A kind of hollow optical fiber microstructure size dynamic nondestructive measurement and control system and method
By combining spectral confocal detection and neural network models, the internal microstructure dimensions of hollow optical fibers can be measured in real time, solving the measurement difficulties in existing technologies and achieving efficient and precise production control.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- YANGTZE OPTICAL FIBRE & CABLE CO LTD
- Filing Date
- 2026-03-16
- Publication Date
- 2026-06-12
AI Technical Summary
Existing technologies cannot measure the internal microstructure dimensions in real time and without damage during the hollow fiber drawing process, resulting in low production efficiency, resource waste, and unstable product quality.
A spectral confocal detection module combined with a neural network model is used to obtain spectral signals from the side of the hollow fiber through the principle of dispersive confocalization. The diameter of the nested glass tube is calculated using the neural network model, and the microstructure size is adjusted in real time through an airflow adjustment module.
It enables precise measurement of the nested structural units inside hollow optical fibers, improving production efficiency and product yield, reducing material and labor costs, and enhancing measurement accuracy.
Smart Images
Figure CN121850352B_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of hollow optical fiber manufacturing, and specifically relates to a dynamic non-destructive measurement and control system and method for the microstructure dimensions of hollow optical fibers. Background Technology
[0002] The light-guiding performance of hollow-core optical fibers is highly sensitive to the geometric dimensions of their internal microstructure, requiring monitoring during the fiber drawing process and timely adjustment of process parameters. Existing online fiber drawing monitoring methods can only measure external dimensions such as the fiber cladding and coating in real time. To view the internal structure, the fiber must be cut during continuous drawing, and the end face must be manually measured using a microscope. Process parameters are then adjusted based on the measurement results. This process is time-consuming, has slow feedback, seriously affects production efficiency, wastes production resources, and may even affect product quality.
[0003] Existing techniques propose a method to analyze the microstructure dimensional parameters of hollow optical fibers by illuminating them with a light source and collecting the scattered spectrum at the other end. This method requires precise adjustment of the angles between the light source, the fiber under test, and the receiver. Due to measurement errors in these angles, the final calculated dimensions also contain some error.
[0004] Spectral confocal sensors can achieve high-precision measurements of transparent materials. However, when directly measuring hollow fiber microstructures using traditional spectral confocal methods, it may be difficult to distinguish the reflection peaks between different microrings and the inner and outer surface peaks of the same microring. Furthermore, since the hollow fiber microstructure consists of nested microrings, additional FP microcavities are formed, resulting in periodic cavity mode superposition in the reflection spectrum, which may lead to misjudgments by traditional algorithms. Summary of the Invention
[0005] In view of the above-mentioned defects or improvement needs of the existing technology, the present invention proposes a dynamic non-destructive measurement and control system and method for the microstructure size of hollow optical fiber.
[0006] To achieve the above objectives, according to one aspect of the present invention, a dynamic non-destructive measurement and control system for the microstructure dimensions of hollow-core optical fibers is provided. The hollow-core optical fiber includes an outer cladding and an inner cladding. The inner cladding is composed of nested structural units, which are arranged circumferentially along the inner wall of the outer cladding and connected to the inner wall of the outer cladding. The central cavity covered by the inner cladding forms the fiber core. Each nested structural unit includes at least one nested glass tube. The system includes:
[0007] The spectral confocal detection module is used to acquire the spectral signals reflected from each interface of the hollow fiber cross-section from the side of the hollow fiber during the hollow fiber drawing process, using the principle of dispersive confocal, and convert them into spectral data; the interface is the reflection interface formed when the incident light emitted by the spectral confocal detection module passes through the outer cladding and inner cladding.
[0008] The data analysis module is used to calculate the axial peak value based on spectral data using a neural network model, and convert it into the diameter of the nested glass tube in the nested structural unit;
[0009] The control module is used to output airflow adjustment commands based on the difference between the diameter of the nested glass tube and the preset value;
[0010] The airflow control module is used to adjust the amount of airflow entering the hollow optical fiber according to the airflow adjustment command, thereby adjusting the diameter of the nested glass tube.
[0011] According to the above scheme, on the cross-section of a hollow optical fiber, the centers of all nested glass tubes in the same nested structural unit are on the same straight line, and are also on the same straight line as the center of the outer cladding.
[0012] The incident light passes through the center of the outer cladding.
[0013] According to the above scheme, the system also includes a rotation follower module, which drives the spectral confocal detection module to rotate around the hollow fiber as an axis, so that each nested structural unit in the hollow fiber can be measured.
[0014] According to the above scheme, the system also includes a position detection module, which moves together with the spectral confocal detection module to detect the measurement position of the hollow fiber in real time.
[0015] The control module is also used to generate position change commands based on the measured position of the detected hollow fiber to control the rotation of the rotation follower module.
[0016] According to the above scheme, there are several spectral confocal detection modules, which are staggered along the circumference of the hollow fiber. The incident light emitted by each spectral confocal detection module is located at different cross sections of the hollow fiber.
[0017] According to the above scheme, the output mapping of the neural network model is as follows:
[0018]
[0019] in, Representative with A neural network mapping operator with parameters, parameter set It consists of the convolutional kernel weights and biases of a deep learning network; This represents the interface reflectance spectrum sequence acquired by the spectral confocal detection module; The continuous response function representing the signal received by the spectral confocal detection module; This represents the activation function.
[0020] According to the above plan, It is constructed by simulating the nonlinear transformation capability of a neural network through a deep learning network, and is expressed as:
[0021]
[0022] Among them, the independent variable Wavelength; Summation term In The number of optical interfaces corresponding to the hollow-core optical fiber under test; coefficient Indicates the first The Fresnel reflection intensity of each interface is encoded by the channel amplitude output by the network; Let be the axial position of the k-th interface. Representing the The standard deviation of each peak is related to the axial chromatic aberration depth of focus of the spectral confocal detection module and is automatically learned by the network; in the linear term... The linear shift representing the response curve, The DC component represents dark current noise.
[0023] According to the above scheme, the neural network model is trained by introducing physical drive until the variance between the calculated value of the axial peak and the true value of the axial peak is minimized.
[0024] The true value of the axial peak was obtained by scanning electron microscopy through the end face.
[0025] According to the above scheme, during training, the microstructure physical model is embedded into gradient descent using the Lagrange multiplier method to construct a composite loss; the composite loss includes:
[0026] The data supervision term measures the deviation between the predicted peak position and the actual value; and
[0027] Physical constraints include wall thickness constraints for each nested glass tube layer and monotonicity constraints for axial position.
[0028] According to another aspect of the present invention, a measurement and control method is provided using the aforementioned hollow-core optical fiber microstructure dynamic non-destructive measurement and control system, comprising the following steps:
[0029] During the hollow fiber drawing process, the principle of dispersive confocal imaging is used to obtain the spectral signals reflected from each interface of the hollow fiber cross-section from the side of the hollow fiber and convert them into spectral data; the interface is the reflection interface formed when the incident light emitted by the spectral confocal detection module passes through the outer cladding and the inner cladding.
[0030] Based on the spectral data, the axial peak value is calculated using a neural network model and converted into the diameter of the nested glass tube in the nested structural unit;
[0031] Based on the difference between the diameter of the nested glass tube and the preset value, an airflow adjustment command is output to adjust the airflow entering the hollow optical fiber, thereby adjusting the diameter of the nested glass tube.
[0032] In summary, compared with the prior art, the above-described technical solutions conceived by this invention can achieve the following beneficial effects:
[0033] 1. By combining spectral confocal detection with neural network models, the diameter of the nested structural units inside hollow optical fibers can be accurately measured. This measurement can be performed during the production process, allowing for timely adjustment of process parameters based on product status, saving material waste and labor costs, and improving production efficiency and product yield.
[0034] 2. By introducing physical constraints into the loss function of the neural network model, the model training is carried out under a dual data-physical drive, making spectral confocal detection more suitable for measuring the nested structural units inside hollow optical fibers and improving measurement accuracy.
[0035] 3. The spectral confocal detection module is rotated circumferentially by a rotation follower module, or multiple spectral confocal detection modules are staggered circumferentially, so that each nested structural unit in the hollow fiber can be measured. Attached Figure Description
[0036] Figure 1 This is a system structure diagram provided in an embodiment of the present invention.
[0037] Figure 2 This is a schematic diagram of a hollow fiber structure to be tested provided in an embodiment of the present invention.
[0038] Figure 3 This is a dispersive confocal optical path diagram provided in an embodiment of the present invention.
[0039] Figure 4 This is a schematic diagram of another hollow fiber structure to be tested provided in the embodiments of the invention.
[0040] In the picture:
[0041] 1-Outer cladding layer, 2-Nested structural unit, 201-First nested glass tube, 202-Second nested glass tube, 203-Third nested glass tube, 211-First support structure, 212-Second support structure, 3-Fiber core. Detailed Implementation
[0042] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other.
[0043] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "clockwise," "counterclockwise," "axial," "radial," and "circumferential" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.
[0044] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.
[0045] In this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," "linking," and "fixing," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components, unless otherwise explicitly limited. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.
[0046] In this invention, unless otherwise explicitly specified and limited, "above" or "below" the second feature can mean that the first feature is in direct contact with the second feature, or that the first feature is in indirect contact with the second feature through an intermediate medium. Furthermore, "above," "over," and "on top" of the second feature can mean that the first feature is directly above or diagonally above the second feature, or simply that the first feature is at a higher horizontal level than the second feature. "Below," "below," and "under" the second feature can mean that the first feature is directly below or diagonally below the second feature, or simply that the first feature is at a lower horizontal level than the second feature.
[0047] The present invention aims to achieve precise measurement of the nested structural units inside the hollow fiber during the drawing process of hollow fiber.
[0048] The hollow optical fiber targeted by this invention includes an outer cladding and an inner cladding. The inner cladding is composed of nested structural units, which are arranged circumferentially along the inner wall of the outer cladding and connected to the inner wall of the outer cladding. The central cavity covered by the inner cladding forms the fiber core. Each nested structural unit includes at least one nested glass tube.
[0049] In some embodiments, the hollow fiber structure is as follows: Figure 2 As shown, the structure includes an outer cladding layer 1 and an inner cladding layer. The inner cladding layer is composed of nested structural units 2, which are arranged circumferentially along the inner wall of the outer cladding layer 1 and connected to the inner wall of the outer cladding layer. The central cavity covered by the inner cladding layer forms a fiber core 3. Each nested structural unit 2 includes two nested glass tubes, namely a first nested glass tube 201 and a second nested glass tube 202. In other embodiments, it may also include only one nested glass tube, or three or more nested glass tubes, all of which are applicable.
[0050] To achieve the above objectives, this embodiment provides a dynamic non-destructive measurement and control system for the dimensions of hollow-core optical fiber microstructures, such as... Figure 1 As shown, the system includes a spectral confocal detection module, a data analysis module, a control module, and an airflow control module.
[0051] The spectral confocal detection module is used during the hollow fiber drawing process to acquire the spectral signals reflected from each interface of the hollow fiber's cross-section using the dispersive confocal principle, and converts them into spectral data. The interface is the reflection interface formed when the incident light emitted by the spectral confocal detection module passes through the outer and inner cladding layers. For example... Figure 3 As shown, since the inner surface of the outer cladding layer 1, the second nested glass tube 202, and the first nested glass tube 201 are bonded together, the reflection interface formed by the incident light emitted by the spectral confocal detection module passing through the hollow optical fiber includes: the outer surface of the outer cladding layer 1, the proximal inner surface of the first nested glass tube 201, the distal inner surface of the first nested glass tube 201, the distal outer surface of the first nested glass tube 201, the distal inner surface of the second nested glass tube 202, and the distal outer surface of the second nested glass tube 202.
[0052] The data analysis module uses a neural network model to calculate the axial peak value based on spectral data and converts it into the diameter of the nested glass tubes in the nested structural unit. The control module outputs airflow adjustment commands based on the difference between the diameter of the nested glass tubes and a preset value.
[0053] The airflow control module is used to adjust the amount of airflow entering the hollow optical fiber according to the airflow adjustment command, thereby adjusting the diameter of the nested glass tube.
[0054] In some embodiments, the spectral confocal detection module includes a spectral confocal sensor probe and a high-resolution spectrometer sensor. The wavelength range of the light beam emitted by the spectral confocal sensor probe is 450nm-900nm, and the spectral signal reflected from each interface is obtained using the dispersive confocal principle. The spectral confocal sensor probe is positioned perpendicular to the running direction of the hollow fiber under test, without contacting the hollow fiber. The device is placed in an area without an external coating layer to measure the hollow fiber under test; at this time, all structural materials are glass.
[0055] In some embodiments, continue as follows Figure 1 As shown, the system also includes a rotation follower module, which drives the spectral confocal detection module to rotate around the hollow fiber as an axis, so that each nested structural unit in the hollow fiber can be measured.
[0056] In some embodiments, continue as follows Figure 1 As shown, the system also includes a position detection module, which moves along with the spectral confocal detection module to detect the measurement position of the hollow fiber in real time. At this time, the control module is also used to generate position change commands based on the detected measurement position of the hollow fiber to control the rotation of the servo module.
[0057] As an alternative, in some other embodiments, there are several spectral confocal detection modules, which are staggered along the circumference of the hollow fiber, and the incident light emitted by each spectral confocal detection module is located at different cross sections of the hollow fiber.
[0058] In some embodiments, continue as follows Figure 1 As shown, the system also includes a human-computer interaction module, which integrates the above-mentioned large ring diameter measured in real time via serial port and software, and displays historical time values. It allows users to choose between managed or manual intervention.
[0059] The spectral confocal detection module is fixed on the rotating follower module. Before measurement, the rotation center of the spectral confocal detection module needs to be calibrated to be located at the hollow fiber under test by controlling the adjustment module through the human-machine interface module. During the measurement, based on the position coordinates updated by the position detection module, the adjustment module issues commands to cause the rotating follower module to adjust its position.
[0060] The invention will be further illustrated below with specific examples.
[0061] In some embodiments, such as Figure 2 and Figure 3As shown, in the cross-section of a hollow optical fiber, the centers of all nested glass tubes located in the same nested structural unit 2 are on the same straight line, and also on the same straight line as the center of the outer cladding layer 1; the incident light of the spectral confocal detection module passes through the center of the outer cladding layer 1. This ensures that the formed reflection spectrum is along the diameter direction of all nested glass tubes, and can be directly converted into diameter during conversion, making the algorithm simpler.
[0062] In some embodiments, the data analysis module receives spectral data generated by the spectral confocal detection module via serial communication, generates axial peak calculation values based on a neural network model, and converts them into the inner and outer diameter values of the nested glass tubes in the nested structural unit. The neural network model predicts the size of the nested glass tubes by relating the actual values to the small distortions of the overlapping peaks, and improves the prediction accuracy by introducing physical constraints.
[0063] The neural network model in this embodiment is specifically as follows:
[0064] First, according to Figure 3 Define the interface reflection spectrum sequence obtained by perpendicular incidence of incident light in the spectral confocal detection module. for:
[0065] ,
[0066] In the formula, for The normalized light intensity value at that point, where N is the total number of wavelengths. and These represent the minimum and maximum wavelengths of the incident light.
[0067] The response function, constructed by simulating the nonlinear transformation capability of deep learning networks, can be expressed as:
[0068]
[0069] in, The continuous response function representing the signal received by the spectral confocal detection module, whose independent variable is... For wavelength, parameter set It consists of the convolutional kernel weights and biases of the deep learning network. (Summation term) In For the number of interfaces corresponding to the material being tested, for a single nested structural unit inside a hollow optical fiber, These respectively represent the outer surface of the outer cladding layer 1, the proximal inner surface of the first nested glass tube 201, the distal inner surface of the first nested glass tube 201, the distal outer surface of the first nested glass tube 201, the distal inner surface of the second nested glass tube 202, and the distal outer surface of the second nested glass tube 202, as shown below. Figure 3 As shown. Coefficients Indicates the first The Fresnel reflection intensity of each interface is encoded by the channel amplitude output by the network; Let be the axial position of the k-th interface. Representing the The standard deviation of each peak is related to the axial chromatic aberration depth of focus of the spectral confocal sensor probe and is automatically learned by the network. In the linear term... The linear shift representing the spectrometer response curve, The linear term represents the DC component of dark current noise. By modeling baseline drift as a linear function of pixel position, wavelength-dimensional calibration is achieved in the digital domain, compensating for measurement biases introduced by the dispersive characteristics of the optical system. Therefore, the output mapping of the neural network model can be defined as:
[0070]
[0071] in Representative with A neural network mapping operator with parameters, containing all trainable weights and biases in the neural network model; The sequence of interface reflectance spectra acquired by the spectral confocal detection module is about The function is actually implemented as a one-dimensional floating-point array, where each element represents the normalized light intensity value at the corresponding wavelength. Represents the activation function, the final output of the network. This represents the optical axial position of each interface, which needs to be further converted into physical distance in micrometers, and the inner and outer diameters of each nested glass tube are calculated. When the sensor is located at the interval of each nested structural unit, it can only detect the spectral peaks reflected from the air-outer wall of the sleeve and the inner wall of the sleeve-air interface. To reduce processing time, the peak positions measured by the spectral confocal sensor are directly used for this simple structure. Therefore, the diameter conversion method can be expressed as:
[0072]
[0073] Corresponding Appendix Figure 2 , Let represent the outer diameter of the second nested glass tube 202. The calculation method for the outer diameter of the second nested glass tube 202 is to divide the optical thickness of each segment by the corresponding refractive index and then convert it into geometric thickness and sum it.
[0074] in This represents the axial position value of the kth reflecting interface predicted by the neural network. and These represent the axial positions of the inner and outer walls of the sleeve as measured by the confocal sensor, respectively. Represents the refractive index of glass; This represents the refractive index of air, with a value of approximately 1. Therefore, the optical thickness of the air layer is approximately equal to its geometric thickness.
[0075] This method introduces physics-driven training of the model, embedding the microstructure physical model into gradient descent using the Lagrange multiplier method to construct a composite loss. The loss function is a constraint during training. For the neural network model itself, the process can be considered as Input → Neural Network → Output. The network output alone is not necessarily the correct value. At this point, it needs to be given a learning object, and the difference between this learning object (Ground Truth) and the output (Output) needs to be measured and continuously reduced. This process is constrained by the loss function (usually called loss), and iterative calculation is performed until the loss decreases to a certain level.
[0076] Composite loss Specifically as follows:
[0077]
[0078] Among them, data supervision items The method used to measure the deviation between the predicted peak position and the actual value can be expressed as:
[0079]
[0080] In the formula This represents the Huber robust loss function, here Representing the true axial position, the value is measured through the end face using a scanning electron microscope. This value is then converted back to the true axial position using the method mentioned earlier for calculating the geometric position from the axial position. This predicted value is... Output from a neural network. Design the regularization term. The sensitivity of the peak position predicted by the constrained neural network to fluctuations in input light intensity is forced to learn low-frequency, physically stable features. This equation is the first one. The predicted peak is related to wavelength. The squared norm of the gradient, The weighting coefficient is set to 0.001 initially and gradually increased during training until the variance between the peak position predicted by the validation set and the true value is minimized.
[0081] Except for data supervision items In addition, the physical constraint term in the composite loss It can be represented as:
[0082]
[0083] This consists of four parts. The first part is the wall thickness constraint of the second nested glass tube 202, in which... The first term represents the actual wall thickness of the second nested glass tube 202; similarly, the second term is the wall thickness constraint for the first nested glass tube 201; the third term is the inter-ring cylindricity gap constraint, where... The refractive index of air, The fourth item is for the inter-ring air gap; it checks the increasing relationship of adjacent peaks to address the axial position monotonicity constraint. If the peak number is misjudged, a penalty is applied to ensure that the order is not reversed. .
[0084] The above is just one example. When the structure of the hollow fiber under test changes, the corresponding neural network model undergoes adaptive transformation and training. Increasing or decreasing the number of nested glass tube layers results in increasing or decreasing reflective interfaces, thus affecting the corresponding... The quantity needs to be adjusted adaptively. For example, nested glass tubes may not be complete circles, but rather arcs. In this case, the converted distance will not be the complete diameter and needs to be calculated based on the actual situation of the nested glass tubes.
[0085] In some embodiments, the structure of the hollow fiber under test is as follows: Figure 4 As shown, each nested structural unit 2 includes a first nested glass tube 201, a second nested glass tube 202, and a third nested glass tube 203. The third nested glass tube 203 is connected to the inner wall of the outer cladding layer 1. The second nested glass tube 202 and the third nested glass tube 203 are connected via a first support structure 211 and a second support structure 212. The inner wall of the first nested glass tube 201 is connected to the inner wall of the second nested glass tube 202. At this point, the connection point between the third nested glass tube 203 and the outer cladding layer, the connection point between the first nested glass tube 201 and the second nested glass tube 202, and the center of the outer cladding layer 1 are all on the same straight line. When the incident light emitted by the spectral confocal detection module passes through this straight line, the reflecting interfaces sequentially include the outer surface of the cladding 1, the proximal inner surface of the third nested glass tube 203, the proximal outer surface of the second nested glass tube 202, the proximal inner surface of the first nested glass tube 201, the distal inner surface of the first nested glass tube 201, the distal outer surface of the first nested glass tube 201, the distal inner surface of the second nested glass tube 202, the distal outer surface of the second nested glass tube 202, the distal inner surface of the third nested glass tube 203, and the distal outer surface of the third nested glass tube 203. During the modeling of this structure, the number of optical interfaces of the hollow fiber under test in the function needs to be adjusted, and adaptive adjustments need to be made when converting the diameter.
[0086] This embodiment also provides a measurement and control method using the aforementioned dynamic non-destructive measurement and control system for the microstructure dimensions of hollow optical fibers, including the following steps:
[0087] S1. During the hollow fiber drawing process, the principle of dispersive confocal optical fiber is used to obtain the spectral signal reflected by each interface of the hollow fiber cross section from the side of the hollow fiber and convert it into spectral data; the interface is the reflection interface formed when the incident light emitted by the spectral confocal detection module passes through the outer cladding and the inner cladding.
[0088] S2. Based on the spectral data, the axial peak value is calculated using a neural network model and converted into the diameter of the nested glass tube in the nested structural unit.
[0089] S3. Based on the difference between the diameter of the nested glass tube and the preset value, output an airflow adjustment command to adjust the airflow entering the hollow optical fiber, thereby adjusting the diameter of the nested glass tube.
[0090] By employing the system and method of this invention, dynamic, non-destructive, and precise measurement of the internal microstructure (i.e., nested structural units) during the hollow fiber drawing process is achieved, and the airflow can be controlled in real time to adjust the size of the nested structural units.
[0091] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.
[0092] It should be noted that, depending on the implementation needs, the various steps / components described in this application can be broken down into more steps / components, or two or more steps / components or parts of the operation of steps / components can be combined into new steps / components to achieve the purpose of this invention.
[0093] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.
Claims
1. A dynamic non-destructive measurement and control system for the microstructure dimensions of hollow optical fibers, wherein the hollow optical fiber includes an outer cladding and an inner cladding, the inner cladding being composed of nested structural units arranged circumferentially along the inner wall of the outer cladding and connected to the inner wall of the outer cladding, and the central cavity covered by the inner cladding forming the fiber core; each nested structural unit includes at least one layer of nested glass tubes; characterized in that: The system includes: The spectral confocal detection module is used to acquire the spectral signals reflected from each interface of the hollow fiber cross-section from the side of the hollow fiber during the hollow fiber drawing process, using the principle of dispersive confocal, and convert them into spectral data; the interface is the reflection interface formed when the incident light emitted by the spectral confocal detection module passes through the outer cladding and inner cladding. The data analysis module is used to calculate the axial peak value based on spectral data using a neural network model, and then convert it into the diameter of the nested glass tubes in the nested structural unit; the output mapping of the neural network model is as follows: in, Representative with A neural network mapping operator with parameters, parameter set It consists of the convolutional kernel weights and biases of a deep learning network; This represents the interface reflectance spectrum sequence acquired by the spectral confocal detection module; The continuous response function representing the signal received by the spectral confocal detection module; Represents the activation function; The control module is used to output airflow adjustment commands based on the difference between the diameter of the nested glass tube and the preset value; The airflow control module is used to adjust the amount of airflow entering the hollow optical fiber according to the airflow adjustment command, thereby adjusting the diameter of the nested glass tube.
2. The dynamic non-destructive measurement and control system for the microstructure dimensions of hollow optical fibers according to claim 1, characterized in that: In the cross-section of a hollow optical fiber, the centers of all the nested glass tubes in the same nested structural unit are on the same straight line, and are also on the same straight line as the center of the outer cladding. The incident light passes through the center of the outer cladding.
3. The dynamic non-destructive measurement and control system for the microstructure dimensions of hollow optical fibers according to claim 1 or 2, characterized in that: The system also includes a rotation follower module, which drives the spectral confocal detection module to rotate around the hollow fiber as an axis, so that each nested structural unit in the hollow fiber can be measured.
4. The dynamic non-destructive measurement and control system for the microstructure dimensions of hollow optical fibers according to claim 3, characterized in that: The system also includes a position detection module, which moves along with the spectral confocal detection module to detect the measurement position of the hollow fiber in real time. The control module is also used to generate position change commands based on the measured position of the detected hollow fiber to control the rotation of the rotation follower module.
5. The dynamic non-destructive measurement and control system for the microstructure dimensions of hollow optical fibers according to claim 1 or 2, characterized in that: There are several spectral confocal detection modules, which are staggered along the circumference of the hollow fiber. The incident light emitted by each spectral confocal detection module is located at a different cross section of the hollow fiber.
6. The dynamic non-destructive measurement and control system for the microstructure dimensions of hollow optical fibers according to claim 1, characterized in that: It is constructed by simulating the nonlinear transformation capability of a neural network through a deep learning network, and is expressed as: Among them, the independent variable Wavelength; Summation term In The number of interfaces corresponding to the hollow fiber under test; coefficient Indicates the first The Fresnel reflection intensity of each interface is encoded by the channel amplitude output by the network; For the first The axial position of each interface Representing the The standard deviation of each peak is related to the axial chromatic aberration depth of focus of the spectral confocal detection module and is automatically learned by the network; in the linear term... The linear shift representing the response curve, The DC component represents dark current noise.
7. The dynamic non-destructive measurement and control system for the microstructure dimensions of hollow optical fibers according to claim 1 or 2, characterized in that: The neural network model is trained by introducing physical drivers until the variance between the calculated axial peak value and the true axial peak value is minimized. The true value of the axial peak was obtained by scanning electron microscopy through the end face.
8. The dynamic non-destructive measurement and control system for the microstructure dimensions of hollow optical fibers according to claim 7, characterized in that: During training, the microstructure physical model is embedded into gradient descent using the Lagrange multiplier method to construct a composite loss; the composite loss includes: The data supervision term measures the deviation between the predicted peak position and the actual value; and Physical constraints include wall thickness constraints for each nested glass tube layer and monotonicity constraints for axial position.
9. A measurement and control method implemented using the dynamic non-destructive measurement and control system for the microstructure dimensions of hollow optical fibers according to any one of claims 1-8, characterized in that: Includes the following steps: During the hollow fiber drawing process, the principle of dispersive confocal imaging is used to obtain the spectral signals reflected from each interface of the hollow fiber cross-section from the side of the hollow fiber and convert them into spectral data; the interface is the reflection interface formed when the incident light emitted by the spectral confocal detection module passes through the outer cladding and the inner cladding. Based on the spectral data, the axial peak value is calculated using a neural network model and converted into the diameter of the nested glass tube in the nested structural unit; Based on the difference between the diameter of the nested glass tube and the preset value, an airflow adjustment command is output to adjust the airflow entering the hollow optical fiber, thereby adjusting the diameter of the nested glass tube.