Method for manufacturing microstructured optical fiber and raman spectrometer light collection system thereof

Abstract

This application relates to the field of spectroscopic detection technology, specifically disclosing a method for fabricating a microstructure-enhanced optical fiber and a light collection system for a Raman spectrometer. The system includes: a light source assembly for generating laser light and collimating the laser light to irradiate a sample to be tested; a detection area assembly for placing and positioning the sample; a collection assembly for collecting Raman scattered light generated by the sample under laser irradiation based on the microstructure-enhanced optical fiber, wherein the fiber end face of the microstructure-enhanced optical fiber is provided with micro / nano structures, and the geometric parameters of the micro / nano structures are determined by a genetic algorithm for global optimization with the optimization objective of maximizing the coupling efficiency for unpolarized diffuse light within a preset wavelength range; and a control and analysis assembly for acquiring signals from the collected Raman scattered light and, based on the signal acquisition results, identifying the component information corresponding to the sample to be tested through a preset component identification model.

Description

Technical Field

[0001] This application relates to the field of spectroscopic detection technology, and in particular to a method for preparing a microstructure-enhanced optical fiber and its Raman spectrometer light collection system. Background Technology

[0002] Raman spectrometer optical collection systems, as powerful non-destructive analytical tools, have demonstrated indispensable value in fields such as chemistry, biomedicine, environmental monitoring, and materials science. Their core technology relies on the efficient capture and precise analysis of Raman scattering signals generated after substances are stimulated. In this system, the optical fiber serves as the light collection channel, and its performance is crucial, directly determining the instrument's sensitivity and reliability in detecting weak signals. In recent years, with the cross-integration of micro-nano optics and fiber optic technology, modulating the light field through microstructures on the fiber endface to improve performance has become an important development direction in this field.

[0003] However, existing Raman signal collection schemes based on microstructured optical fibers still face significant limitations. On the one hand, traditional fiber end-face structures (such as regular gratings) are usually based on simplified models or single-condition designs, making it difficult to maintain high coupling efficiency over a wide spectral range, resulting in severe loss of signals in some frequency bands during Raman scattering. On the other hand, Raman signals themselves are extremely weak and exhibit diffuse characteristics, while existing microstructures show a sharp decrease in response to large-angle incident light, making the system extremely sensitive to small shifts in sample position, severely affecting the repeatability and stability of quantitative analysis. Summary of the Invention

[0004] In view of this, this application provides a method for fabricating microstructure-enhanced optical fibers and a light collection system for a Raman spectrometer. By globally optimizing the micro-nano structure of the end face of the microstructure-enhanced optical fiber using a genetic algorithm, Fresnel reflection can be effectively suppressed and the angular receiving range can be greatly widened. This allows for the capture of more scattered photons as a whole, maximizing the coupling of unpolarized diffuse light over a wide spectral range, effectively reducing signal loss, and improving the response to large-angle incident light. This results in a higher tolerance for small deviations in the focusing position of the sample to be detected, reducing the sensitivity to sample position shifts, improving the performance of the optical fiber, and thus enhancing the stability, sensitivity, and reliability of the entire system in detecting weak signals.

[0005] According to one aspect of this application, a light collection system for a Raman spectrometer based on microstructure-enhanced optical fiber is provided, comprising:

[0006] A light source assembly is used to generate laser light and collimate the laser light so that the collimated laser light illuminates the sample to be tested.

[0007] A detection area component is used to place the sample to be detected and to position the sample to be detected.

[0008] A collection component is used to collect the Raman scattered light generated by the sample to be tested under laser irradiation. It includes a microstructure-enhanced optical fiber. The fiber end face of the microstructure-enhanced optical fiber is provided with a micro-nano structure. The geometric parameters of the micro-nano structure are determined by a genetic algorithm for global optimization with the optimization objective of maximizing the coupling efficiency of non-polarized diffuse light within a preset wavelength range.

[0009] The control and analysis component is used to acquire signals from the collected Raman scattered light and, based on the signal acquisition results, identify the component information corresponding to the sample to be detected through a preset component identification model.

[0010] According to another aspect of this application, a method for fabricating a microstructure-enhanced optical fiber is provided, the method comprising:

[0011] The target band of the Raman scattered light is determined, and the target band is decomposed according to a preset band length to obtain multiple sub-bands;

[0012] A fitness function for the genetic algorithm is defined, wherein the fitness function is used to evaluate the coupling efficiency of the micro / nano structure for unpolarized diffuse light within a preset wavelength range, and the coupling efficiency is proportional to the collection efficiency of the Raman scattered light.

[0013] For each sub-band, a genetic algorithm is used to optimize the geometric parameters of the micro / nano structure corresponding to the sub-band, with the fitness function as the optimization objective. The optimized geometric parameters of the micro / nano structure corresponding to the sub-band are output, wherein the geometric parameters include at least one of the period, height, duty cycle, and tilt angle of the micro / nano structure.

[0014] Based on the optimized geometric parameters of the micro-nano structures corresponding to each sub-band, micro-nano structure arrays are fabricated on the end face of the optical fiber body using micro-nano fabrication methods to obtain microstructure-enhanced optical fibers that can enhance the Raman scattering light collection efficiency of the target band.

[0015] According to another aspect of this application, a storage medium is provided that stores a computer program thereon, which, when executed by a processor, implements the above-described method for fabricating microstructure-enhanced optical fibers.

[0016] According to another aspect of this application, a computer device is provided, including a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, wherein the processor executes the program to implement the above-described method for fabricating microstructure-enhanced optical fibers.

[0017] By employing the above technical solutions, this application provides a method for fabricating microstructure-enhanced optical fibers and a light collection system for a Raman spectrometer. Through global optimization of the micro-nano structure of the microstructure-enhanced optical fiber end face using a genetic algorithm, Fresnel reflection can be effectively suppressed, and the angular receiving range can be greatly widened. This allows for the capture of more scattered photons overall, maximizing the coupling of unpolarized diffuse light over a wide spectral range, effectively reducing signal loss, and improving the response to large-angle incident light. This results in higher tolerance for small deviations in the focusing position of the sample to be detected, reduced sensitivity to sample position shifts, and improved fiber performance. Consequently, the stability, sensitivity, and reliability of the entire system in detecting weak signals are enhanced.

[0018] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0019] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0020] Figure 1 This illustration shows a schematic diagram of a Raman spectrometer light collection system based on microstructure-enhanced optical fiber, according to an embodiment of this application.

[0021] Figure 2 This paper shows a schematic diagram of another Raman spectrometer light collection system based on microstructure-enhanced optical fiber provided in an embodiment of this application;

[0022] Figure 3 This paper shows an integral graph of the coupling efficiency measurement of a microstructure-enhanced optical fiber under 900 nm Gaussian incident light, according to an embodiment of this application.

[0023] Figure 4 This illustration shows a schematic diagram of collecting weak signals using an optical fiber endface and a three-zone grating, according to an embodiment of this application.

[0024] Figure 5 A schematic flowchart of a method for fabricating a microstructure-enhanced optical fiber according to an embodiment of this application is shown;

[0025] Figure 6 This illustration shows a schematic diagram of the structure of a neural network model provided in an embodiment of this application;

[0026] Figure 7 This illustration shows a spectrum obtained by detecting cerium oxide (CeO) using a microstructure-enhanced fiber Raman spectrometer, as provided in an embodiment of this application.

[0027] Figure 8 A schematic diagram of the device structure of a computer device provided in an embodiment of this application is shown. Detailed Implementation

[0028] The present application will be described in detail below with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in the embodiments of the present application can be combined with each other.

[0029] This embodiment provides a Raman spectrometer light collection system based on microstructure-enhanced optical fiber, such as... Figure 1 As shown, the system includes:

[0030] A light source assembly is used to generate laser light and collimate the laser light so that the collimated laser light illuminates the sample to be tested.

[0031] A detection area component is used to place the sample to be detected and to position the sample to be detected.

[0032] A collection component is used to collect the Raman scattered light generated by the sample to be tested under laser irradiation. It includes a microstructure-enhanced optical fiber. The fiber end face of the microstructure-enhanced optical fiber is provided with a micro-nano structure. The geometric parameters of the micro-nano structure are determined by a genetic algorithm for global optimization with the optimization objective of maximizing the coupling efficiency of non-polarized diffuse light within a preset wavelength range.

[0033] The control and analysis component is used to acquire signals from the collected Raman scattered light and, based on the signal acquisition results, identify the component information corresponding to the sample to be detected through a preset component identification model.

[0034] This application provides a Raman spectrometer light collection system based on microstructure-enhanced optical fiber, which can be applied to fields requiring high sensitivity, such as mineral analysis, biomedical analysis (e.g., non-invasive detection of low concentrations of cancer biomarkers in urine and blood), environmental monitoring analysis, and food safety analysis. It provides a microstructure-enhanced optical fiber that combines intelligent design, high-precision manufacturing, and superior optical performance, overcoming the efficiency bottleneck of traditional optical fibers in Raman spectrometer light collection. Specifically, it can include: a light source component, a detection zone component, a collection component, and a control and analysis component.

[0035] Light source assembly: The core function of this assembly is to generate laser light, which is the fundamental energy source for Raman spectrometer detection. After generating the laser, it can be collimated to ensure that the laser accurately illuminates the sample to be tested. The collimated laser has better directionality and focusing, and can act on the sample to be tested in a more concentrated form, providing stable excitation conditions for the subsequent excitation of Raman scattered light from the sample, making the excitation process more efficient and accurate, thus laying a good foundation for the entire detection process.

[0036] Detection Zone Component: The main function of this component is to provide a suitable placement space for the sample to be tested. In actual testing, the sample needs to be accurately placed in a specific location so that the laser can irradiate it as intended. Simultaneously, the detection zone component also has the function of positioning the sample. Accurate positioning ensures that the sample remains in a fixed position during testing, avoiding deviations in test results due to sample movement. This ensures that the laser accurately targets the same area of ​​the sample in each test, improving the accuracy and repeatability of the test.

[0037] Collection Component: This component collects Raman scattered light based on microstructure-enhanced optical fiber. The fiber end face of the microstructure-enhanced fiber is equipped with micro / nanostructures. The geometric parameters of the micro / nanostructures are not arbitrarily set, but determined through global optimization using a genetic algorithm. The genetic algorithm aims to maximize the coupling efficiency for unpolarized diffuse light within a preset wavelength range. By simulating natural selection and genetic mechanisms in biological evolution, iteratively optimizes the geometric parameters of the micro / nanostructures. The optimized micro / nanostructures can more effectively collect Raman scattered light generated by the sample under laser irradiation, improving light collection efficiency and reducing light signal loss, thereby providing richer optical signal data for subsequent accurate component analysis.

[0038] Control and Analysis Component: This component is responsible for extracting useful information from the collected Raman scattered light. Specifically, it first acquires the signal, converting the collected Raman scattered light into a processable electrical signal or other form of signal. Then, based on the acquired signal results, it uses a preset component identification model to identify the components. The preset component identification model can be obtained by learning and training on the Raman scattered light signals and corresponding component information of a large number of known samples. It can quickly and accurately identify the component information corresponding to the sample to be detected based on the characteristics of the input Raman scattered light signal, realizing the conversion from light signal to sample component, and completing the detection and analysis functions of the entire Raman spectrometer light collection system.

[0039] In a specific embodiment, such as Figure 2The image shows another Raman spectrometer light collection system based on microstructure-enhanced fiber provided in this application embodiment. This light collection system comprises four core components: A light source component: Laser 1 outputs laser light with wavelengths of 190nm-1064nm, which is collimated by incident fiber 2 and a first lens group 3 before illuminating the sample to be tested; A detection area component: A high-throughput sample disk 4 is driven by a stepper motor to achieve automatic positioning of the sample to be tested; A collection component: A filter element 5 filters out Rayleigh scattered light, a second lens group 6 collects Raman scattered light, and the end face of the microstructure-enhanced fiber 7 is provided with a trapezoidal-spiral composite grating optimized by a genetic algorithm to further collect Raman scattered light; A control and analysis component: A spectral measurement device 8 acquires signals from the collected Raman scattered light and uses a built-in deep learning model (i.e., a preset component recognition model) to achieve rapid component identification of the sample to be tested. Figure 3 The figure shows an integral graph of the coupling efficiency measurement of a microstructure-enhanced optical fiber provided in an embodiment of this application under 900 nm Gaussian incident light.

[0040] By applying the technical solution of this embodiment, the micro-nano structure of the fiber end face is enhanced through global optimization of the microstructure using a genetic algorithm. This effectively suppresses Fresnel reflection and greatly expands the angle receiving range, thereby capturing more scattered photons as a whole. This maximizes the coupling of unpolarized diffuse light across a wide spectral range, effectively reducing signal loss and improving the response to large-angle incident light. It also increases the tolerance for small deviations in the focusing position of the sample to be detected, reduces the sensitivity to sample position shift, and improves fiber performance. In turn, it enhances the stability, sensitivity, and reliability of the entire system in detecting weak signals.

[0041] In this embodiment of the application, optionally, the microstructure-enhanced optical fiber includes an optical fiber body and a micro / nano structure disposed on the end face of the optical fiber body; wherein, the micro / nano structure includes a three-dimensional structure, the three-dimensional structure is an aperiodic structure or a composite periodic structure, and the geometric parameters of the micro / nano structure are determined by optimization using a genetic algorithm, wherein the fitness function of the genetic algorithm is configured to maximize the coupling efficiency for unpolarized diffuse light within a preset wavelength range.

[0042] In this embodiment, the microstructure-enhanced fiber uses the fiber body as its basic carrier. The fiber body possesses the basic function of transmitting optical signals in conventional optical fibers, providing a channel for optical transmission for the entire optical collection system. By constructing micro- and nanostructures on the end face of the fiber body, the optical properties of the original fiber end face can be altered, thereby creating conditions for improving optical collection performance and enabling the fiber to interact with external optical signals in a more efficient manner.

[0043] Micro- and nanostructures can be three-dimensional and possess special properties, namely, aperiodic or composite periodic structures. Aperiodic structures break the regularity of traditional periodic structures, enabling the generation of more complex and diverse light field distributions. This complex light field distribution helps to better capture and couple light signals with different directions and characteristics, especially for diffuse light, which exhibits strong directional randomness; aperiodic structures can provide more coupling opportunities. Composite periodic structures, on the other hand, combine some of the ordered properties of periodic structures with the flexibility of aperiodic structures. They can utilize the efficient coupling advantages of periodic structures under specific conditions and, through composite design, adapt to more complex light signal scenarios, further enhancing the ability to collect diffuse light.

[0044] The geometric parameters of the micro / nano structures are determined through optimization using a genetic algorithm. Specifically, the genetic algorithm aims to maximize the coupling efficiency for unpolarized diffuse light within a preset wavelength range. This involves filtering and iterating through a large number of geometric parameter combinations to continuously find the optimal combination that achieves the best coupling efficiency. This global optimization approach avoids getting trapped in local optima, resulting in micro / nano structure geometric parameters that better meet practical requirements.

[0045] The fitness function of the genetic algorithm is configured within a preset wavelength range, with the core objective of maximizing the coupling efficiency for unpolarized diffuse light. The fitness function is the criterion used in the genetic algorithm to evaluate the merits of each individual (i.e., different geometric parameters). Within the preset wavelength range, the fitness function can quantitatively evaluate the coupling efficiency corresponding to each individual. The higher the coupling efficiency, the higher the fitness of that individual in the genetic algorithm, and the more likely it is to be retained for subsequent crossover and mutation operations. In this way, the genetic algorithm can guide the parameter search towards improving coupling efficiency, ultimately determining the micro / nanostructure geometric parameters that can achieve efficient coupling of unpolarized diffuse light within the preset wavelength range, thereby improving the overall microstructure-enhanced fiber's Raman scattered light collection performance.

[0046] In a specific embodiment, if an enhanced micro / nanostructure in the 400-700nm wavelength range is required, preset wavelength ranges can be set as 400nm~500nm, 500nm~600nm, and 600nm~700nm, with a preset band length of 100nm. Micro / nanostructures for each preset wavelength range are optimized using a genetic algorithm, and then these three micro / nanostructures are spliced ​​together to form the desired composite periodic structure. For example... Figure 4 The diagram illustrates a method for collecting (or detecting) weak signals using an optical fiber end face and a three-segment spliced ​​grating. The three-segment spliced ​​grating is a composite periodic structure composed of three parts, each corresponding to a micro / nano structure within a preset wavelength range.

[0047] In this embodiment of the application, the micro / nano structure may optionally be a grating, a nanopore array, or a plasmonic resonance structure.

[0048] In this embodiment, the micro / nano structure is set as a grating, a nanopore array, or a plasmon resonance structure. Each of these three structures has unique optical properties, meeting the needs of Raman scattering light collection in different scenarios. Grating: With periodic structural features, it can produce a diffraction effect on incident light. By rationally designing the grating parameters, such as period and duty cycle, the direction and intensity of the diffracted light can be controlled, thereby achieving efficient coupling and collection of Raman scattering light of specific wavelengths. Nanopore array: Composed of a series of nanoscale holes, a near-field enhancement effect occurs when light shines on the nanopore array. This effect can enhance the interaction between light and matter, improving the excitation and collection efficiency of Raman scattering light, especially suitable for detecting weak signals. Plasmon resonance structure: Utilizing the plasmon resonance characteristics of metal nanoparticles or nanostructures. When the frequency of the incident light matches the plasmon resonance frequency of the metal nanostructure, a strong local electric field enhancement is generated, significantly increasing the intensity of Raman scattering and enhancing the ability to collect Raman signals.

[0049] In this embodiment of the application, optionally, when the micro-nano structure is a grating, the grating is fabricated on the end face of the optical fiber body based on a micro-nano fabrication method, and the cross-sectional type of the grating includes triangle, circle, square, trapezoid and combination thereof.

[0050] In this embodiment, different types of cross-sections can have different effects on light propagation and coupling. Triangular cross-sections: possess unique diffraction characteristics; their sharp edges generate a strong scattering effect, helping to enhance the collection of Raman scattered light. Simultaneously, the direction and intensity of diffracted light can be flexibly controlled by adjusting parameters such as side length and angle. Circular cross-sections: possess rotational symmetry, resulting in relatively uniform optical properties in all directions. This structure can reduce the polarization dependence of light during propagation, improving the coupling efficiency for unpolarized Raman scattered light. Square cross-sections: simple in structure and easy to manufacture. Square cross-section gratings have the same period in both horizontal and vertical directions, producing a more regular diffraction pattern, facilitating the directional collection and analysis of Raman scattered light. Trapezoidal cross-sections: combining some characteristics of triangles and rectangles, the diffraction efficiency and coupling characteristics of the grating can be optimized by adjusting parameters such as the upper base, lower base, and height of the trapezoid. Combining different types of cross-sections allows for the comprehensive utilization of the advantages of various cross-sections, further enhancing the grating's ability to collect Raman scattered light. For example, combining triangular and circular cross sections can enhance the scattering effect while reducing polarization dependence and improving the overall performance of the system.

[0051] Optionally, in the embodiments of this application, the micro-nano fabrication method of the grating includes electron beam lithography, focused ion beam etching, and two-photon polymerization 3D printing; when the micro-nano fabrication method of the grating is two-photon polymerization 3D printing, the two-photon laser parameters are: wavelength of 700~1000nm; laser power of 10~50mW; and laser scanning speed of 1000~25000μm / s.

[0052] This embodiment combines the powerful design capabilities of genetic algorithms with the precision manufacturing capabilities of two-photon polymerization 3D printing for the first time, solving the dilemmas of "optimal designs being unmanufacturable" and "manufacturable designs being suboptimal." This integrated design-manufacturing process can create novel optical components with performance far exceeding the limitations of traditional processes.

[0053] In one specific embodiment, a trapezoidal grating multimode fiber fabrication process based on two-photon polymerization (TPP) is proposed: A femtosecond laser (35mW power, 5000μm / s scanning speed) with a wavelength of 700nm-1000nm is used to expose IP-Dip photoresist, achieving 200nm feature size control (SEM measured deviation <5%). Exposure parameters are adjusted via real-time confocal microscopy feedback to ensure processing consistency (batch-to-batch deviation <5%). Performance testing of the fabricated fiber using a Raman spectrometer light collection system shows that the spectral response improvement in the 850nm-950nm band is superior to that of bare fiber.

[0054] Optionally, in this embodiment, the size of the micro / nano structure is adapted to the optical fiber body, and the constituent material of the micro / nano structure is a metal, dielectric, polymer material or a combination thereof; the optical fiber body includes at least one of single-mode optical fiber, multimode optical fiber, multi-core optical fiber, microstructured optical fiber and on-chip optical waveguide.

[0055] Furthermore, as Figure 1 In terms of specific system implementation, this application provides a method for fabricating microstructure-enhanced optical fibers, such as... Figure 5 As shown, the method includes:

[0056] Step 101: Determine the target band of the Raman scattered light, and decompose the target band according to the preset band length to obtain multiple sub-bands.

[0057] Step 102: Set the fitness function of the genetic algorithm, wherein the fitness function is used to evaluate the coupling efficiency of the micro / nano structure for unpolarized diffuse light within a preset wavelength range, and the coupling efficiency is proportional to the collection efficiency of the Raman scattered light.

[0058] Step 103: For each sub-band, a genetic algorithm is used to optimize the geometric parameters of the micro / nano structure corresponding to the sub-band, with the fitness function as the optimization objective. A set of optimized geometric parameters of the micro / nano structure corresponding to the sub-band is output, wherein the geometric parameters include at least one of the period, height, duty cycle, and tilt angle of the micro / nano structure.

[0059] Step 104: Based on the optimized geometric parameters of the micro / nano structures corresponding to each sub-band, micro / nano structure arrays are fabricated on the end face of the optical fiber body using micro / nano fabrication methods to obtain microstructure-enhanced optical fibers that can enhance the Raman scattering light collection efficiency of the target band.

[0060] This application provides a method for fabricating a microstructure-enhanced optical fiber. First, the target wavelength band of the Raman scattered light can be determined. Raman scattered light often contains rich spectral information, and different wavelength bands correspond to different material characteristics and physical processes in the sample. Therefore, the target wavelength band can be decomposed according to a preset wavelength length to obtain multiple sub-bands. This operation aims to refine the processing of complex spectral problems. Since Raman scattered light in different wavelength bands has different characteristics, such as intensity and polarization state, after decomposing the target wavelength band into sub-bands, independent micro / nano structure designs can be performed for the characteristics of each sub-band. This allows for more precise optimization of the light collection performance of each sub-band, improving the overall collection efficiency and accuracy of Raman scattered light.

[0061] Next, the fitness function of the genetic algorithm can be set. The fitness function is used to evaluate the coupling efficiency of the micro / nanostructure to unpolarized diffuse light within a preset wavelength range. In Raman spectroscopy detection, the Raman signal is typically unpolarized diffuse light with relatively weak intensity. The coupling efficiency directly determines the amount of Raman scattered light that can be collected, thus affecting the sensitivity and accuracy of the detection. By setting a fitness function with coupling efficiency as the evaluation index, the genetic algorithm has a clear optimization direction: finding the micro / nanostructure geometric parameters that maximize the coupling efficiency, thereby ensuring that the fabricated microstructure-enhanced fiber can efficiently collect Raman scattered light. It is important to note that the preset wavelength range here refers to the wavelength range corresponding to each sub-band. A set of optimal geometric parameters can be obtained for each sub-band using the genetic algorithm. Here, the geometric parameters can include at least one of the micro / nanostructure's period, height, duty cycle, and tilt angle.

[0062] Furthermore, for each sub-band obtained from the decomposition, a genetic algorithm is employed, with a set fitness function as the optimization objective, to optimize the geometric parameters of the corresponding micro / nano structure for that sub-band. The genetic algorithm is an optimization algorithm that simulates the biological evolution process. It performs a global search in the parameter space by simulating operations such as natural selection, crossover, and mutation. In this process, the algorithm generates a large number of geometric parameters (individuals) and evaluates the quality of each individual based on the fitness function. Individuals with high fitness (i.e., geometric parameters with high coupling efficiency) can be retained for subsequent crossover and mutation operations to generate new individuals. After multiple iterations, the algorithm gradually converges to the optimal geometric parameters, outputting a set of optimized geometric parameters corresponding to the micro / nano structure of each sub-band. This global search approach avoids getting trapped in local optima and increases the probability of finding the optimal solution.

[0063] Finally, based on the optimized geometric parameters of the micro / nano structures corresponding to each sub-band, micro / nano structure arrays were fabricated on the fiber end face of the optical fiber body using micro / nano fabrication methods. Micro / nano fabrication methods are characterized by high precision and high resolution, enabling the accurate fabrication of micro / nano structures according to the optimized geometric parameters. By splicing together the micro / nano structures corresponding to different sub-bands to form a complete micro / nano structure array, the optimized performance of each sub-band can be comprehensively utilized to achieve efficient collection of Raman scattered light across the entire target wavelength band. This splicing method considers both the characteristics of different sub-bands and ensures the integrity and consistency of the micro / nano structures on the fiber end face, ultimately resulting in a high-performance microstructure-enhanced optical fiber.

[0064] Optionally, in this embodiment, step 103, "using a genetic algorithm with the fitness function as the optimization objective to optimize the geometric parameters of the micro / nano structure corresponding to the sub-band and output a set of optimized geometric parameters of the micro / nano structure corresponding to the sub-band," includes: initializing multiple candidate geometric parameters and generating corresponding micro / nano structure cross-sectional images based on each candidate geometric parameter; inputting the micro / nano structure cross-sectional images corresponding to each candidate geometric parameter into a pre-trained neural network model and outputting the predicted coupling efficiency values ​​for unpolarized diffuse light corresponding to each candidate geometric parameter; updating the parameter values ​​of the multiple candidate geometric parameters based on the predicted coupling efficiency values ​​corresponding to each candidate geometric parameter to obtain new candidate geometric parameters, and recalculating the predicted coupling efficiency values ​​corresponding to each new candidate geometric parameter until the iteration termination condition is met, and using the candidate geometric parameter with the largest predicted coupling efficiency value as the optimized geometric parameters of the micro / nano structure corresponding to the sub-band.

[0065] In this embodiment, multiple candidate geometric parameters are first initialized. Each candidate geometric parameter may include key elements such as the period, height, duty cycle, and tilt angle of the micro / nano structure.

[0066] Once these candidate geometric parameters are obtained, corresponding micro / nanostructure cross-sectional diagrams can be generated based on each candidate parameter. This transforms abstract geometric parameters into intuitive graphical representations, facilitating subsequent analysis and processing. The process of generating micro / nanostructure cross-sectional diagrams can be implemented using computer-aided design software. Based on the input candidate geometric parameters, the software can accurately draw the cross-sectional shape of the micro / nanostructure. For example, if the geometric parameters specify that the micro / nanostructure is a rectangular columnar structure and provide parameters such as period and height, the software can draw the corresponding rectangular columnar micro / nanostructure cross-sectional diagram according to these parameters.

[0067] The pre-trained neural network model, having learned from a large amount of training data, has mastered the complex relationship between the geometric parameters of micro / nano structures and the coupling efficiency of unpolarized diffuse light. After obtaining the cross-sectional images of the micro / nano structures corresponding to each candidate geometric parameter, these cross-sectional images can be provided as input data to the pre-trained neural network model. The neural network model can process and analyze the input cross-sectional images. Specifically, through multiple neurons and layer structures, it can extract and transform the geometric features in the cross-sectional images of the micro / nano structures, and then, based on the learned knowledge, output a predicted value of the coupling efficiency for unpolarized diffuse light corresponding to each candidate geometric parameter. This predicted value reflects the level of coupling efficiency of the micro / nano structure for unpolarized diffuse light under given geometric parameters. For example, if the predicted coupling efficiency value output by the neural network model after inputting the cross-sectional image of the micro / nano structure corresponding to a certain candidate geometric parameter is high, it indicates that the micro / nano structure has a good coupling ability for unpolarized diffuse light under that geometric parameter.

[0068] After obtaining the predicted coupling efficiency value for each candidate geometric parameter, the candidate geometric parameters can be updated based on these predictions. The update strategy in the genetic algorithm includes operations such as selection, crossover, and mutation. First, candidate geometric parameters are selected based on the predicted coupling efficiency values, retaining those with higher predicted values ​​and eliminating those with lower values. Then, a crossover operation is performed on the retained candidate geometric parameters; that is, two candidate geometric parameters are randomly selected, and some of their features are exchanged to generate new candidate geometric parameters. Simultaneously, a mutation operation is performed on some candidate geometric parameters, which involves applying a certain degree of random perturbation to the current parameter values ​​to increase parameter diversity.

[0069] These operations yield new candidate geometric parameters. Then, based on these new candidate geometric parameters, corresponding micro / nano structure cross-sectional images are generated again. These new cross-sectional images are input into a pre-trained neural network model, and the predicted coupling efficiency value for each new candidate geometric parameter is recalculated. This process can be repeated until an iteration termination condition is met. The iteration termination condition could be reaching a preset number of iterations, or the change in the predicted coupling efficiency value being less than a certain threshold, etc.

[0070] When the iteration termination condition is met, the candidate geometric parameter with the largest predicted coupling efficiency is found from all candidate geometric parameters, and this candidate geometric parameter is used as the optimized geometric parameter of the micro / nano structure corresponding to this sub-band.

[0071] This application's embodiments achieve efficient optimization of the geometric parameters of micro / nano structures by combining genetic algorithms and pre-trained neural network models. Specifically, multiple candidate geometric parameters are initialized and cross-sectional images of the micro / nano structures are generated, providing abundant initial samples for subsequent optimization. The pre-trained neural network model is used to quickly and accurately predict coupling efficiency, avoiding complex direct optical simulation calculations and significantly improving optimization efficiency. The candidate geometric parameters are updated and iterated based on the predicted coupling efficiency values, gradually approaching the optimal solution. Finally, the parameter with the largest predicted coupling efficiency value is selected as the optimized geometric parameter, ensuring that the resulting micro / nano structure has high coupling efficiency for unpolarized diffuse light, thereby improving the collection efficiency of Raman scattered light.

[0072] In this embodiment, optionally, the pre-trained neural network model is trained in the following manner: Training samples are obtained, wherein the training samples include multiple geometric parameter samples, each geometric parameter sample corresponding to a coupling efficiency label, the coupling efficiency label being calculated based on the geometric parameter samples through electromagnetic simulation; for each geometric parameter sample, geometric filling is performed on a discrete two-dimensional grid according to the geometric parameters of the micro / nano structure defined by the geometric parameter sample, generating a binary image characterizing the cross-sectional shape of the micro / nano structure, wherein the pixels of the binary image are used to distinguish between the micro / nano structure entity region and the background medium region; using the binary image as input and the corresponding coupling efficiency label as the target output, supervised training is performed on the initial neural network model, and when the training stopping condition is met, the pre-trained neural network model is obtained.

[0073] In this embodiment, during the construction of the pre-trained neural network model, training samples are first acquired. The quality and quantity of training samples directly affect the effectiveness of subsequent model training. Specifically, the key features of the micro / nano structure covered by each geometric parameter sample can be determined, such as period, height, duty cycle, and tilt angle. To obtain sufficiently diverse geometric parameter samples, multiple combinations can be selected within the possible value range of each parameter through random generation, uniform sampling, or other methods to form multiple geometric parameter samples.

[0074] For each geometric parameter sample, its corresponding coupling efficiency label can be determined through electromagnetic simulation. Electromagnetic simulation is a technique that uses numerical methods to simulate the distribution and propagation of electromagnetic fields, enabling precise calculation of the coupling efficiency of micro / nano structures to unpolarized diffuse light. During the simulation, the micro / nano structure defined by the geometric parameter sample is input into the electromagnetic simulation software, and the corresponding boundary conditions, light source parameters, etc., are set before running the simulation. The simulation software can calculate the coupling efficiency of the micro / nano structure to unpolarized diffuse light based on electromagnetic field theory, and use this value as the coupling efficiency label for that geometric parameter sample. In this way, each geometric parameter sample is assigned an accurate coupling efficiency label, providing a data foundation for supervised learning in subsequent model training.

[0075] After obtaining the geometric parameter samples and their corresponding coupling efficiency labels, the geometric parameter samples can be transformed into an input format suitable for processing by neural network models. Since neural network models typically process structured data such as images, the micro / nano structures defined by the geometric parameter samples can be geometrically filled onto a discrete two-dimensional grid to generate a binary image.

[0076] Specifically, binary images can be generated as follows: First, a suitable two-dimensional grid size and resolution are determined, which can be based on the size of the micro / nano structure and the required precision. Then, geometric filling is performed on the two-dimensional grid according to the geometric parameters of the micro / nano structure defined in the geometric parameter sample, such as period, height, and duty cycle. During the filling process, the pixel value corresponding to the solid region of the micro / nano structure is set to a specific value (e.g., 1), indicating that the region is the solid part of the micro / nano structure; the pixel value corresponding to the background medium region is set to another value (e.g., 0), indicating that the region is the background medium. In this way, a binary image that can accurately characterize the cross-sectional shape of the micro / nano structure can be generated. This binary image intuitively displays the geometric features of the micro / nano structure, facilitating feature extraction and learning by neural network models.

[0077] Furthermore, the binary image can also be generated as follows: based on the period parameter in the geometric parameter sample, the period range of the two-dimensional grid in the horizontal direction is determined; based on the duty cycle parameter in the geometric parameter sample, the horizontal width of the micro / nano structure entity region is determined within the period range; based on the height parameter in the geometric parameter sample, the vertical height of the micro / nano structure entity region is determined; on the two-dimensional grid, the rectangular area determined by the horizontal width and vertical height is filled with a first pixel value to represent the micro / nano structure entity region; the remaining area on the two-dimensional grid is filled with a second pixel value to represent the background medium region; wherein, the first pixel value and the second pixel value are different.

[0078] After preparing the training samples (i.e., binary images and corresponding coupling efficiency labels), supervised training of the initial neural network model can begin. The initial neural network model can be a neural network with a specific structure and parameters, such as a convolutional neural network (CNN), which can automatically learn features from the input data and make predictions. The binary image is input into the initial neural network model, which processes and extracts features from the input image through its internal multiple neurons and layers. Then, the initial neural network model makes a prediction based on the learned features, outputting a coupling efficiency prediction value. This coupling efficiency prediction value is compared with the corresponding coupling efficiency label to calculate the prediction error. Based on the prediction error, optimization algorithms (such as gradient descent) are used to adjust the parameters of the initial neural network model to reduce the prediction error.

[0079] The training process is repeated, continuously inputting binary images into the newly adjusted neural network model to calculate the prediction error and update the model parameters. Training can stop when a preset number of training epochs is reached, or when the prediction error falls below a certain preset threshold. When the training stop condition is met, the neural network model has learned the mapping relationship from the binary image of the micro / nano structure to the coupling efficiency; the resulting model is the pre-trained neural network model. This model can be used in subsequent optimization of the geometric parameters of the micro / nano structure to quickly and accurately predict the coupling efficiency corresponding to different geometric parameters.

[0080] In this embodiment, firstly, coupling efficiency labels are obtained through electromagnetic simulation calculations, ensuring the accuracy and reliability of the labels and providing a high-quality data foundation for supervised training of the neural network model. Secondly, geometric parameter samples are converted into binary images as input, enabling the neural network model to intuitively learn the geometric features of micro / nano structures, avoiding the difficulties caused by directly processing complex geometric parameters. Thirdly, supervised training is employed, allowing the model to learn from the input binary images and corresponding coupling efficiency labels, gradually mastering the relationship between geometric features and coupling efficiency, thus improving the model's prediction accuracy. Finally, the pre-trained neural network model obtained when the training stopping condition is met has the ability to quickly and accurately predict the coupling efficiency corresponding to different geometric parameters, providing efficient support for subsequent optimization of micro / nano structure geometric parameters and contributing to improving the efficiency and quality of the entire microstructure-enhanced fiber fabrication method.

[0081] In this embodiment, optionally, the neural network model includes multiple feature extraction modules, a multilayer perceptron, a forward propagation module, and an output module; the step of "inputting the micro / nano structure cross-sectional image corresponding to each candidate geometric parameter into the pre-trained neural network model and outputting the predicted coupling efficiency value for non-polarized diffuse light corresponding to each candidate geometric parameter" includes: for each candidate geometric parameter corresponding to a micro / nano structure cross-sectional image, inputting the micro / nano structure cross-sectional image into multiple sequentially connected feature extraction modules of the neural network model, and extracting multi-scale high-dimensional features related to optical coupling performance from the micro / nano structure cross-sectional image level by level through the feature extraction modules. Figure 1 shows the process of inputting the multi-scale high-dimensional feature map output by the last feature extraction module into the multilayer perceptron of the neural network model. The multilayer perceptron performs nonlinear transformation and feature recombination on the multi-scale high-dimensional feature map to obtain a recombined feature vector. The recombined feature vector is then input into the forward propagation module of the neural network model. The forward propagation module optimizes the recombined feature vector to obtain a target feature vector. The target feature vector is then input into the output module of the neural network model. The output module converts the target feature vector into a predicted coupling efficiency value for non-polarized diffuse light corresponding to the candidate geometric parameters.

[0082] In this embodiment, such as Figure 6The diagram illustrates a neural network model provided in an embodiment of this application. When the neural network model processes micro / nano structure cross-sectional images corresponding to candidate geometric parameters, the micro / nano structure cross-sectional images are first input into multiple feature extraction modules connected in series. Each feature extraction module is a key component of the entire neural network model for feature extraction. For the first feature extraction module, the micro / nano structure cross-sectional image is used as its input feature map. It can extract shallow features related to optical coupling performance from the input micro / nano structure cross-sectional image to obtain an output feature map. For each subsequent feature extraction module, it can receive the output feature map from the previous adjacent feature extraction module, use it as its input feature map, and extract specific-level features related to optical coupling performance from this input feature map to obtain an output feature map.

[0083] In one specific embodiment, each feature extraction module includes a convolutional layer, a Gaussian error linear unit activation function, and a pooling layer. The convolutional layer extracts feature vectors related to optical coupling performance at a specific level from the input feature map. The Gaussian error linear unit activation function performs a nonlinear transformation on the feature vectors. The pooling layer performs dimensionality reduction processing on the nonlinear transformation result to obtain an output feature map. The multi-scale high-dimensional feature map output by the last feature extraction module is input into the multilayer perceptron of the neural network model. The multilayer perceptron performs nonlinear transformation and feature recombination on the multi-scale high-dimensional feature map to obtain a recombined feature vector.

[0084] Specifically, each feature extraction module may contain convolutional layers, Gaussian Error Linear Unit (GELU) activation functions, and pooling layers. The convolutional layer is the core of feature extraction; it uses a series of learnable convolutional kernels to slide across the input feature map, extracting feature vectors related to optical coupling performance at specific levels. These kernels can capture local features in cross-sectional images of micro / nano structures, such as edges and textures, which are crucial for understanding the optical coupling performance of micro / nano structures. Next, the GELU activation function performs a nonlinear transformation on the feature vectors extracted by the convolutional layer. Nonlinear transformation is key to enabling neural networks to learn complex functional relationships. The GELU activation function, through its specific mathematical form, introduces nonlinearity, allowing the neural network to better fit the complex relationship between input data and output targets, thereby enhancing the model's expressive power. Finally, the pooling layer performs dimensionality reduction on the nonlinear transformation results. Dimensionality reduction reduces the size and computational cost of the nonlinear transformation results while preserving important feature information. Pooling layers typically employ max pooling or average pooling, taking the maximum or average value of feature values ​​within local regions to obtain the output feature map. Through the step-by-step processing of multiple feature extraction modules, multi-scale high-dimensional features in the cross-sectional images of micro and nano structures are gradually extracted, providing rich feature information for subsequent processing.

[0085] After processing by multiple feature extraction modules, the multi-scale high-dimensional feature map output by the last feature extraction module can be input into the multilayer perceptron of the neural network model. The multilayer perceptron can include sequentially cascaded Linear + Gelu + Linear modules, enabling deeper nonlinear transformations and feature recombination of the input multi-scale high-dimensional feature map. This feature recombination further optimizes the feature representation, allowing the recombined feature vector to better reflect the optical coupling performance of the micro / nano structure. Through the processing of the multilayer perceptron, the input multi-scale high-dimensional feature map is transformed into a recombined feature vector, providing a more suitable feature representation for subsequent outputs.

[0086] The reconstructed feature vector is then fed into the forward propagation module of the neural network model. The main function of the forward propagation module is to optimize the reconstructed feature vector to obtain the target feature vector. Specifically, the forward propagation module can include specific optimization algorithms or structures, such as attention mechanisms. Attention mechanisms can dynamically adjust the weights of features based on the importance of different parts of the reconstructed feature vector, allowing the model to focus more on features closely related to optical coupling performance. By optimizing the reconstructed feature vector, the forward propagation module can further improve the quality and effectiveness of the features, removing redundant or irrelevant feature information, making the target feature vector more accurately reflect the optical coupling performance of the micro / nano structure.

[0087] Finally, the target feature vector is input into the output module of the neural network model. The main task of the output module is to convert the target feature vector into a predicted coupling efficiency value for unpolarized diffuse light corresponding to the candidate geometric parameters. The output module can use one or more fully connected layers to map the target feature vector to a scalar value, which is the predicted coupling efficiency value. During the mapping process, the weights and bias parameters of the fully connected layers are trained to accurately establish a linear or nonlinear relationship between the target feature vector and the coupling efficiency. In this way, the neural network model can quickly and accurately predict the corresponding coupling efficiency value based on the input micro / nano structure cross-sectional image, providing an important reference for subsequent geometric parameter optimization.

[0088] In this embodiment, the cascaded use of multiple feature extraction modules enables the step-by-step extraction of multi-scale, high-dimensional features from the cross-sectional images of micro / nano structures, comprehensively capturing information related to optical coupling performance and improving the feature extraction capability of the neural network model for complex micro / nano structures. The introduction of a multilayer perceptron further performs nonlinear transformations and recombination of features, uncovering deeper feature relationships and enhancing the model's expressive power. Optimized processing by the forward propagation module improves the quality and effectiveness of features, removes redundant information, and makes the target feature vector more accurately reflect optical coupling performance. The output module converts the target feature vector into a predicted coupling efficiency value, achieving efficient conversion from image input to numerical output. This structurally designed neural network model can quickly and accurately predict the coupling efficiency corresponding to different geometric parameters, providing strong support for optimizing the geometric parameters of micro / nano structures and contributing to improving the efficiency and quality of the entire microstructure-enhanced fiber fabrication method.

[0089] In this embodiment of the application, optionally, the fitness function is the mean square error between the predicted coupling efficiency and the preset coupling efficiency.

[0090] In this embodiment, the fitness function can be specifically set as the mean square error between the predicted coupling efficiency and the preset coupling efficiency. The preset coupling efficiency can be the desired target value, representing the coupling capability of the micro / nanostructure to unpolarized diffuse light under ideal conditions. The mean square error measures the deviation between the predicted coupling efficiency and the preset coupling efficiency; the smaller the deviation, the closer the performance of the micro / nanostructure corresponding to the candidate geometric parameters is to the ideal state. By using the mean square error as the fitness function, the genetic algorithm can tend to select candidate geometric parameters that minimize the mean square error during the optimization process, thereby gradually approaching the optimal solution and finding candidate geometric parameters that enable the micro / nanostructure to achieve the expected coupling efficiency.

[0091] In a specific embodiment of the application of mineral composition detection:

[0092] Cerium oxide (CeO2) powder was pressed into Φ10mm×1mm sheet-shaped samples for testing and placed in a container as shown in the image. Figure 2 The high-throughput sample disk 4 of the system shown.

[0093] Result: As Figure 7 As shown in the figure, the CeO2 Raman spectra collected using the microstructure-enhanced fiber of this application embodiment and using a bare fiber of the same type with a flat end face are compared. It can be clearly seen from the figure that the system using this application embodiment achieves a significant improvement in signal intensity across the entire 825-875 nm band compared to the bare fiber. This demonstrates that the technical solution of this application embodiment can effectively overcome the challenge of weak signal detection and achieve excellent beneficial effects.

[0094] This application also provides a computer device, which may specifically be a personal computer, a server, a network device, etc. Figure 8 As shown, the computer device includes a bus, a processor, memory, and a communication interface, and may also include an input / output interface and a display device. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database stores location information. The network interface allows communication with external terminals via a network connection. When the computer program is executed by the processor, it implements the steps in the various method embodiments.

[0095] Those skilled in the art will understand that Figure 8 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0096] In one embodiment, a computer-readable storage medium is provided, which may be non-volatile or volatile, having stored thereon a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0097] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0098] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0099] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0100] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0101] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A Raman spectrometer light collection system based on microstructured fiber, characterized in that, The method comprises the following steps: A light source assembly is used to generate laser light and collimate the laser light to irradiate a sample to be detected; A detection area assembly is used to place the sample to be detected and position the sample to be detected; A collection assembly is used to collect Raman scattered light generated by the sample to be detected under laser irradiation, wherein the collection assembly comprises a microstructure enhanced optical fiber, an end face of the microstructure enhanced optical fiber is provided with a micro-nano structure, and geometric parameters of the micro-nano structure are determined based on the following manner: a plurality of candidate geometric parameters are initialized, and a corresponding micro-nano structure cross-sectional view is generated based on each candidate geometric parameter respectively; each candidate geometric parameter corresponding micro-nano structure cross-sectional view is input into a pre-trained neural network model, and a coupling efficiency prediction value of each candidate geometric parameter corresponding to the non-polarized diffuse light is output; Based on the coupling efficiency prediction value corresponding to each candidate geometric parameter, the parameter values of the plurality of candidate geometric parameters are updated to obtain new candidate geometric parameters, and the coupling efficiency prediction value corresponding to each new candidate geometric parameter is recalculated until the iteration termination condition is met, and the candidate geometric parameter with the maximum coupling efficiency prediction value is taken as the optimized geometric parameter of the corresponding micro-nano structure; A control analysis assembly is used to collect signals from the collected Raman scattered light, and based on the signal collection result, a preset component recognition model is used to recognize component information corresponding to the sample to be detected.

2. The system of claim 1, wherein, The microstructure enhanced optical fiber comprises an optical fiber body and a micro-nano structure arranged on an end face of the optical fiber body; The micro-nano structure comprises a three-dimensional structure, the three-dimensional structure is a non-periodic structure or a composite periodic structure, and the geometric parameters of the micro-nano structure are determined by a genetic algorithm optimization, and a fitness function of the genetic algorithm is configured to maximize the coupling efficiency of the non-polarized diffuse light in a preset wavelength range.

3. The system of claim 2, wherein, The micro-nano structure is a grating, a nano-pore array or a plasmonic resonance structure; When the micro-nano structure is a grating, the grating is processed on the end face of the optical fiber body based on a micro-nano processing method, and a cross-sectional type of the grating comprises a triangle, a circle, a square, a trapezoid and a combination thereof; Correspondingly, the micro-nano processing method of the grating comprises electron beam lithography, focused ion beam etching and two-photon polymerization 3D printing; In the case that the micro-nano processing method of the grating is two-photon polymerization 3D printing, the two-photon laser parameters are as follows: wavelength is 700-1000 nm; laser power is 10-50 mW; and laser scanning speed is 1000-25000 μm / s.

4. The system of claim 2 or 3, wherein, The size of the micro-nano structure is adapted to the optical fiber body, and the constituent material of the micro-nano structure is metal, dielectric, polymer material or a combination thereof. The optical fiber body comprises at least one of a single-mode optical fiber, a multi-mode optical fiber, a multi-core optical fiber, a microstructure optical fiber and an optical waveguide on a chip.

5. A method of making a microstructured fiber, comprising: The method comprises the following steps: A target waveband of Raman scattered light is determined, and the target waveband is disassembled according to a preset waveband length to obtain a plurality of sub-wavebands; Setting a fitness function of a genetic algorithm, wherein the fitness function is used to evaluate the coupling efficiency of the micro-nano structure to the non-polar diffuse light in a preset wavelength range, and the coupling efficiency is proportional to the collection efficiency of the Raman scattered light; For each sub-band, a genetic algorithm is used to optimize the geometric parameters of the micro-nano structure corresponding to the sub-band with the fitness function as the optimization target, and a set of optimized geometric parameters of the micro-nano structure corresponding to the sub-band is output, wherein the geometric parameters include at least one of the period, height, duty cycle, and inclination angle of the micro-nano structure; Based on the optimized geometric parameters of the micro-nano structure corresponding to each sub-band, a micro-nano structure array is prepared on the end face of the optical fiber body by a micro-nano processing method to obtain a micro-structure enhanced optical fiber capable of enhancing the Raman scattered light collection efficiency of the target wavelength band; The genetic algorithm is used to optimize the geometric parameters of the micro-nano structure corresponding to the sub-band with the fitness function as the optimization target, and a set of optimized geometric parameters of the micro-nano structure corresponding to the sub-band is output, including: Initializing a plurality of candidate geometric parameters, and generating a corresponding micro-nano structure cross-sectional view based on each candidate geometric parameter; Each candidate geometric parameter corresponding to the micro-nano structure cross-sectional view is input into a pre-trained neural network model to output a coupling efficiency prediction value of each candidate geometric parameter corresponding to the non-polar diffuse light; Based on the coupling efficiency prediction value corresponding to each candidate geometric parameter, the parameter values of the plurality of candidate geometric parameters are updated to obtain new candidate geometric parameters, and the coupling efficiency prediction value corresponding to each new candidate geometric parameter is recalculated until the iteration termination condition is met, and the candidate geometric parameter with the maximum coupling efficiency prediction value is taken as the optimized geometric parameter of the micro-nano structure corresponding to the sub-band.

6. The method of claim 5, wherein, The pre-trained neural network model is trained based on the following method: Obtain training samples, wherein the training samples include a plurality of geometric parameter samples, and each geometric parameter sample corresponds to a coupling efficiency label, which is calculated based on the geometric parameter sample by electromagnetic simulation; For each geometric parameter sample, perform geometric filling on a discrete two-dimensional grid according to the geometric parameters of the micro-nano structure defined by the geometric parameter sample to generate a binary image representing the cross-sectional shape of the micro-nano structure, wherein the pixels of the binary image are used to distinguish the micro-nano structure solid region and the background medium region; The binary image is used as input and the corresponding coupling efficiency label is used as target output to supervise the training of an initial neural network model, and when the training stopping condition is met, the pre-trained neural network model is obtained.

7. The method according to claim 5 or 6, characterized in that, The neural network model includes a plurality of feature extraction modules, a multi-layer perceptron, a forward propagation module, and an output module; and the micro-nano structure cross-sectional view corresponding to each candidate geometric parameter is input into the pre-trained neural network model to output the coupling efficiency prediction value of each candidate geometric parameter corresponding to the non-polar diffuse light, including: For the micro-nano structure cross-section drawing corresponding to each candidate geometric parameter, the micro-nano structure cross-section drawing is input into a plurality of feature extraction modules connected in sequence in the neural network model, and a plurality of multi-scale high-dimensional feature maps related to the optical coupling performance are extracted from the micro-nano structure cross-section drawing by the feature extraction modules; The multi-scale high-dimensional feature map output by the last feature extraction module is input into a multi-layer perceptron of the neural network model, and the multi-scale high-dimensional feature map is subjected to nonlinear transformation and feature recombination by the multi-layer perceptron to obtain a recombined feature vector; The recombined feature vector is input into a forward propagation module of the neural network model, and the recombined feature vector is subjected to optimization processing by the forward propagation module to obtain a target feature vector; The target feature vector is input into an output module of the neural network model, and the target feature vector is converted into a predicted value of the coupling efficiency of the candidate geometric parameter for unpolarized diffuse light by the output module.

8. A storage medium having stored thereon a computer program, characterized in that The computer program is executed by the processor to implement the method of any one of claims 5-7.

9. A computer device comprising a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, characterized in that, The processor executes the computer program to implement the method of any one of claims 5-7.

Images