Design method of hollow-core anti-resonance optical fiber temperature sensor based on artificial intelligence

Abstract

The invention discloses a hollow-core anti-resonance optical fiber temperature sensor design method based on artificial intelligence, relates to the technical field of optical fiber sensor design, and aims to avoid tedious multi-time simulation calculation by training a forward performance prediction model and a reverse structure design model, and save calculation resources and time while ensuring design precision. In combination with forward and reverse models, the optical fiber structure can be flexibly adjusted according to different performance requirements, and the requirements of performance changes of the sensor can be quickly met. Due to the flexibility, the optical fiber sensor can be widely and accurately applied under different environment conditions. Through the combination of the forward performance prediction model and the reverse structure design model, the closed loop of structure design and performance verification is realized, and the accuracy and reliability of the design scheme in practical application are ensured.

Description

Technical Field

[0001] This invention relates to the field of fiber optic sensor design technology, specifically to a design method for a hollow anti-resonant fiber optic temperature sensor based on artificial intelligence. Background Technology

[0002] Sensors have a wide range of applications, covering almost all modern technologies, industrial fields, and daily life. Fiber optic sensors, as a type of sensor based on the principle of optical signal transmission and modulation, use various types of optical fibers as their core transmission medium. Changes in the external environment, such as temperature, magnetic fields, and stress, will cause changes in the optical signal transmitted in the fiber. By measuring changes in parameters such as signal loss and phase in the fiber transmission, parameters such as temperature, magnetic field, and stress in the external environment can be indirectly obtained. Therefore, they possess characteristics such as high sensitivity, resistance to electromagnetic interference, corrosion and high temperature resistance, and the ability to measure multiple parameters.

[0003] Hollow-core antiresonant optical fibers confine optical signals within an air core, further enhancing their transmission performance. Sensor design based on hollow-core antiresonant fibers holds promise for improving sensor performance. Currently, the design of hollow-core antiresonant fiber sensors all require starting with numerical simulation. Finite element method (FEM) simulation is the most commonly used method, but it demands significant computational resources and time for complex structures. In the optimization design of hollow-core antiresonant fiber sensors, multiple structural parameters need to be optimized to achieve optimal sensor performance. When there are many structural parameters, each with a large value range, the optimization parameter space for fiber optic sensors becomes enormous. Performing exhaustive optimization of all parameters would require orders of magnitude more computation. Due to the computational resource demands of FEM simulation, comprehensive parameter optimization is difficult to achieve. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a design method for a hollow anti-resonant fiber optic temperature sensor based on artificial intelligence, which solves the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a design method for a hollow-core anti-resonant fiber optic temperature sensor based on artificial intelligence, comprising the following steps:

[0006] S1. Determine the basic structure, parameters to be optimized, and optimization range of the hollow anti-resonant fiber optic temperature sensor.

[0007] S2. Dataset D is constructed using finite element numerical simulation.

[0008] S3. Train a positive performance prediction model using dataset D, and combine the positive performance prediction model to construct an inverse structure design model;

[0009] S4. Utilize positive performance prediction models to achieve large-scale parameter optimization;

[0010] S5. Implement the structural design of a fiber optic temperature sensor with specific performance using a reverse structural design model.

[0011] Preferably, S1 includes S11 and S12;

[0012] S11. Determine the structure of the hollow-core anti-resonant fiber optic temperature sensor, including the diameter of the fiber core, the structure of the cladding, whether it is filled with liquid, and the type and location of the liquid;

[0013] The core of a hollow anti-resonant optical fiber is usually made of air, while the cladding tube is filled with different materials, such as ethanol, depending on the design requirements, and the ethanol refractive index is obtained.

[0014] The fiber core has a diameter of 30 μm, the cladding tube is made of quartz tubing, and the filling liquid is ethanol.

[0015] The refractive index of ethanol is obtained using the following formula:

[0016] n = n0 - a(T - T0);

[0017] In the formula, n represents the refractive index of ethanol at temperature T, n0 represents the refractive index of ethanol at reference temperature T0, and a represents the thermo-optical coefficient of ethanol, with a value of 3.94 × 10⁻⁶. -4 / ℃, T represents the experimental temperature, and T0 represents the reference temperature;

[0018] S12. Determine the structural parameters that need to be optimized in the sensor design, and set the optimization range for each parameter, including the cladding gap g and the cladding wall thickness t. tube and gold layer thickness t Au These parameters directly affect the transmission characteristics of optical fibers, such as loss limitation and sensor sensitivity. To ensure that the sensor can operate efficiently within a temperature range of 20℃ to 40℃, the optimization range of these parameters will be precisely set.

[0019] The cladding gap g ranges from 2 μm to 6 μm, and this parameter directly affects the anti-resonance effect and loss characteristics of the optical fiber; the cladding wall thickness t... tube The range is 0.5 μm to 0.8 μm, which is a key parameter affecting the stability and sensitivity of the optical fiber structure; the gold layer thickness t Au The range is from 20nm to 50nm. The thickness of the gold layer will affect the surface plasmon resonance effect of the sensor, and thus affect the temperature sensitivity of the sensor.

[0020] Preferably, S2 includes S21 and S22;

[0021] S21. Using the finite element method, the hollow-core anti-resonant optical fiber with different structural configurations is simulated. The geometric structure of the optical fiber is defined, including the cladding gap g and the cladding wall thickness t. tube and gold layer thickness t Au The limiting loss of the optical fiber in the wavelength range of 1.44μm to 1.66μm was calculated using finite element simulation at different temperature ranges from 20℃ to 40℃ to evaluate the sensing performance of the optical fiber.

[0022] In the simulation process, the cladding tube gap g of the optical fiber is 2μm to 6μm, and the cladding tube wall thickness t is taken into account. tube The gold layer thickness ranges from 0.5 μm to 0.8 μm. Au The effects of 20nm to 50nm and temperature on optical fiber performance;

[0023] Preferably, in step S22, fiber confinement loss data are collected at temperatures between 20°C and 40°C and wavelengths between 1.44μm and 1.66μm; each simulation result, including structural parameters, temperature and wavelength, is compiled into a data table, with each row representing the fiber structure at a temperature and its corresponding loss data.

[0024] Finally, the different temperatures, wavelengths, and their corresponding limiting loss values ​​obtained from the simulation are organized to form a complete dataset D;

[0025] Dataset D is obtained using the following formula:

[0026] ;

[0027] In the formula, p1, p2, ..., pn represent the optical fiber structure parameters and corresponding wavelengths. The optical fiber structure includes the cladding gap g and the cladding wall thickness t. tube and gold layer thickness t Au .

[0028] Preferably, S3 includes S31 and S32;

[0029] S31. Normalize the acquired dataset D to ensure that the data is within a reasonable range in order to improve the training effect of the model. Construct and train a positive performance prediction model through dataset D. The positive performance prediction model is a fully connected neural network model.

[0030] Input the cladding gap g and cladding wall thickness t into the forward performance prediction model tube Gold layer thickness t Au The output layer provides 12 numbers corresponding to the limiting loss at different wavelengths, along with the experimental temperature; the optimized activation function ReLU is selected, the learning rate is 0.0005, the number of training epochs is 750, and the test set ratio is set to 10% for verification;

[0031] The calculation formula for the positive performance prediction model is as follows: y = ReLU(WD + b);

[0032] In the formula, y represents the output value, W represents the weight matrix, and b represents the bias term;

[0033] Correlation coefficient R 2 This is used to evaluate the positive performance prediction model based on the output value y, and the judgment method is as follows;

[0034] When 0.8 < R 2 A value ≤1 indicates that the prediction result is accurate;

[0035] When 0 < R 2 A value ≤0.8 indicates that the prediction result is inaccurate.

[0036] Preferably, S32, based on the output of the positive performance prediction model, train and construct the inverse structure design model;

[0037] The inputs to the inverse structural design model are the predicted limit loss curve and the experimental temperature, and the outputs are the structural parameters of the hollow antiresonant fiber, including the cladding gap g and the cladding wall thickness t. tube and gold layer thickness t Au By using the reverse structural design model, the optical fiber structure parameters that meet the requirements can be calculated in reverse based on the target sensitivity.

[0038] The reverse structure design model is a fully connected neural network model that uses the ReLU activation function. By adjusting the model parameters, the fiber structure can be accurately deduced from the predicted loss curve.

[0039] The calculation formula for the inverse model is as follows:

[0040] ;

[0041] In the formula, f -1 This represents the inverse function, and zp represents the output structural parameters, including the optimal cladding gap g and the optimal cladding wall thickness t. tube and optimal gold layer thickness t Au .

[0042] Preferably, in step S4, the determined optimization parameter range is optimized over a wide range using a positive performance prediction model, including the cladding gap g and the cladding wall thickness t. tube and gold layer thickness t Au ;

[0043] By using a forward performance prediction model, different parameter combinations are predicted, and the corresponding limiting losses are predicted. A grid search optimization algorithm is used to find the structural configuration that maximizes the sensor sensitivity. This process uses a forward model to quickly evaluate and adjust the design, reducing the tedious simulation and experimental processes in traditional optimization methods.

[0044] Based on the structural configuration of the sensor sensitivity, select the optimal fiber optic design parameters;

[0045] The sensitivity calculation formula is as follows: S = Δλ / ΔT;

[0046] In the formula, S represents the sensitivity of the sensor, specifically the degree of response of the limiting loss to temperature changes, Δλ represents the amount of change in limiting loss, which is obtained by the difference in limiting loss between adjacent temperatures, and ΔT represents the amount of temperature change.

[0047] The formula relating loss to structural parameters is as follows:

[0048] L(A,T) = fx(g,t) tube , t Au (T)

[0049] In the formula, fx represents the mapping function.

[0050] Preferably, S5 includes S51 and S52;

[0051] S51. Using the sensitivity S of the sensor, the reverse structural design model is used to calculate the structural parameters that meet the sensitivity requirement.

[0052] Specifically, the sensor's sensitivity S is determined by the ratio between the limiting loss CL and the temperature change ΔT;

[0053] The sensor's sensitivity S is input into the reverse structure design model, and the limiting loss curve and temperature T in the wavelength range of the optical fiber from 1.44μm to 1.66μm are also input into the reverse structure design model.

[0054] The reverse structural design model calculates the fiber structure parameters nzp that meet the sensitivity requirements based on the reverse of the input, including optimizing the cladding gap ng and the cladding tube wall thickness nt. tube and optimize gold layer thickness nt Au ;

[0055] The fiber optic structure parameter nzp is obtained using the following formula:

[0056] .

[0057] Preferably, in step S52, the obtained fiber structure parameter nzp is verified using a positive performance prediction model to ensure that its sensitivity meets expectations.

[0058] The forward performance prediction model calculates the new confinement loss nL(A,T) based on the fiber structure parameter nzp, and further calculates the new sensitivity nS;

[0059] The obtained sensitivity nS is compared with the preset sensitivity threshold TS to determine the success of the design.

[0060] When the sensitivity nS ≥ the sensitivity threshold TS, the design is considered successful.

[0061] If the sensitivity nS < the sensitivity threshold TS, the design has failed and needs to be readjusted.

[0062] An artificial intelligence-based hollow-core anti-resonant fiber optic temperature sensor includes an optical fiber jacket tube, a cladding tube disposed inside the optical fiber jacket tube, a gold plating layer disposed on the inner wall of the cladding tube, a gap between adjacent cladding tubes of the optical fiber and the optical fiber jacket tube, an internal region of the optical fiber disposed inside the cladding tube, and the distance between the cladding tubes being the diameter of the optical fiber core.

[0063] This invention provides a design method for a hollow anti-resonant fiber optic temperature sensor based on artificial intelligence, which has the following advantages:

[0064] (1) By training the forward performance prediction model and the reverse structural design model, tedious multiple simulation calculations are avoided, which greatly saves computing resources and time while ensuring design accuracy. Combining the forward and reverse models, the fiber optic structure can be flexibly adjusted according to different performance requirements to quickly respond to the needs of sensor performance changes. This flexibility makes the application of fiber optic sensors more extensive and accurate under different environmental conditions. By combining the forward performance prediction model and the reverse structural design model, a closed loop of structural design and performance verification is realized, ensuring the accuracy and reliability of the design scheme in practical applications. The optimization and verification steps of the model can effectively improve the temperature response sensitivity of the sensor.

[0065] (2) By introducing the refractive index formula of ethanol and considering the effect of temperature change on its refractive index, and using the thermo-optic coefficient for correction, this design method can accurately simulate the effect of ethanol on the optical fiber performance at different temperatures, thus improving the accuracy of temperature sensing. By clearly defining the optimization range, such as cladding gap, tube wall thickness, and gold layer thickness, and setting reasonable optimization intervals for each parameter, parameter optimization can be performed more efficiently. This refined optimization method avoids the fuzzy and unsystematic parameter adjustments in traditional design methods, ensuring the high performance of the sensor.

[0066] Finite element method (FEM) simulation was used to calculate optical fibers with different structural configurations, particularly focusing on loss limiting within the wavelength range of 1.44 μm to 1.66 μm. The FEM simulation simultaneously considered the effects of various factors on fiber performance, including temperature variations, cladding tube gap, tube wall thickness, and gold layer thickness, making the sensor design more adaptable across multiple dimensions, especially for efficient sensing of temperature changes.

[0067] (3) By using the correlation coefficient R 2 This metric measures the accuracy of the model's predictions, ensuring a close match between the model's output and actual data. It provides a standardized quantitative basis for evaluating model quality, ensuring the reliability of the predictions. Using the loss-limiting curve from the forward performance prediction model's output, the inverse structural design model is further trained, allowing for the reverse calculation of optimized fiber optic structure parameters based on the predicted loss data. This process avoids the trial-and-error methods of traditional design, making fiber optic design more precise and efficient.

[0068] (4) By using a positive performance prediction model and combining it with a grid search optimization algorithm, the cladding gap g and the cladding wall thickness t are evaluated. tube and gold layer thickness t Au The structural parameters are optimized over a wide range. This method efficiently finds the optimal configuration within the combination space of structural parameters, thereby maximizing the sensor's sensitivity and improving the accuracy and speed of the optimization process. The sensitivity calculation formula directly quantifies the fiber optic sensor's response to temperature changes. Accurate sensitivity calculation allows the design process to precisely align with actual performance requirements, ensuring efficient sensor operation within the desired temperature range.

[0069] Combining a forward performance prediction model with a grid search optimization algorithm allows for the traversal of all possible combinations of structural parameters, avoiding the inefficiencies of manual adjustment and random selection. This automated optimization method ensures that the fiber optic structure can quickly reach optimal performance. The reverse structural design model is used to calculate the fiber optic structure parameters that meet the sensitivity requirements, and this is then verified using the forward performance prediction model. This design method achieves a closed-loop design from target performance to structural parameters, avoiding repeated trial and error and improving the reliability and efficiency of the design. Attached Figure Description

[0070] Figure 1 This is a schematic diagram illustrating the steps of a design method for a hollow anti-resonant fiber optic temperature sensor based on artificial intelligence according to the present invention.

[0071] Figure 2 This is a schematic diagram of the forward prediction model structure of the present invention;

[0072] Figure 3 This is a schematic diagram of the reverse structural design model of the present invention;

[0073] Figure 4 This is a schematic diagram of the structure of a hollow anti-resonant fiber optic temperature sensor based on artificial intelligence according to the present invention.

[0074] Legend:

[0075] 1. Fiber optic outer tube; 2. Cladding tube; 3. Internal region of fiber optic cladding tube; 4. Gold plating inside fiber optic cladding tube; 5. Gap between adjacent cladding tubes of the fiber; 6. Fiber optic core diameter; Detailed Implementation

[0076] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0077] Example 1

[0078] This invention provides a design method for a hollow-core anti-resonant fiber optic temperature sensor based on artificial intelligence. Please refer to [link / reference]. Figures 1 to 4 This includes the following steps:

[0079] S1. Determine the basic structure, parameters to be optimized, and optimization range of the hollow anti-resonant fiber optic temperature sensor.

[0080] S2. Dataset D is constructed using finite element numerical simulation.

[0081] S3. Train a positive performance prediction model using dataset D, and combine the positive performance prediction model to construct an inverse structure design model;

[0082] S4. Utilize positive performance prediction models to achieve large-scale parameter optimization;

[0083] S5. Implement the structural design of a fiber optic temperature sensor with specific performance using a reverse structural design model.

[0084] In this embodiment, by introducing artificial intelligence technology and combining forward performance prediction models and reverse structural design models, intelligent design and optimization of the fiber optic sensor structure are achieved, reducing manual intervention and improving the automation and efficiency of the design process. A combination of finite element numerical simulation and machine learning is used to optimize multi-dimensional fiber optic structure parameters, avoiding the problems of reliance on experience and over-simulation in traditional methods. Through large-scale parameter optimization, the optimal design scheme can be quickly found.

[0085] By training both the forward performance prediction model and the reverse structural design model, tedious multiple simulation calculations are avoided, significantly saving computational resources and time while ensuring design accuracy. Combining the forward and reverse models allows for flexible adjustment of the fiber optic structure according to different performance requirements, quickly responding to changes in sensor performance. This flexibility enables fiber optic sensors to be used more widely and accurately under various environmental conditions. The combination of the forward performance prediction model and the reverse structural design model achieves a closed loop in structural design and performance verification, ensuring the accuracy and reliability of the design scheme in practical applications. The model optimization and verification steps effectively improve the sensor's temperature response sensitivity.

[0086] Through the aforementioned intelligent design and optimization methods, a fiber optic temperature sensor with high sensitivity can be designed to meet more stringent temperature measurement requirements, providing a more efficient solution for precision measurement and industrial applications.

[0087] The introduction of artificial intelligence has accelerated the design process of fiber optic sensors, reducing the time spent on manual debugging and calculations, and improving the efficiency of the entire design cycle. By reducing the amount of computation and number of experiments required in traditional methods, it lowers the cost of the design process, saves development time and resources, and makes the application of the technology more economical. This method not only improves the sensitivity of the sensor but also ensures the stability and reliability of the structure, making it suitable for a wider range of industrial applications. It allows for flexible adjustment of the sensor's structure and performance according to actual application requirements, improving the applicability of fiber optic sensors in complex environments, especially in applications requiring high sensitivity and high accuracy temperature measurement.

[0088] Example 2

[0089] This embodiment is an explanation based on Embodiment 1. Please refer to it. Figure 1 Specifically: S1 includes S11 and S12;

[0090] S11. Determine the structure of the hollow-core anti-resonant fiber optic temperature sensor, including the diameter of the fiber core, the structure of the cladding, whether it is filled with liquid, and the type and location of the liquid;

[0091] The core of a hollow anti-resonant optical fiber is usually made of air, while the cladding tube is filled with different materials, such as ethanol, depending on the design requirements, and the ethanol refractive index is obtained.

[0092] The fiber core has a diameter of 30 μm, the cladding tube is made of quartz tubing, and the filling liquid is ethanol.

[0093] The refractive index of ethanol is obtained using the following formula:

[0094] n = n0 - a(T - T0);

[0095] In the formula, n represents the refractive index of ethanol at temperature T, n0 represents the refractive index of ethanol at reference temperature T0, and a represents the thermo-optical coefficient of ethanol, with a value of 3.94 × 10⁻⁶. -4 / ℃, T represents the experimental temperature, and T0 represents the reference temperature;

[0096] S12. Determine the structural parameters that need to be optimized in the sensor design, and set the optimization range for each parameter, including the cladding gap g and the cladding wall thickness t. tube and gold layer thickness t Au ;

[0097] The cladding gap g ranges from 2 μm to 6 μm; the cladding wall thickness t tube The range is 0.5 μm to 0.8 μm; gold layer thickness t Au The range is from 20nm to 50nm.

[0098] S2 includes S21 and S22;

[0099] S21. Using the finite element method, the hollow-core anti-resonant optical fiber with different structural configurations is simulated. The geometric structure of the optical fiber is defined, including the cladding gap g and the cladding wall thickness t. tube and gold layer thickness t Au The limiting loss of the optical fiber in the wavelength range of 1.44μm to 1.66μm was calculated using finite element simulation at different temperature ranges from 20℃ to 40℃ to evaluate the sensing performance of the optical fiber.

[0100] In the simulation process, the cladding tube gap g of the optical fiber is 2μm to 6μm, and the cladding tube wall thickness t is taken into account. tube The gold layer thickness ranges from 0.5 μm to 0.8 μm. Au The effects of 20nm to 50nm and temperature on optical fiber performance;

[0101] In this embodiment, during the design process, the core of the hollow anti-resonant optical fiber is composed of air, and a liquid, such as ethanol, is selected as the filling material according to requirements. This gives the sensor greater flexibility and customizability in design. By specifying the fiber core diameter, cladding structure, and liquid filling material and location, the sensing performance of the optical fiber can be precisely controlled.

[0102] By introducing the refractive index formula of ethanol and considering the effect of temperature changes on its refractive index, and using the thermo-optic coefficient for correction, this design method can accurately simulate the effect of ethanol on optical fiber performance at different temperatures, improving the accuracy of temperature sensing. By clearly defining the optimization range, such as cladding gap, tube wall thickness, and gold layer thickness, and setting reasonable optimization intervals for each parameter, parameter optimization can be performed more efficiently. This refined optimization method avoids the fuzzy and unsystematic parameter adjustments found in traditional design methods, ensuring the high performance of the sensor.

[0103] Using the finite element method (FEM) to calculate optical fibers with different structural configurations, especially for loss-limiting calculations in the wavelength range of 1.44 μm to 1.66 μm, allows for a more accurate evaluation of the fiber's sensing performance. This avoids the limitations of relying on experimental data and improves design efficiency. The FEM simulation simultaneously considers the effects of multiple factors on fiber performance, such as temperature changes, cladding tube gap, tube wall thickness, and gold layer thickness, making sensor design more adaptable across multiple dimensions, particularly for efficient sensing of temperature variations.

[0104] This embodiment employs a flexible and precise design approach. By specifically setting the range of each optimization parameter, a more accurate fiber optic structure design can be achieved to meet the needs of different application scenarios. By considering the influence of temperature on the fiber's refractive index, the sensor's sensitivity under different temperature conditions is improved, enabling the fiber optic temperature sensor to operate effectively over a wider temperature range. The design method, combining finite element simulation and artificial intelligence, avoids extensive experimental verification work, reduces time and resource consumption, and improves the efficiency and economy of the design process.

[0105] Example 3

[0106] This embodiment is an explanation based on Embodiment 2. Please refer to it. Figure 3 and Figure 4 Specifically: S22, collect fiber confinement loss data at temperatures between 20°C and 40°C and wavelengths between 1.44μm and 1.66μm; organize each simulation result, including structural parameters, temperature and wavelength, into a data table, with each row representing the fiber structure and its corresponding loss data at a temperature;

[0107] Finally, the different temperatures, wavelengths, and their corresponding limiting loss values ​​obtained from the simulation are organized to form a complete dataset D;

[0108] Dataset D is obtained using the following formula:

[0109] ;

[0110] In the formula, p1, p2, ..., pn represent optical fiber structural parameters, cladding gap g, cladding tube wall thickness t, etc. tube and gold layer thickness t Au .

[0111] S3 includes S31 and S32;

[0112] S31. Normalize the obtained dataset D, and construct and train a positive performance prediction model using the dataset D. The positive performance prediction model is a fully connected neural network model.

[0113] Input the cladding gap g and cladding wall thickness t into the forward performance prediction model tube Gold layer thickness t Au In conjunction with experimental temperature and environmental parameters, the output layer provides 12 numbers corresponding to the limiting loss at different wavelengths; the optimized activation function ReLU, learning rate, and training epochs are selected, and the test set is set to account for 10% for validation;

[0114] The calculation formula for the positive performance prediction model is as follows: y = ReLU(WD + b);

[0115] In the formula, y represents the output value, W represents the weight matrix, and b represents the bias term;

[0116] The positive performance prediction model is evaluated using the output value y, and the correlation coefficient R is obtained. 2 ;

[0117] Through the correlation coefficient R 2 To assess the accuracy of the prediction results from the positive performance prediction model;

[0118] The judgment method is as follows;

[0119] When 0.8 < R 2 A value ≤1 indicates that the prediction result is accurate;

[0120] When 0 < R 2 A value ≤0.8 indicates that the prediction result is inaccurate.

[0121] S32. Based on the output of the positive performance prediction model, train and construct the inverse structure design model;

[0122] The inputs to the inverse structural design model are the predicted limit loss curve and the experimental temperature, and the outputs are the structural parameters of the hollow antiresonant fiber, including the cladding gap g and the cladding wall thickness t. tube and gold layer thickness t Au ;

[0123] The reverse structure design model is a fully connected neural network model that uses the ReLU activation function. By adjusting the model parameters, the fiber structure can be accurately deduced from the predicted loss curve.

[0124] The calculation formula for the inverse model is as follows:

[0125] ;

[0126] In the formula, f -1 This represents the inverse function, and zp represents the output structure parameters.

[0127] In this embodiment, fiber optic confinement loss data were collected at different temperatures (20°C to 40°C) and wavelengths (1.44μm to 1.66μm), and each simulation result was organized into a data table to ensure data integrity and structure. This precise data management method provides a high-quality data foundation for subsequent model training. Normalization of the collected dataset D ensured data stability and consistency, which helps improve the training effect of the neural network model. Data normalization is a key step in ensuring that the model maintains efficiency and accuracy when processing data of different magnitudes.

[0128] The forward performance prediction model is trained using a fully connected neural network and employs hyperparameters such as the ReLU activation function, optimized learning rate, and number of training epochs to maximize prediction accuracy. This approach enables rapid evaluation of confinement loss and performance under different fiber structures, avoiding the cumbersome physical simulation process.

[0129] By using the fitting index R 2 This metric measures the accuracy of the model's predictions, ensuring a close match between the model's output and actual data. It provides a standardized quantitative basis for evaluating model quality, ensuring the reliability of the predictions. Using the loss-limiting curve from the forward performance prediction model's output, the inverse structural design model is further trained, allowing for the reverse calculation of optimized fiber optic structure parameters based on the predicted loss data. This process avoids the trial-and-error methods of traditional design, making fiber optic design more precise and efficient.

[0130] By employing fully connected neural networks (i.e., forward and inverse models), the complex relationships in the fiber optic design process (such as the mapping between structural parameters and performance) are transformed into efficient machine learning problems. This significantly improves the automation level of the design and optimization process.

[0131] By employing data-driven artificial intelligence models, the design of fiber optic sensors can be quickly and accurately predicted and optimized without relying on extensive physical experiments and repeated simulations, thus significantly improving design efficiency and accuracy. The use of simulation data and machine learning methods reduces the dependence on large amounts of experimental and computational resources in traditional design processes, lowering R&D costs. Normalization processing and model training methods effectively optimize the resources required for computation and experimentation.

[0132] By combining forward and inverse models, flexible adjustments and optimizations can be made according to different performance requirements (such as sensitivity requirements), enabling fiber optic temperature sensors to accurately meet the requirements of specific application scenarios and have wider adaptability.

[0133] By fitting the index R 2 Validating the forward performance prediction model ensures the accuracy of the model output, enhances the reliability and stability of the prediction results, and provides stronger data support for design decisions. Using the reverse design model, the corresponding fiber optic structure parameters can be quickly calculated based on the expected sensitivity or other performance requirements. This method significantly shortens the design cycle, improves response speed, and is particularly suitable for complex and variable engineering design tasks.

[0134] Example 4

[0135] This embodiment is an explanation based on Embodiment 3. Please refer to it. Figure 1 Specifically: S4. Utilize a positive performance prediction model to perform large-scale optimization of the above-determined optimization parameter range, including cladding gap g and cladding wall thickness t. tube and gold layer thickness t Au ;

[0136] By using a positive performance prediction model, different parameter combinations are predicted, and the corresponding limiting losses are predicted. A grid search optimization algorithm is used to find the structural configuration that maximizes the sensor sensitivity.

[0137] Based on the structural configuration of the sensor sensitivity, select the optimal fiber optic design parameters;

[0138] The sensitivity calculation formula is as follows: S = Δλ / ΔT;

[0139] In the formula, S represents the sensitivity of the sensor, Δλ represents the change in limiting loss, and ΔT represents the change in temperature.

[0140] The formula relating loss to structural parameters is as follows:

[0141] L(A,T) = fx(g,t) tube , t Au (T)

[0142] In the formula, fx represents the mapping function.

[0143] S5 includes S51 and S52;

[0144] S51. Using the sensitivity S of the sensor, the reverse structural design model is used to calculate the structural parameters that meet the sensitivity requirement.

[0145] Specifically, the sensor's sensitivity S is determined by the ratio between the limiting loss CL and the temperature change ΔT;

[0146] The sensor's sensitivity S is input into the reverse structure design model, and the limiting loss curve and temperature T in the wavelength range of the optical fiber from 1.44μm to 1.66μm are also input into the reverse structure design model.

[0147] The reverse structural design model calculates the fiber structure parameters nzp that meet the sensitivity requirements based on the reverse of the input, including optimizing the cladding gap ng and the cladding tube wall thickness nt. tube and optimize gold layer thickness nt Au ;

[0148] The fiber optic structure parameter nzp is obtained using the following formula:

[0149] .

[0150] S52. The obtained fiber structure parameter nzp is verified by a positive performance prediction model to ensure that its sensitivity meets expectations.

[0151] The forward performance prediction model calculates the new confinement loss nL(A,T) based on the fiber structure parameter nzp, and further calculates the new sensitivity nS;

[0152] The new limiting loss nL(A,T) is obtained by the following formula:

[0153] nL(A,T)=fx(ng,nt) tube nt Au (nT)

[0154] In the formula, nT represents the optimized temperature;

[0155] Sensitivity nS is obtained using the following formula:

[0156] ;

[0157] The obtained sensitivity nS is compared with the preset sensitivity threshold TS to determine the success of the design.

[0158] When the sensitivity nS ≥ the sensitivity threshold TS, the design is considered successful.

[0159] If the sensitivity nS < the sensitivity threshold TS, the design has failed and needs to be readjusted.

[0160] In this embodiment, a positive performance prediction model is used in conjunction with a grid search optimization algorithm to analyze the cladding gap g and the cladding wall thickness t. tube and gold layer thickness t Au The structural parameters are optimized over a wide range. This method efficiently finds the optimal configuration within the combination space of structural parameters, thereby maximizing the sensor's sensitivity and improving the accuracy and speed of the optimization process. The sensitivity calculation formula directly quantifies the fiber optic sensor's response to temperature changes. Accurate sensitivity calculation allows the design process to precisely align with actual performance requirements, ensuring efficient sensor operation within the desired temperature range.

[0161] Combining a forward performance prediction model with a grid search optimization algorithm allows for the traversal of all possible combinations of structural parameters, avoiding the inefficiencies of manual adjustment and random selection. This automated optimization method ensures that the fiber optic structure can quickly reach optimal performance. The reverse structural design model is used to calculate the fiber optic structure parameters that meet the sensitivity requirements, and this is then verified using the forward performance prediction model. This design method achieves a closed-loop design from target performance to structural parameters, avoiding repeated trial and error and improving the reliability and efficiency of the design.

[0162] By setting a sensitivity threshold TS, the success of the design is ensured. If the designed sensitivity meets the preset requirements, this verification mechanism ensures the sensor's performance is up to standard. This verification step provides clear performance standards and judgment criteria for the design, ensuring the controllability and repeatability of the design results. By combining a forward performance prediction model with optimization algorithms, fiber optic sensor design can be automated, avoiding the reliance on extensive experiments and simulation calculations in traditional design methods. The optimization process is more efficient, finding the optimal structural parameter configuration in a short time.

[0163] By reducing manual adjustments and experimental verification, the optimized combination of algorithms and models significantly lowers design and testing costs. This offers substantial economic benefits for the development and application of fiber optic temperature sensors, especially in situations requiring extensive experimentation. A closed-loop process of reverse design and forward verification ensures that the sensor design meets expected performance requirements. Each design step is validated, thereby improving the reliability of the final design and preventing designs that fail to meet performance requirements.

[0164] Through sensitivity calculations and optimized design, the fiber optic structure can be flexibly adjusted according to different temperature response requirements. This flexibility allows fiber optic sensors to adapt widely to various practical application scenarios, meeting sensing needs across different temperature ranges and performance requirements. By adjusting and verifying design parameters in real time, the performance of the fiber optic sensor can be continuously optimized, ensuring stable operation under various environmental conditions. Each optimization brings performance improvements, resulting in long-term stability and maintainability of the sensor.

[0165] Example 5

[0166] A hollow-core anti-resonant fiber optic temperature sensor based on artificial intelligence, please refer to... Figure 4 Specifically, it includes an optical fiber outer tube 1, a cladding tube 2 is provided inside the optical fiber outer tube 1, a gold plating layer 4 is provided on the inner wall of the cladding tube 2, a gap 5 between adjacent cladding tubes of the optical fiber is provided between the cladding tube 2 and the optical fiber outer tube 1, an internal region 3 of the optical fiber is provided inside the cladding tube 2, and the distance between the cladding tubes 2 is the optical fiber core diameter 6.

[0167] In this embodiment, the temperature sensing performance of the optical fiber is enhanced by precisely designing its multi-layered structure, especially the layout of the outer sheath, cladding tube, and gold plating. The gold plating effectively utilizes the surface plasmon resonance effect to improve the fiber's sensitivity, enabling it to provide higher measurement accuracy in complex environments. The arrangement of the cladding tube, as well as the design of the fiber's internal region and core, optimizes the fiber's transmission characteristics. Optimizing the cladding tube gap reduces fiber optic sensor loss, improves signal transmission efficiency, and enhances its temperature response performance.

[0168] Through meticulous fiber optic structure design (including the design of the fiber jacket, cladding tube, gold layer, and core diameter), the thermal stability and anti-interference capability of the fiber optic cable are effectively improved, providing more stable performance for the fiber optic temperature sensor. The gold plating and cladding tube design enhance the sensor's thermal response sensitivity while ensuring its stability under long-term use. The fiber optic structure enables accurate temperature measurement across different temperature ranges, ensuring the sensor's reliability in harsh environments.

[0169] By employing a well-designed gold plating layer and cladding tube, the optical fiber becomes more sensitive to temperature changes, enhancing the temperature sensor's sensitivity. This allows the fiber optic sensor to measure temperature variations more accurately, making it particularly suitable for applications requiring high-precision temperature monitoring. Through the introduction of the gold plating layer and optimized cladding tube design, the fiber effectively suppresses external interference, ensuring stable operation under high temperature, humidity, or other environmental pressures. This significantly improves the sensor's reliability in practical applications, making it especially suitable for fields such as industry and scientific research where high temperature accuracy is crucial.

[0170] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A design method for a hollow-core anti-resonant fiber optic temperature sensor based on artificial intelligence, characterized in that: Includes the following steps: S1. Determine the initial structure, parameters to be optimized, and optimization range of the hollow-core anti-resonant fiber optic temperature sensor; S2. Dataset D is constructed using finite element numerical simulation. S3. Train a positive performance prediction model using dataset D, and combine the positive performance prediction model to construct an inverse structure design model; S4. Utilize positive performance prediction models to achieve large-scale parameter optimization; S5. Implement the structural design of a fiber optic temperature sensor with specific performance using a reverse structural design model.

2. The design method of a hollow-core anti-resonant fiber optic temperature sensor based on artificial intelligence according to claim 1, characterized in that: S1 includes S11 and S12; S11. Determine the structure of the hollow-core anti-resonant fiber optic temperature sensor, including the diameter of the fiber core, the structure of the cladding, whether it is filled with liquid, and the type and location of the liquid; The air portion of the hollow anti-resonant optical fiber is filled with different materials, such as ethanol, depending on the design requirements, and the refractive index of ethanol is obtained. The fiber core has a diameter of 30 μm, the cladding tube is made of quartz tubing, and the filling liquid is ethanol. The refractive index of ethanol is obtained using the following formula: n = n0 - a(T - T0); In the formula, n represents the refractive index of ethanol at temperature T, n0 represents the refractive index of ethanol at reference temperature T0, a represents the thermo-optic coefficient of ethanol, T represents the experimental temperature, and T0 represents the reference temperature. S12. Determine the structural parameters that need to be optimized in the sensor design, and set the optimization range for each parameter, including the cladding gap g and the cladding wall thickness t. tube and gold layer thickness t Au ; The cladding gap g ranges from 2 μm to 6 μm; the cladding wall thickness t tube The range is 0.5 μm to 0.8 μm; gold layer thickness t Au The range is from 20nm to 50nm.

3. The design method of a hollow-core anti-resonant fiber optic temperature sensor based on artificial intelligence according to claim 2, characterized in that: S2 includes S21 and S22; S21. Using the finite element method, the hollow-core anti-resonant optical fiber with different structural configurations is simulated. The geometric structure of the optical fiber is defined, including the cladding gap g and the cladding wall thickness t. tube and gold layer thickness t Au Furthermore, within a temperature range of 20℃ to 40℃, the limiting loss CL of the optical fiber in the wavelength range of 1.44μm to 1.66μm was calculated using finite element simulation to evaluate the sensing performance of the optical fiber. In the simulation process, the cladding tube gap g of the optical fiber is 2μm to 6μm, and the cladding tube wall thickness t is taken into account. tube The gold layer thickness ranges from 0.5 μm to 0.8 μm. Au The effects of 20nm to 50nm and temperature on optical fiber performance.

4. The design method of a hollow-core anti-resonant fiber optic temperature sensor based on artificial intelligence according to claim 3, characterized in that: S22. Collect fiber optic confinement loss data between 20℃ and 40℃ and between 1.44μm and 1.66μm wavelengths; compile each simulation result, including structural parameters, temperature, wavelength, and corresponding confinement loss CL, into a complete dataset D. Dataset D is obtained using the following formula: ； In the formula, p1, p2, ..., pn represent the optical fiber structure parameters and corresponding wavelengths, and the cladding gap g and cladding wall thickness t represent the optical fiber structure parameters. tube and gold layer thickness t Au .

5. The design method of a hollow-core anti-resonant fiber optic temperature sensor based on artificial intelligence according to claim 4, characterized in that: S3 includes S31 and S32; S31. Normalize the obtained dataset D, and construct and train a positive performance prediction model using the dataset D. The positive performance prediction model is a fully connected neural network model. Input the cladding gap g and cladding wall thickness t into the forward performance prediction model tube Gold layer thickness t Au The output layer provides 12 numbers corresponding to the limiting loss at different wavelengths, along with the experimental temperature; the optimized activation function ReLU, learning rate, and training epochs are selected, and the test set is set to account for 10% for validation; The calculation formula for the positive performance prediction model is as follows: y = ReLU(WD + b); In the formula, y represents the output value, W represents the weight matrix, and b represents the bias term; Correlation coefficient R 2 This is used to evaluate the positive performance prediction model based on the output value y, and the judgment method is as follows; When 0.8 < R 2 A value ≤1 indicates that the prediction result is accurate; When 0 < R 2 A value ≤0.8 indicates that the prediction result is inaccurate.

6. The design method of a hollow-core anti-resonant fiber optic temperature sensor based on artificial intelligence according to claim 5, characterized in that: S32. Based on the output of the positive performance prediction model, train and construct the inverse structure design model; The inputs to the inverse structural design model are the estimated limit loss curve and the experimental temperature, and the outputs are the structural parameters of the hollow antiresonant fiber, including the cladding gap g and the cladding wall thickness t. tube and gold layer thickness t Au ; The reverse structure design model is a fully connected neural network model that uses the ReLU activation function. By adjusting the model parameters, the fiber structure can be accurately deduced from the predicted loss curve. The calculation formula for the inverse model is as follows: ； In the formula, f -1 This represents the inverse function, and zp represents the output structure parameters.

7. The design method of a hollow-core anti-resonant fiber optic temperature sensor based on artificial intelligence according to claim 6, characterized in that: S4. Utilize a positive performance prediction model to perform large-scale optimization of the above-determined parameter range, including the cladding gap g and the cladding wall thickness t. tube and gold layer thickness t Au ; By using a positive performance prediction model, different parameter combinations are predicted, and the corresponding limiting losses are predicted. A grid search optimization algorithm is used to find the structural configuration that maximizes the sensor sensitivity. Based on the structural configuration of the sensor sensitivity, select the optimal fiber optic design parameters; The sensitivity calculation formula is as follows: S = Δλ / ΔT; In the formula, S represents the sensitivity of the sensor, Δλ represents the change in wavelength corresponding to the limiting loss, and ΔT represents the temperature change.

8. The design method of a hollow-core anti-resonant fiber optic temperature sensor based on artificial intelligence according to claim 7, characterized in that: S5 includes S51 and S52; S51. Using the sensor's sensitivity S, the reverse structural design model calculates the structural parameters that meet the sensitivity requirements. Specifically, the sensor's sensitivity S is determined by the ratio between the wavelength offset Δλ corresponding to the limiting loss and the temperature change ΔT. The sensor's sensitivity S is input into the reverse structure design model, and the limiting loss curve and temperature T in the wavelength range of the optical fiber from 1.44μm to 1.66μm are also input into the reverse structure design model. The reverse structural design model calculates the fiber structure parameters nzp that meet the sensitivity requirements based on the reverse of the input, including optimizing the cladding gap ng and the cladding tube wall thickness nt. tube and optimize gold layer thickness nt Au ; The fiber optic structure parameter nzp is obtained using the following formula: 。 9. The design method of a hollow-core anti-resonant fiber optic temperature sensor based on artificial intelligence according to claim 8, characterized in that: S52. The obtained fiber structure parameter nzp is verified by a positive performance prediction model to ensure that its sensitivity meets expectations. The forward performance prediction model calculates the new confinement loss CL based on the fiber structure parameter nzp, and further calculates the new sensitivity nS; The obtained sensitivity nS is compared with the preset sensitivity threshold TS to determine the success of the design. When the sensitivity nS ≥ the sensitivity threshold TS, the design is considered successful. If the sensitivity nS < the sensitivity threshold TS, the design has failed and needs to be readjusted.

10. A design method for a hollow-core anti-resonant fiber optic temperature sensor based on artificial intelligence, applied to the hollow-core anti-resonant fiber optic temperature sensor based on artificial intelligence as described in any one of claims 1 to 9, characterized in that: The optical fiber includes an outer jacket tube (1), inside which a cladding tube (2) is provided. On the inner wall of the cladding tube (2), a gold plating layer (4) is provided inside the cladding tube. A gap (5) between adjacent cladding tubes of the optical fiber is provided between the cladding tube (2) and the outer jacket tube (1). An inner region (3) of the optical fiber is provided inside the cladding tube (2). The distance between the cladding tubes (2) is the diameter of the optical fiber core (6).

Images