new initial conditions to update the TCAD model used for single-event effect simulation, thus constructing a "TID pre-damage device model".

[0049] 2.3.2 Comparative Simulation and Mechanism Analysis

[0050] Single-particle transient simulations were performed on both the "novel device model" and the "TID pre-damaged device model" using identical heavy ion incident conditions (including scan incident at different locations of adjacent cells). The following synergistic mechanisms were systematically compared and analyzed:

[0051] Critical charge reduction mechanism: The negative drift of the transistor threshold voltage (for NMOS) leads to an enhanced conduction capability, which reduces the critical charge (Qcrit) required for the memory node to maintain its logic state, making it more susceptible to single-event transients.

[0052] Charge sharing enhancement mechanism: The increase in leakage current induced by TID and the change in electric field in the depletion region may alter the efficiency and range of charge sharing between adjacent transistors. This makes it easier for excess carriers generated by a single particle incident to be collected by multiple neighboring nodes, significantly increasing the probability of multi-bit flips and mode complexity.

[0053] Equivalent shortening of sensitive node spacing: Due to the enhanced charge sharing effect mentioned above, the "effective spacing" of physically isolated memory cells in the layout is shortened in electrical terms, causing incident particles that might have only caused single-particle flips to trigger multi-particle flips under TID cooperation.

[0054] 2.3.2 Quantitative Assessment

[0055] Output quantitative metrics, including: the relationship between the multi-bit flip cross section and the total pre-dose, the change in the probability distribution of MBU modes (such as two-bit flip and four-bit flip), and the percentage decrease in critical charge, thereby directly assessing the degradation of circuit-level reliability under synergistic effects.

[0056] 2.4 Circuit-level impact assessment

[0057] The mechanistic analysis results are then evaluated at the circuit and system level. The SPICE model parameters after TID degradation and the representative MBU transient current injection waveforms obtained from synergistic effect analysis are imported into a circuit simulation environment (such as Cadence Spectre). Transient fault injection simulations are performed in a complete SRAM array or register file netlist, and statistical analysis is conducted on the actual error rates of unit bit flips and multi-bit flips under the influence of synergistic effects; the failure probability of error detection and correction circuits; and the risk level of system functional interruption. This completes a comprehensive and quantitative evaluation chain from physical damage mechanisms to circuit functional failure.

[0058] III. Radiation Effect Data Platform and Prediction Based on Relational Databases and Hybrid Neural Networks

[0059] 3.1 Building a Relational Database Based on SQL

[0060] A "Space Radiation Effect Database" was designed and established using a MySQL database management system. The database mainly contains the following logically related table structures: Environmental Parameters Table: stores radiation field conditions from the digital twin (particle type, energy value, microwave power value, flux ratio, etc.); Device Structure Parameters Table: stores key design parameters of simulated / experimental devices (process node, channel length / width, oxide layer thickness, doping concentration, etc.); Simulation Results Table: stores the raw results and extracted features (such as electrical characteristic curves, SET pulse parameters) of TCAD / SPICE simulations, and is linked to the Environmental Parameters Table and Device Structure Table via foreign keys; Physical Experiment Results Table: stores test data from historical irradiation experiments (electrical parameter test records, failure phenomenon records); Synergistic Effect Analysis Results Table: specifically stores comparative data on the sudden failure response of "non-degraded devices" and "pre-degraded devices," quantifying synergistic factors.

[0061] 3.2 Performance Prediction Model Based on Hybrid Neural Networks

[0062] Network structure selection: Considering that radiation effect data has both static attributes (device structure, radiation type) and dynamic sequences (performance degradation over time / flux), a hybrid model of convolutional neural network (CNN) and long short-term memory network (LSTM) was designed and used.

[0063] The CNN part is used to automatically extract deep spatial correlation patterns from input features, such as processing device doping profiles (which can be converted into two-dimensional data) or fluence distributions of particles with different energies. The LSTM part is used to learn and memorize long-term dependencies of performance parameters over time or cumulative fluence, making it ideal for predicting degradation trajectories.

[0064] The specific implementation architecture of the model is as follows: The input layer is designed to fuse heterogeneous input features, including numerical device structural parameters and radiation environment parameters, as well as preprocessed time-series fluence sequence data. Based on this, the hidden layers adopt a hybrid structure of convolution followed by recurrent layers: first, one to two layers of convolutional neural networks perform spatial abstraction and deep pattern extraction on the input features; then, the output feature sequence is input into a long short-term memory network layer to learn the dynamic temporal dependence of performance parameters on time or fluence evolution; finally, a fully connected layer is connected for high-order nonlinear fitting to integrate the features and map them to the output space. The output layer is flexibly configured according to the specific prediction task; it can be set as a regression output to predict specific values ​​of threshold voltages under a specific fluence, or as a categorical output to assess the probability level of single-event flip under given cooperative radiation conditions. The training and optimization of the model are based on the radiation effect database constructed above. The historical simulation and experimental data in the database are used to divide the training set and the validation set. The Adam optimizer is used to update the parameters, and the mean squared error loss function or the cross-entropy loss function is selected according to the task type. In order to improve the generalization ability, a Dropout layer is introduced into the network to prevent overfitting, and an early stopping strategy is adopted to automatically terminate the training when the performance on the validation set no longer improves, thereby determining the optimal number of training rounds.

[0065] In summary, this invention achieves device-level high-fidelity digital twins through "SolidWorks / COMSOL + Unity3D + Socket", achieves device-level deep mechanism twins through "Sentaurus TCAD + SPICE collaborative analysis mechanism", and achieves data-driven intelligent prediction through "MySQL + CNN-LSTM hybrid neural network". These three technologies are tightly coupled through standardized data flow, forming a complete, closed-loop, and efficient solution for assessing and predicting spatial radiation effects.

[0066] Example 1

[0067] Taking low-energy particle irradiation of SRAM cells as an example, the embodiments of the present invention will be described in detail with reference to the accompanying drawings:

[0068] Step 1: Accurate Reproduction of the Radiation Environment Based on Digital Twins

[0069] First, for a low-energy electron irradiation device, a 3D geometric model of its ECR ion source, beam pipeline, and vacuum target chamber was constructed using SolidWorks, and its internal electromagnetic field and plasma distribution were obtained through COMSOL simulation. After importing the model into the Unity3D engine, an interactive scene including a virtual control console and device status panel was built. When conducting SRAM cell irradiation experiments, the operator sets the electron energy (e.g., 110 keV) and beam intensity in the virtual scene, and the commands are sent to the physical device for execution via a custom Socket protocol through an optical fiber link. Simultaneously, real-time data collected by the beam detector and dosimeter installed in the target area are fed back to the digital twin through the same optical fiber path, driving the virtual instruments to update synchronously, and ultimately generating a standardized digital radiation field parameter file that accurately describes the current "proton energy-flux rate-spatial distribution".

[0070] Step 2: Physical simulation and quantitative analysis of the cooperative radiation failure mechanism

[0071] Using the radiation field parameters generated in the first step as input, device-level mechanism simulation is initiated. A detailed 3D model of a 180 nm SRAM cell, consistent with the experimental sample fabrication process, is constructed in Synopsys Sentaurus TCAD. The simulation is divided into two sequential stages:

[0072] Total dose degradation simulation: Simulate the long-term effects of SRAM cells under cumulative electron flux (e.g., 5e10 p / cm²), calculate oxide trap charge generation and interface state density increase, and extract the degradation of critical transistor threshold voltage (Vth).

[0073] Synergistic Analysis of Single-Particle Effects: Degraded electrical parameters and defect distribution were imported into a "TID pre-damage" model. Subsequently, a single-particle transient induced by single-proton incidence was simulated on this model. The simulation results of the "novel" and "pre-damage" models were compared to quantitatively analyze the synergistic effects: The results show that due to the negative drift of Vth, the node critical charge is reduced by about 25%, and the charge sharing range is expanded due to the increase in leakage current, resulting in an increase of about one order of magnitude in the sensitive cross section of the two-bit flip (MBU).

[0074] Step 3: Data-driven performance prediction

[0075] The simulation results (radiation conditions, device structure, ΔVth, MBU cross-section, etc.) were entered into a MySQL database along with historical experimental data. Using this data, a CNN-LSTM hybrid neural network model was trained. The CNN layers learned the device structure features, while the LSTM layers learned the performance evolution over time / flux. After training, when a rapid evaluation of the performance of a newly designed 90 nm SRAM cell under similar irradiation conditions was needed, only its layout parameters and radiation conditions needed to be input. The trained model could then predict its threshold voltage degradation curve and MBU risk probability within seconds, and the prediction results were consistent with the trends verified by subsequent physical experiments.

Claims

1. A method for predicting the performance of space radiation synergistic effects ground simulation and electronic devices, characterized by, The method comprises the following steps: S1, constructing a digital twin of a ground simulation device, and realizing two-way real-time interaction between the digital twin and a physical device control system to obtain and output standardized digital radiation field parameters; S2, taking the digital radiation field parameters as input, constructing a physical model of a target device based on a semiconductor process and device simulation tool, simulating total dose degradation effects and single particle burst failure, and analyzing the influence mechanism of the synergistic effect on multi-bit flip sensitivity, and outputting quantitative evaluation results; S3, constructing a relational database containing environmental parameters, device structure parameters and simulation and test results, and based on the database, using a hybrid model of a convolutional neural network and a long short-term memory network to predict the performance evolution of the device under radiation.

2. A method of predicting the performance of space radiation synergistic effects ground simulation and electronic devices according to claim 1, characterized in that, In step S1, the digital twin is constructed, specifically including: using a three-dimensional design software to perform fine three-dimensional geometric modeling on key subsystems of the ground simulation device; importing the geometric model into a virtual reality engine for lightweight processing and building an immersive virtual scene; and integrating physical field distribution data based on finite element simulation.

3. A method of predicting the effects of space radiation synergistically on the ground and electronic device performance according to claim 1 or 2, characterized in that, In step S1, the two-way real-time interaction is realized through a hybrid scheme of a self-defined Socket communication protocol combined with optical fiber transmission, specifically including: converting sensor data of the physical device into optical signals through an optical transceiver and transmitting the optical signals to a data collection server through an optical fiber; connecting a communication client in the virtual reality engine to the server through a Socket protocol, receiving real-time data to update the state of the digital twin and generating the digital radiation field parameters, and sending control instructions from the digital twin to the physical device for execution.

4. The method of claim 1, wherein the method further comprises: In step S2, the influence mechanism of the synergistic effect on multi-bit flip sensitivity specifically includes: establishing a pre-damage device model after total dose irradiation; comparing the single particle transient response of a brand new device model and the pre-damage device model under the same heavy ion incidence conditions, and analyzing the mechanisms of critical charge reduction, charge sharing enhancement and effective shortening of sensitive node spacing.

5. A method of predicting the performance of space radiation synergistic effects ground simulation and electronic devices according to claim 4, characterized in that, Step S2 further includes: importing the degradation amount of the electrical parameters of the pre-damage device model and the multi-bit flip transient current injection waveform obtained by synergistic effect analysis into a circuit simulation environment for transient fault injection simulation in a complete circuit netlist to evaluate the error rate and functional failure risk at the circuit level.

6. The method of predicting the performance of space radiation synergistic effects ground simulation and electronic devices according to claim 1, characterized in that, In step S3, the relational database at least includes logically associated environmental parameter tables, device structure parameter tables, simulation result tables, physical test result tables and synergistic effect analysis result tables.

7. A method of predicting the effects of space radiation synergistically on the ground and electronic device performance according to claim 1 or 6, characterized in that, In step S3, the convolutional neural network is used to extract spatial correlation patterns in input features, and the long short-term memory network is used to learn the time-dependent relationship of performance parameters over time or cumulative fluence evolution.

8. A method of predicting the performance of space radiation synergistic effects ground simulation and electronic devices according to claim 7, characterized in that, The specific architecture of the hybrid model is: the input layer fuses numerical parameters and preprocessed time series data; the hidden layer adopts a structure of a convolutional neural network layer followed by a long short-term memory network layer; and the output layer is configured as a regression type or a categorical output according to the prediction task.

9. A method of predicting the performance of space radiation synergistic effects ground simulation and electronic devices according to claim 8, characterized in that, The training of the hybrid model is based on historical data in the aforementioned relational database, uses an Adam optimizer, and introduces a Dropout layer and an early stopping strategy to prevent overfitting.

Images