Refractive index prediction method and related device for cascaded long-period fiber gratings

Abstract

The present application discloses a refractive index prediction method and related device for cascaded long-period fiber gratings. The method includes: obtaining the resonant wave center wavelength and transmission loss of the central interference peak of the optical signal of the current cascaded long-period fiber grating; using an improved fully connected refractive index prediction model to perform feature analysis on the resonant wave center wavelength and transmission loss, and predict the current refractive index; the improved fully connected refractive index prediction model is constructed after training the optimal weights based on the Dropout algorithm and the relu activation function. The parameters based on the model are in line with the actual situation and can ensure the reliability of the predicted refractive index; in terms of the prediction method adopted, the model is specifically constructed and trained to ensure the accuracy of the prediction results. Therefore, the present application can solve the technical problem that the existing cascaded long-period fiber grating technology is difficult to obtain accurate and reliable refractive index, resulting in limited application of cascaded long-period fiber gratings.

Description

Technical Field

[0001] The present application relates to the field of neural network technology, and in particular to a refractive index prediction method and related device for cascaded long-period fiber gratings. Background Art

[0002] Long-period gratings (LPGs), as lossy fiber filters distinct from Bragg gratings, have found widespread application in fiber-optic communications and fiber-optic sensing. Periodic fiber gratings can be categorized into short-period and long-period types based on the length of their period. A cascaded LPG can be considered a combination of two LPG segments separated by a cascade fiber of length d and an initial phase shift of Φ.

[0003] In the refractive index sensing measurement of cascaded long-period fiber gratings, it is difficult to calculate an accurate and reliable refractive index based on the obtained spectral wavelength according to the refractive index sensitivity expression formula; this will lead to limited applications of cascaded long-period fiber gratings in fields such as optical fiber sensors, which is not conducive to its development. Summary of the Invention

[0004] The present application provides a refractive index prediction method and related devices for cascaded long-period fiber gratings, which are used to solve the technical problem that the existing cascaded long-period fiber grating technology is difficult to obtain accurate and reliable refractive index, resulting in limited application of cascaded long-period fiber gratings.

[0005] In view of this, the first aspect of the present application provides a refractive index prediction method for cascaded long-period fiber gratings, comprising:

[0006] Obtaining the resonant wave center wavelength and transmission loss of the central interference peak of the optical signal of the current cascade long-period fiber grating;

[0007] Using an improved fully connected refractive index prediction model to perform characteristic analysis on the central wavelength of the resonant wave and the transmission loss, and predicting the current refractive index;

[0008] The improved fully connected refractive index prediction model is constructed after training the optimal weights based on the Dropout algorithm and the ReLU activation function.

[0009] Preferably, the improved fully connected refractive index prediction model is used to perform characteristic analysis on the resonant wave center wavelength and the transmission loss, and predict the current refractive index, and the method further includes:

[0010] Acquire multiple sets of historical resonant wave center wavelengths and historical transmission losses to construct a refraction training set, wherein the refraction training set includes an actual refractive index;

[0011] Based on the Dropout algorithm, the relu activation function and the prediction error, the initial fully connected refractive index prediction model is predicted and trained according to the refractive training set to obtain an improved fully connected refractive index prediction model.

[0012] Preferably, the step of obtaining multiple sets of historical resonant wave center wavelengths and historical transmission losses to construct a refraction training set further includes:

[0013] Normalization processing is performed on the historical resonant wave center wavelength and the historical transmission loss.

[0014] Preferably, the method further comprises: performing prediction training on the initial fully connected refractive index prediction model according to the refractive index training set based on the Dropout algorithm, the relu activation function and the prediction error to obtain an improved fully connected refractive index prediction model;

[0015] The prediction error is calculated based on the prediction result obtained from the prediction training and the actual refractive index.

[0016] A second aspect of the present application provides a refractive index prediction device for cascaded long-period fiber gratings, comprising:

[0017] A data acquisition unit, configured to acquire a resonant wave center wavelength and a transmission loss of a central interference peak of an optical signal of a current cascaded long-period fiber grating;

[0018] a refraction prediction unit, configured to perform characteristic analysis on the central wavelength of the resonant wave and the transmission loss using an improved fully connected refractive index prediction model, and predict a current refractive index;

[0019] The improved fully connected refractive index prediction model is constructed after training the optimal weights based on the Dropout algorithm and the ReLU activation function.

[0020] Preferably, it also includes:

[0021] A data set construction unit is used to obtain multiple sets of historical resonant wave center wavelengths and historical transmission losses to construct a refraction training set, wherein the refraction training set includes an actual refractive index;

[0022] The model training unit is used to perform prediction training on the initial fully connected refractive index prediction model according to the refractive index training set based on the Dropout algorithm, the Relu activation function and the prediction error to obtain an improved fully connected refractive index prediction model.

[0023] Preferably, it also includes:

[0024] A normalization processing unit is used to perform normalization processing on the historical resonant wave center wavelength and the historical transmission loss.

[0025] Preferably, it also includes:

[0026] The error calculation unit is used to calculate the prediction error based on the prediction result obtained by the prediction training and the actual refractive index.

[0027] A third aspect of the present application provides a refractive index prediction device for cascaded long-period fiber gratings, the device comprising a processor and a memory;

[0028] The memory is used to store program code and transmit the program code to the processor;

[0029] The processor is configured to execute the refractive index prediction method for cascaded long-period fiber gratings according to the instructions in the program code.

[0030] In a fourth aspect, the present application provides a computer-readable storage medium for storing program code, wherein the program code is used to execute the refractive index prediction method for cascaded long-period fiber gratings described in the first aspect.

[0031] It can be seen from the above technical solutions that the embodiments of the present application have the following advantages:

[0032] In the present application, a refractive index prediction method for cascaded long-period fiber gratings is provided, comprising: obtaining the central wavelength of the resonant wave and the transmission loss of the central interference peak of the optical signal of the current cascaded long-period fiber grating; using an improved fully connected refractive index prediction model to perform feature analysis on the central wavelength of the resonant wave and the transmission loss, and predict the current refractive index; the improved fully connected refractive index prediction model is constructed after training the optimal weights based on the Dropout algorithm and the ReLU activation function.

[0033] The refractive index prediction method for cascaded long-period fiber gratings provided in this application uses an improved fully connected refractive index prediction model to predict the refractive index of the cascaded long-period fiber grating. The input to the model is the key parameters of the cascaded long-period fiber grating, namely the center wavelength of the resonant wave and the transmission loss. The parameters are based on actual conditions and can ensure the reliability of the predicted refractive index. In terms of the prediction method, the model is specifically constructed and trained to ensure the accuracy of the prediction results. Therefore, this application can solve the technical problem that the existing cascaded long-period fiber grating technology has difficulty in obtaining accurate and reliable refractive index, which has led to limited application of cascaded long-period fiber gratings. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 A schematic flow chart of a method for predicting the refractive index of cascaded long-period fiber gratings provided in an embodiment of the present application;

[0035] Figure 2A schematic structural diagram of a refractive index prediction device for cascaded long-period fiber gratings provided in an embodiment of the present application;

[0036] Figure 3 A schematic diagram of the network structure of the improved fully connected refractive index prediction model provided in an embodiment of the present application;

[0037] Figure 4 Schematic diagram of the actual refractive index and predicted refractive index curve provided in the embodiment of the present application;

[0038] Figure 5 Schematic diagram of the relative loss curve of the improved fully connected refractive index prediction model provided in an embodiment of the present application. DETAILED DESCRIPTION

[0039] In order to help those skilled in the art better understand the present invention, the following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of this application.

[0040] For easier understanding, see Figure 1 , an embodiment of the refractive index prediction method for cascaded long-period fiber gratings provided in this application includes:

[0041] Step 101: Obtain the resonance wave center wavelength and transmission loss of the center interference peak of the optical signal of the current cascaded long period fiber grating.

[0042] A cascaded long-period grating can be viewed as a combination of two long-period gratings separated by a cascaded optical fiber of length d and an initial phase shift of Φ. It can be applied in fiber grating sensors, and the refractive index prediction method provided in this embodiment can provide good theoretical support for its application.

[0043] It should be noted that the central interference peak is selected because the resolution here is the highest. The central wavelength of the resonant wave and the transmission loss can be obtained by calculation or direct acquisition. These two variables are used as model inputs and can be expressed as the first input x1 and the second input x2, respectively.

[0044] Step 102: Use the improved fully connected refractive index prediction model to perform characteristic analysis on the central wavelength of the resonant wave and the transmission loss, and predict the current refractive index.

[0045] The improved fully connected refractive index prediction model is constructed after training the optimal weights based on the Dropout algorithm and the relu activation function.

[0046] Improve the network structure of the fully connected refractive index prediction model Figure 3 As shown, the input is the resonant wave center wavelength and transmission loss. The data first enters the neural network from the input layer, then undergoes nonlinear calculation and mapping in the hidden layer. Finally, the data is exported through the output layer to obtain the refractive index prediction result, namely the current refractive index. In this embodiment, the current cascaded long-period fiber grating and the current refractive index both represent unknown prediction tasks. The improved fully connected refractive index prediction model is trained based on a specific method and can be directly applied to unknown prediction tasks to achieve refractive index prediction.

[0047] The dropout algorithm adjusts the training structure of the fully connected network, thereby enhancing the model's generalization ability and preventing overfitting. The improved fully connected refractive index prediction model utilizes the ReLU (Linear Rectification Function) activation function for fast approximation. The model was trained using Keras on Windows. The model parameters set the batch size of the training set to 64, the number of iterations to 3000, and 60% of the dataset was randomly selected as the training set, while 40% of the data was selected as the test set.

[0048] Furthermore, the previous also includes:

[0049] Obtain multiple sets of historical resonant wave center wavelengths and historical transmission losses to construct a refraction training set, which includes the actual refractive index;

[0050] Based on the Dropout algorithm, ReLU activation function and prediction error, the initial fully connected refractive index prediction model is trained according to the refractive training set to obtain an improved fully connected refractive index prediction model.

[0051] Furthermore, multiple sets of historical resonant wave center wavelengths and historical transmission losses are obtained to construct a refraction training set, which also includes:

[0052] The historical resonant wave center wavelength and historical transmission loss are normalized.

[0053] Furthermore, the initial fully connected refractive index prediction model is trained based on the refractive index training set based on the Dropout algorithm, the ReLU activation function and the prediction error to obtain an improved fully connected refractive index prediction model, which also includes:

[0054] The prediction error is calculated based on the prediction results obtained from the prediction training and the actual refractive index.

[0055] The pre-training process for the improved fully connected refractive index prediction model is as follows: data first enters the neural network from the input layer, then nonlinear calculations and mapping are performed in the hidden layer. Finally, data is exported through the output layer. The difference between the output and the true result is compared, and backpropagation is performed to update the parameters. The minimum error between the output and the true result is found, and the neural network parameters are then saved. Detailed data related to the specific model training is shown in Table 1.

[0056] Table 1 List of model input and output parameters

[0057]

[0058]

[0059] The process of calculating the prediction error based on the prediction results obtained from the prediction training and the actual refractive index is the difference between the actual refractive index and the predicted refractive index. For the changes between the predicted refractive index and the actual refractive index, please refer to Figure 4 The vertical axis represents the true refractive index of the original data and the predicted refractive index obtained by the prediction model. The difference between the true and predicted refractive indices can be seen. The horizontal axis represents the number of samples. Please refer to Table 2 for the training sample parameters.

[0060] Table 2 List of parameters related to true refractive index and predicted refractive index

[0061]

[0062]

[0063] According to Table 2, the relative error is obtained by comparing the actual refractive index n with the refractive index y output by the model prediction. In the 9 sets of data, the absolute values ​​of the relative errors are small, all less than 0.6%. Among them, the absolute values ​​of the relative errors of 8 sets of data are controlled within 0.5%. When the actual refractive index y is 1.358, the prediction result of the improved fully connected refractive index prediction model has the highest accuracy, and the absolute value of the relative error is only 0.14%.

[0064] When training a model using the Dropout algorithm, several neurons in the hidden layer are randomly removed. With each iteration, all input and output connections of the removed neurons are temporarily discarded. These removed neurons will not be included in the network and will not be included in this round of model training. Each neuron in the same layer has an equal chance of being removed, so the choice of which neurons to remove is random. As a result, the neural network will change to varying degrees with each training session, preventing it from overfitting to a specific local feature.

[0065] Because the units of the resonant center wavelength at the central interference peak and the transmission loss data input variables differ, they may have different effects during the calculation process. To prevent the model prediction results from being affected by the data dimension differences, the data needs to be normalized. Normalization can both reduce the range of sample data and shorten model training time. The improved fully connected refractive index prediction model uses normalized data for training, which not only accelerates model convergence but also further improves the prediction accuracy of the entire model.

[0066] Furthermore, during training, the number of network layers and neurons is continuously adjusted by assessing the accuracy of model training and the magnitude of loss. The fully connected neural network refractive index prediction model utilizes the relu (Linear Rectification Function) activation function for rapid approximation. The model was trained using Keras on Windows. The model parameters were set to 64 batch size, 3000 iterations, and 60% of the dataset was randomly selected as the training set, while 40% was used as the test set.

[0067] See also Figure 5 , where lower relative loss values ​​indicate a more accurate description of the experimental data by the improved fully-connected refractive index prediction model. The relative loss curve for the improved fully-connected refractive index prediction model shows that the training results of the improved fully-connected refractive index prediction model have a high accuracy rate. A higher accuracy rate indicates that the improved fully-connected refractive index prediction model better fits the experimental data. As the number of samples increases, the relative loss approaches 0.0185. As the number of training times increases, the root mean square error gradually decreases and converges, with the mean absolute error remaining stable at 0.0445, indicating that the prediction model training is effective. After training, the improved fully-connected refractive index prediction model is obtained.

[0068] The refractive index prediction method for cascaded long-period fiber gratings provided in the embodiments of the present application uses an improved fully connected refractive index prediction model to predict the refractive index of the cascaded long-period fiber grating. The inputs to the model are the key parameters of the cascaded long-period fiber grating, namely the center wavelength of the resonant wave and the transmission loss. The parameters are based on actual conditions and can ensure the reliability of the predicted refractive index. In terms of the prediction method, the model is specifically constructed and trained to ensure the accuracy of the prediction results. Therefore, the embodiments of the present application can solve the technical problem that the existing cascaded long-period fiber grating technology has difficulty in obtaining an accurate and reliable refractive index, which has led to limited application of cascaded long-period fiber gratings.

[0069] For easier understanding, see Figure 2The present application provides an embodiment of a refractive index prediction device for cascaded long-period fiber gratings, comprising:

[0070] The data acquisition unit 201 is used to acquire the central wavelength of the resonant wave and the transmission loss of the central interference peak of the optical signal of the current cascaded long-period fiber grating;

[0071] The refraction prediction unit 202 is configured to perform characteristic analysis on the central wavelength of the resonant wave and the transmission loss using an improved fully connected refractive index prediction model, and predict the current refractive index;

[0072] The improved fully connected refractive index prediction model is constructed after training the optimal weights based on the Dropout algorithm and the relu activation function.

[0073] Furthermore, it also includes:

[0074] A data set construction unit 203 is used to obtain multiple sets of historical resonant wave center wavelengths and historical transmission losses to construct a refraction training set, where the refraction training set includes an actual refractive index;

[0075] The model training unit 204 is used to perform prediction training on the initial fully connected refractive index prediction model according to the refractive index training set based on the Dropout algorithm, the ReLU activation function and the prediction error to obtain an improved fully connected refractive index prediction model.

[0076] Furthermore, it also includes:

[0077] The normalization processing unit 205 is used to perform normalization processing on the historical resonant wave center wavelength and the historical transmission loss.

[0078] Furthermore, it also includes:

[0079] The error calculation unit 206 is used to calculate the prediction error according to the prediction result obtained by the prediction training and the actual refractive index.

[0080] The present application also provides a refractive index prediction device for cascaded long-period fiber gratings, the device comprising a processor and a memory;

[0081] The memory is used to store program codes and transmit the program codes to the processor;

[0082] The processor is configured to execute the refractive index prediction method for cascaded long-period fiber gratings in the above method embodiment according to instructions in the program code.

[0083] The present application also provides a computer-readable storage medium for storing program code, and the program code is used to execute the refractive index prediction method for cascaded long-period fiber gratings in the above method embodiment.

[0084] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0085] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0086] In addition, the functional units in the various embodiments of the present application may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0087] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for executing all or part of the steps of the method described in each embodiment of the present application through a computer device (which can be a personal computer, server, or network device, etc.). The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (English full name: Read-Only Memory, English abbreviation: ROM), random access memory (English full name: Random Access Memory, English abbreviation: RAM), disk or optical disk and other media that can store program code.

[0088] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A refractive index prediction method for cascaded long-period fiber gratings, characterized in that: include: Obtaining the resonant wave center wavelength and transmission loss of the central interference peak of the optical signal of the current cascaded long-period fiber grating, wherein the current cascaded long-period fiber grating is regarded as a long-period grating combination of two long-period gratings separated by a cascaded optical fiber of length d and an initial phase shift of size Φ, and is applied in a fiber grating sensor; Using an improved fully connected refractive index prediction model to perform characteristic analysis on the central wavelength of the resonant wave and the transmission loss, and predicting the current refractive index; The improved fully connected refractive index prediction model is constructed after training the optimal weights based on the Dropout algorithm and the ReLU activation function.

2. The refractive index prediction method for cascaded long-period fiber gratings according to claim 1, characterized in that: The improved fully connected refractive index prediction model is used to perform characteristic analysis on the central wavelength of the resonant wave and the transmission loss, and to predict the current refractive index, and the method also includes: Acquire multiple sets of historical resonant wave center wavelengths and historical transmission losses to construct a refraction training set, wherein the refraction training set includes an actual refractive index; Based on the Dropout algorithm, the relu activation function and the prediction error, the initial fully connected refractive index prediction model is predicted and trained according to the refractive training set to obtain an improved fully connected refractive index prediction model.

3. The refractive index prediction method for cascaded long-period fiber gratings according to claim 2, characterized in that: The method further includes obtaining multiple sets of historical resonant wave center wavelengths and historical transmission losses to construct a refraction training set, and then further including: Normalization processing is performed on the historical resonant wave center wavelength and the historical transmission loss.

4. The refractive index prediction method for cascaded long-period fiber gratings according to claim 2, characterized in that: The method further includes: performing prediction training on the initial fully connected refractive index prediction model according to the refractive index training set based on the Dropout algorithm, the ReLU activation function and the prediction error to obtain an improved fully connected refractive index prediction model; The prediction error is calculated based on the prediction result obtained from the prediction training and the actual refractive index.

5. A refractive index prediction device for cascaded long-period fiber gratings, characterized in that: include: a data acquisition unit, configured to acquire the central wavelength of the resonant wave and the transmission loss of the central interference peak of the optical signal of the current cascaded long-period fiber grating (LPFBG), wherein the current cascaded LPFG is regarded as a LPFG combination in which two LPFGs are separated by a cascaded optical fiber of length d and an initial phase shift of size Φ, and is applied in a fiber grating sensor; a refraction prediction unit, configured to perform characteristic analysis on the central wavelength of the resonant wave and the transmission loss using an improved fully connected refractive index prediction model, and predict a current refractive index; The improved fully connected refractive index prediction model is constructed after training the optimal weights based on the Dropout algorithm and the ReLU activation function.

6. The refractive index prediction device for cascaded long-period fiber gratings according to claim 5, characterized in that: Also includes: A data set construction unit is used to obtain multiple sets of historical resonant wave center wavelengths and historical transmission losses to construct a refraction training set, wherein the refraction training set includes an actual refractive index; The model training unit is used to perform prediction training on the initial fully connected refractive index prediction model according to the refractive index training set based on the Dropout algorithm, the Relu activation function and the prediction error to obtain an improved fully connected refractive index prediction model.

7. The refractive index prediction device for cascaded long-period fiber gratings according to claim 6, characterized in that: Also includes: A normalization processing unit is used to perform normalization processing on the historical resonant wave center wavelength and the historical transmission loss.

8. The refractive index prediction device for cascaded long-period fiber gratings according to claim 6, characterized in that: Also includes: The error calculation unit is used to calculate the prediction error based on the prediction result obtained by the prediction training and the actual refractive index.

9. A refractive index prediction device for cascaded long-period fiber gratings, characterized in that: The device includes a processor and a memory; The memory is used to store program code and transmit the program code to the processor; The processor is configured to execute the refractive index prediction method for cascaded long-period fiber gratings according to any one of claims 1 to 4 according to the instructions in the program code.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium is used to store program codes, and the program codes are used to execute the refractive index prediction method for cascaded long-period fiber gratings according to any one of claims 1 to 4.

Images