FNO-based functional link resource autonomous matching test signal generation apparatus and method

Abstract

Disclosed in the present invention are an FNO-based functional link resource autonomous matching test signal generation apparatus and method, belonging to the technical field of electronic testing. The present invention organically combines a broadband radio frequency signal generation mechanism with a deductive reasoning neural network algorithm, and dynamically identifies a signal parameter set and utilizes same, so as to achieve autonomous matching and situation deduction for signal generation. The solution not only adaptively extracts signal parameter features and data features and utilizes same, but also uses environmental parameters and a signal pattern specified by a user as other input features for processing; then, the solution extracts high-level features layer by layer by using Fourier transform and inverse transform operations, and explores the logical relationship, spatial correlation, and temporal characteristics among input data, such that a model has a deductive reasoning capability; and finally, the solution autonomously generates matching simulation data according to test scenarios, so as to complete radio frequency generation of signals, thereby eliminating high dependence on manual operations, and achieving high-precision simulation of complex electromagnetic signals and autonomous matching generation for multi-application scenarios.

Description

FNO-based Functional Link Resource Autonomous Matching Test Signal Generation Device and Method Technical Field

[0001] This invention belongs to the field of electronic testing technology, specifically relating to a device and method for generating test signals for autonomous matching of functional link resources based on FNO. Background Technology

[0002] The primary method for generating complex electromagnetic signals is to simulate electromagnetic signal patterns and generate data using digital domain simulation technology, and then generate the electromagnetic signals through digital-to-analog conversion and I / Q modulation. Traditional approaches involve setting a series of general signal parameters based on the ideal mathematical expression of electromagnetic signals to meet signal requirements under specific conditions or scenarios, and then generating simulation data within a fixed framework. The generated signals are approximations of real signals under ideal conditions. However, real-world electromagnetic applications are complex and diverse, with multiple types of signals overlapping and signal parameters changing in real time. The aforementioned methods are limited by ideal mathematical formulas and parameter sets, resulting in electromagnetic signals that differ from actual scene signals, failing to meet the need for high-fidelity simulation of complex electromagnetic environments. Furthermore, the generation of relevant data and the operation of the signal generator are highly dependent on the experience and proficiency of the test personnel, limiting the signal generator's ability to quickly perceive environmental changes and simulate signals in real time across different application scenarios. In recent years, deep learning-based signal modeling methods have constructed a framework different from traditional mathematical models by using signal data and neural networks. They can adaptively extract the parameter features and data features of signals to achieve tasks such as signal classification, recognition and noise reduction. They can also simulate some electromagnetic signals with low complexity, but their ability to identify and predict unknown complex signals and infer future trends of signals is relatively weak.

[0003] Complex electromagnetic signal simulation technology is widely used to generate various electromagnetic signals and is one of the key technologies in fields such as radar, navigation, and communication. Combining software simulation systems based on digital complex electromagnetic environment data databases with complex signal generation hardware platforms (such as signal generators) can generate ideal complex-pattern signals for equipment performance testing. The main method for generating complex-pattern electromagnetic signals involves selecting an appropriate signal pattern generation model to produce signal data, then performing digital-to-analog conversion and I / Q modulation, and finally using a signal generation instrument to generate the electromagnetic signal. Electromagnetic signal generation models based on ideal mathematical expressions and deep neural network frameworks based on big data can generate simulated signal data that conforms to specific conditions or scenarios, approximating actual signals under ideal conditions.

[0004] Currently, the simulation of electromagnetic signals is typically based on ideal mathematical expressions, generating simulation data through various pre-set parameters. For example, the following is a sinusoidal expression...

[0005]

[0006] Used to generate radio frequency signals. Wherein represents the amplitude of the signal component. The frequency of the signal components and This represents the phase of the signal component. This technique is an approximation of the real signal under ideal assumptions. Taking the simulation of electromagnetic radiation signals as an example, users generate waveform data through various parameters to simulate and generate signals, as shown in Figure 1. Users can set relevant parameters on the control panel according to their needs. The software parses the set data set, compiles it, and generates electromagnetic radiation signals of the specified type and parameters, storing them in the system storage space as files for users to retrieve and call. In addition, the generated waveform data can also be directly transmitted to the signal generator via LAN communication to achieve radio frequency output. It can be seen that this scheme is based on the ideal mathematical formula of a specific type of electromagnetic signal to simulate and generate simulation data, which is an approximation of the real signal under ideal assumptions. This also leads to a large error between the simulated electromagnetic radiation signal and the real space, which is inconvenient to evaluate and cannot meet the relevant requirements of electromagnetic signal simulation. The types of parameters set through the panel are limited and fixed. If other types of excitation signals are to be generated, the entire test process needs to be repeated, relying entirely on the tester's subjective calculation and modeling. Such methods are objectively limited by the fixed mechanism and parameter set of ideal mathematical formulas, and are also greatly affected by the knowledge level and testing experience of the testers, which poses a hidden danger to the scientificity and efficiency of related testing work. Technical issues

[0007] The main drawbacks of existing technologies include:

[0008] 1. Traditional electromagnetic signal simulation methods based on ideal mathematical formulas are objectively limited by fixed formula mechanisms and parameter sets, and cannot simulate actual electromagnetic environment signals with high fidelity.

[0009] 2. Traditional electromagnetic signal simulation methods rely heavily on the experience and knowledge of the testers and their skill in operating the instruments. They cannot make independent corrections to cope with changes in the signal environment, which limits the efficiency of signal simulation.

[0010] 3. Changes in real-world application scenarios require signals with different characteristics. Modeling methods based on ideal mathematical expressions and conventional neural networks cannot meet the simulation requirements of environmental parameter evolution and are difficult to autonomously match to different application scenarios. Technical solutions

[0011] To address the aforementioned technical problems in the existing technology, this invention proposes a functional link resource autonomous matching test signal generation device and method based on FNO (Fourier Neural Operator). The design is reasonable, overcomes the shortcomings of the existing technology, and has good results.

[0012] To achieve the above objectives, the present invention adopts the following technical solution:

[0013] The FNO-based functional link resource autonomous matching test signal generator includes a feature preprocessing module, a feature generalization learning module, and a signal feature fusion and training module.

[0014] The feature preprocessing module is configured to process user-defined signal features or factors that affect the generated signal, as well as the basic features of the acquired signal.

[0015] The feature generalization learning module is configured to process the basic features in the preprocessing stage according to different categories and characteristics to form standardized feature data, and then use Fourier layers to perform step-by-step generalization to autonomously construct the implicit space of signal features.

[0016] The signal feature fusion and training module is configured to perform affine transformations and scaling on the data in the latent feature space to adapt to changing target signals.

[0017] Furthermore, this invention also mentions a functional link resource autonomous matching test signal generation method based on FNO. This method uses the functional link resource autonomous matching test signal generation device based on FNO as described above, and specifically includes the following steps:

[0018] Step 1: The feature preprocessing module processes user-defined signal features or factors affecting the generated signal, as well as the basic features of the acquired signal.

[0019] Step 2: Through the feature generalization learning module, the basic features in the preprocessing stage are processed according to different categories and characteristics to form standardized feature data. Then, Fourier layers are used to perform step-by-step generalization to autonomously construct the implicit space of signal features.

[0020] Step 3: Through the signal feature fusion and training module, affine transformation and scaling are performed on the data in the latent feature space, and the data is compared with the features of the target signal to form a loss function, thereby achieving autonomous learning and feature matching.

[0021] Preferably, step 1 specifically includes the following steps:

[0022] Step 1.1: Perform data cleaning on the custom style file and influencing factors;

[0023] Step 1.2: Perform word segmentation and sorting on the remaining characters;

[0024] Step 1.3: Digitize the segmented words and put them into the bag of words to form part of the signal features; for the collected signals, directly extract their basic features and then import them into the feature library.

[0025] Preferably, step 2 specifically includes the following steps:

[0026] Step 2.1: Process the basic features using a fully connected artificial neural network;

[0027] Step 2.2: Use stacked Fourier layers to extract features step by step and generalize;

[0028] Step 2.3: Use a fully connected network to transform the generalized features to the target space, thereby achieving implicit construction of the input features.

[0029] Preferably, in step 2, each Fourier layer includes a Fourier transform, a convolution operation, an inverse Fourier transform, and a linear operation. Beneficial effects

[0030] The beneficial technical effects of this invention are as follows:

[0031] This invention organically combines the broadband radio frequency signal generation mechanism with the deductive operator FNO network algorithm, which can adaptively process different input features, explore and construct the logical and spatial temporal relationships between feature data, establish a feature-based signal generation mechanism, and is not limited by the fixed mechanism and parameter set of signal generation based on ideal mathematical formulas, thus realizing high-fidelity simulation of complex electromagnetic signals.

[0032] This invention enables autonomous matching of functional link resources across multiple application scenarios, eliminating the heavy reliance on manual operation and expert experience throughout the entire testing task chain. It can autonomously learn and correct based on the signal feature library of actual environmental signals, thereby improving the simulation efficiency and accuracy of environmental signals. Attached Figure Description

[0033] Figure 1 is a schematic diagram of the electromagnetic radiation signal simulation generation principle;

[0034] Figure 2 is a schematic block diagram of the device of the present invention;

[0035] Figure 3 is a schematic diagram of a Fourier-induced deep neural network. Embodiments of the present invention

[0036] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments:

[0037] This invention proposes a test excitation signal generation technology based on FNO deductive reasoning network for autonomous matching of functional link resources. It organically combines a broadband radio frequency signal generation mechanism with a deductive reasoning neural network algorithm to dynamically identify and utilize signal parameter sets, achieving autonomous matching and situational deduction of signal generation. This scheme breaks through the traditional mathematical formula-based signal generation architecture. It can not only adaptively extract and utilize signal parameter features and data features, but also process environmental parameters and user-specified signal patterns as other input features. Then, it uses Fourier transform and inverse transform operations to extract high-level features layer by layer, exploring the logical relationships, spatial connections, and temporal sequences between input data, enabling the model to have deductive reasoning capabilities. Finally, it autonomously generates matching simulated data according to the test scenario, thereby completing the radio frequency signal generation. This eliminates the high dependence on manual operation, achieving high-precision simulation of complex electromagnetic signals and autonomous matching generation for multiple application scenarios.

[0038] This invention proposes a test excitation signal generation architecture based on autonomous matching of functional link resources using an FNO deductive network, combining the FNO network used for solving parameterized partial differential equations with a classical signal generation mechanism. The model mainly consists of three parts: a feature preprocessing module, a feature generalization learning module with deductive reasoning capabilities, and a signal feature fusion and training module, as shown in Figure 2. The feature preprocessing module handles user-defined signal features or factors influencing the generated signal, as well as the basic features of the acquired signal. This module first uses traditional techniques to clean the custom style file and influencing factors, then segments and arranges the remaining characters, and finally digitizes the segmented words and places them into a bag-of-words system, forming part of the signal features. Furthermore, for the acquired signal, we can directly extract its basic features and then import them into a feature library. The feature generalization learning module based on Fourier operations processes the basic features from the preprocessing stage according to their categories and characteristics, forming standardized feature data. Then, it uses Fourier layers for progressive generalization, autonomously constructing an implicit space for signal features. The signal feature fusion and training module performs affine stretching or compression based on the implicit space features, enabling the signal generated by the signal generation component to exhibit different amplitude, phase, and frequency variation trends. The data generated by the model is processed by the signal generation component to obtain the predicted signal, which is compared with the features of the target signal to form a loss function, thereby achieving autonomous learning and feature matching for this device.

[0039] The generalization learning module is an operator deductive network with a stacked structure. This module first processes the basic features using a fully connected artificial neural network, then uses stacked Fourier layers to extract and generalize features step-by-step. Finally, a fully connected network transforms the generalized features to the target space, thus implicitly constructing the input features and providing the signal generation module with the necessary refined input. Generally, the parameters used to generate signals differ greatly in characteristics and scale. Magnitude normalization does not conform to the physical laws of signal generation and affects model performance. Furthermore, environmental parameters differ significantly from signal parameters. Therefore, we classify these parameters according to their properties (e.g., amplitude, frequency, phase, pulse, and others), and then use multiple simple networks to process these feature groups respectively, increasing the dimensionality of the input data and enriching its expressive meaning, thus expanding the model's search space. Further, we apply the Fourier activation function to the first hidden layer of these simple network models to better capture features with similarity within feature groups and distinguish features with high similarity, as shown in Figure 3.

[0040] Basic features with different properties are processed by multiple sub-neural networks to form new features, which are then fed into Fourier layers. Each Fourier layer includes a Fourier transform, convolution operation, inverse Fourier transform, and a linear operation. The results from the two paths are fused together, then processed by an activation function, and finally output. Successive Fourier layers refine and extract features step by step, enhancing the model's expressive and generalization abilities. Mathematically, an operator deductive network with an iterative structure is represented as follows:

[0041]

[0042]

[0043]

[0044] in, and Two shallow neural networks are used to increase and decrease feature dimensionality. Representing operator operations. The Fourier layer implements the kernel integral operator, which has the ability to aggregate information around a point of interest. The relevant concepts and forms are described below.

[0045] Definition of a kernel operator: A kernel integral operator over a region D

[0046]

[0047] in It is a continuous kernel function, such as the Green kernel function which has global aggregation properties. Feature processing of continuous signals is a global operation, so a global kernel function is a good choice. Furthermore, we can implement a global convolution kernel using Fourier transform.

[0048] Fourier kernel integral operator: Let and They represent continuous functions respectively. The Fourier transform and inverse Fourier transform, then for continuous input variables... and nuclear We have the following expression

[0049]

[0050] The signal fusion generation and training module performs affine transformations and scaling on the data in the latent feature space to adapt to changing target signals, especially multi-tone composite signals. Finally, this data is played back through the signal generation component.

[0051] For a given training sample points After a series of processing steps in the above model, the corresponding predicted signal can be obtained. Therefore, we can obtain the cost function of the model:

[0052]

[0053] in and These represent the network input and the actual signal labels, respectively. Represents the signal generated by the network. This represents the set of all weights and bias parameters of a neural network. Regularization terms are used to prevent overfitting in neural networks. The sparsity of the sparsity factor weights can adjust the sparsity and overfitting during training.

[0054] By continuously optimizing the cost function through the stochastic gradient descent algorithm, the model's network embedding layer can better encode the features of the input data and make these inputs exhibit their own unique properties. The Fourier layer processes and aggregates the logical relationships, spatial connections, and temporal sequences between the input data step by step, transforming the features of the user input into the latent feature space, and providing suitable feature inputs for the signal generation component.

[0055] Through iterative training, a network model is obtained that can reasonably process input features and automatically generate excitation signals for various application scenarios. Especially when the feature input is incomplete, the network can automatically construct the input for the signal generation module, heuristically generating the signal pattern desired by the user. Furthermore, some features of complex electromagnetic signals generated by the adaptive network can also serve as input to the network, providing training data for further optimization of the network model. This enables the network to possess self-learning capabilities, gradually improving the model's robustness and generalization performance.

[0056] In summary, the test excitation signal generation method based on the autonomous matching of functional link resources of the FNO operator deductive network proposed in this invention realizes the iterative optimization process from feature to signal as shown in Figure 2. It solves the problems of low simulation fidelity and single application scenario caused by the high dependence of electromagnetic space construction on traditional ideal mathematical formulas, and achieves the electronic testing goals of "one specialty with multiple functions" and "scenario adaptability".

[0057] This invention presents a functional link resource autonomous matching test excitation signal simulation generation framework that organically combines a broadband radio frequency signal generation mechanism with a deductive reasoning network (FNO).

[0058] This invention addresses the significant differences in signal and environmental parameters by designing a multi-DNN channel feature embedding submodule for the FNO network module. This submodule performs feature scale separation and information smoothing, resolving the issues of feature scale imbalance and information oscillation caused by a large number of different parameters, and improving the FNO's ability to handle complex and ever-changing signal environments and signal parameters.

[0059] This invention proposes an adaptive scaling strategy for the autonomous matching mechanism of functional link resources in signal generators, generating signals with different amplitude and frequency scales, thereby improving the multi-scenario adaptive matching capability of the proposed network model.

[0060] Of course, the above description is not intended to limit the present invention, and the present invention is not limited to the examples given above. Any changes, modifications, additions or substitutions made by those skilled in the art within the scope of the present invention should also fall within the protection scope of the present invention.

Claims

1. A functional link resource autonomous matching test signal generating device based on FNO, characterized in that: The feature preprocessing module, the feature generalization learning module and the signal feature fusion and training module are included. The feature preprocessing module is configured to process user-defined signal features or factors affecting the generation of signals and basic features of collected signals. The feature generalization learning module is configured to process basic features in the preprocessing stage according to different categories and characteristics, form standardized feature data, and then perform step-by-step generalization using Fourier layers to autonomously construct an implicit space of signal features. The signal feature fusion and training module is configured to perform affine transformation and scaling of data in the implicit feature space to adapt to changing target signals.

2. The method for generating test signals based on the autonomous matching of functional link resources of FNO, characterized in that: The FNO-based functional link resource autonomous matching test signal generation device according to claim 1 specifically includes the following steps: Step 1: The feature preprocessing module processes user-defined signal features or factors affecting the generation of signals and basic features of collected signals. Step 2: The feature generalization learning module processes basic features in the preprocessing stage according to different categories and characteristics, forms standardized feature data, and then performs step-by-step generalization using Fourier layers to autonomously construct an implicit space of signal features. Step 3: The signal feature fusion and training module performs affine transformation and scaling of data in the implicit feature space, compares it with the features of the target signal, forms a loss function, and thus realizes autonomous learning and feature perfect matching.

3. The method of claim 2, wherein the FNO-based functional link resource self- matching test signal generation method is characterized by: Step 1 specifically includes the following steps: Step 1.1: Data cleaning is performed on the self-defined style file and influencing factors. Step 1.2: The remaining characters are processed and arranged. Step 1.3: The divided words are digitized and placed in a bag of words to form part of the signal features.

4. The method of claim 2, wherein the FNO-based functional link resource self- matching test signal generation method is characterized by: Step 2 specifically includes the following steps: Step 2.1: The basic features are processed using a fully connected artificial neural network. Step 2.2: The stacked Fourier layers are used to extract features and generalize step by step. Step 2.3: A fully connected network is used to transform the generalized features to the target space, thereby realizing implicit construction of the input features.

5. The method of claim 2, wherein the FNO-based functional link resource self- matching test signal generation method is characterized by: Each Fourier layer in Step 2 includes Fourier transform, convolution operation, inverse Fourier transform, and a linear operation.

Images