Method for producing porous glass base material, method for producing transparent glass base material, and device for producing porous glass base material

Abstract

Provided is a method for producing a porous glass base material, in which glass microparticles generated by a burner flame formed by a burner are deposited at the tip of a starting material that rotates about a vertical axis, and the starting material is pulled upward, thereby forming a porous glass base material. The manufacturing method comprises the following steps that one or more cameras are used for shooting the deposition face of the glass microparticles in the starting material and the flame of the combustor, the image of the deposition face and the image of the flame of the combustor are projected to the same coordinate plane, and obtained image data are obtained; and calculating a characteristic quantity of a spatial relationship between the deposition surface and the burner flame by performing image processing on the image data.

Description

Technical Field

[0001] This invention relates to a method for manufacturing a porous glass substrate, a method for manufacturing a transparent glass substrate, and an apparatus for manufacturing a porous glass substrate. Background Technology

[0002] Patent Document 1 describes a process in which "during glass microparticle deposition, the deposition shape of the glass microparticle deposit 14 is monitored and the lifting speed of the glass microparticle deposit 14 is controlled. At the same time, at least one of the following is controlled: the flow rate of glass raw material gas, the flow rate of flame forming gas, and the position of the core burner 17 and the cladding burner 18 relative to the glass microparticle deposit 14, which are used as burners for generating glass microparticles, so as to achieve a target deposition shape" (Abstract). Patent Document 2 describes that "image data representing at least the state of a flame or particle flow is extracted from a two-dimensional image obtained by an imaging device... Specifically, after adjusting the brightness of the two-dimensional image from the signal processing unit 104, the outline of the flame or particle flow is determined. In addition, the calculation unit 105 performs regression prediction on the refractive index distribution of the transparent glass substrate 140 as the target variable based on explanatory variables that include at least the data obtained by coordinateizing the outline of the flame or particle flow. More specifically, the outline data of the flame or particle flow, the burner setting position (the burner setting position along the X-axis of the burner, the burner setting angle), the flow rate of the glass raw material gas (raw material containing a dopant for refractive index adjustment), the flow rate of the fuel gas (H2), the flow rate of the combustion-supporting gas (O2), the dehydration and sintering conditions (temperature, gas flow rate) when the glass microparticle deposit 14 is made transparent and vitrified are set as explanatory variables, and the data characterizing the refractive index distribution is set as the target variable." (Paragraph

[0032] ). Patent document 3 describes "detecting the edge shape of the deposition surface from an image of the deposition surface acquired by an imaging device, and quantifying the deformation of the edge shape to determine the quality of the glass microparticle deposit" (abstract). Patent document 4 describes "measuring the brightness of the porous glass substrate and its surroundings along the axial and / or radial directions from at least one direction, and detecting the measurement position where the maximum value of the brightness change rate in the axial and / or radial directions is obtained as the shape and position of the substrate" (paragraph

[0011] ). [Background Technical Documents] [Patent Literature] [Patent Document 1] Japanese Patent Application Publication No. 2012-91965 [Patent Document 2] WO2021 / 235408 [Patent Document 3] WO2022 / 186341 [Patent Document 4] Japanese Patent Application Publication No. 10-72225 Summary of the Invention

[0003] In a first aspect of the invention, a method for manufacturing a porous glass matrix is ​​provided, wherein a porous glass matrix is ​​formed by depositing glass microparticles generated by a burner flame formed by a burner onto the front end of a starting material rotating about a vertical axis, and then pulling the starting material upward. The method for manufacturing the porous glass matrix includes the following steps: photographing the deposition surface of the glass microparticles in the starting material and the burner flame using one or more cameras, projecting the images of the deposition surface and the burner flame onto the same coordinate plane to obtain image data; and calculating characteristic quantities of the spatial relationship between the deposition surface and the burner flame by performing image processing on the image data.

[0004] The manufacturing method of the porous glass matrix may further include the following steps: by performing image processing on the image data, calculating the characteristic quantities of the shape of the burner flame and the characteristic quantities of the shape of the deposition surface.

[0005] In any of the methods for manufacturing a porous glass substrate, the optical axis of the camera that captures the deposition surface in one or more cameras can have an angle of 45 degrees to 90 degrees relative to the central axis of the porous glass substrate. Specifically, it can have an angle of 60 degrees to 90 degrees, and more specifically, it can have an angle of 80 degrees to 90 degrees. In any of the methods for manufacturing a porous glass substrate, the optical axis of the camera that captures the burner flame in one or more cameras can have an angle of 45 degrees to 90 degrees relative to the central axis of the burner. Specifically, it can have an angle of 60 degrees to 90 degrees, and more specifically, it can have an angle of 80 degrees to 90 degrees.

[0006] In any of the methods for manufacturing porous glass matrix, the angle between the optical axis of the camera that photographs the deposition surface and the optical axis of the camera that photographs the burner flame can be 0 degrees or more and 20 degrees or 160 degrees or more and 180 degrees. More specifically, it can be 0 degrees or more and 10 degrees or 170 degrees or more and 180 degrees. More specifically, it can be 0 degrees or more and 5 degrees or 175 degrees or more and 180 degrees.

[0007] In any of the methods for manufacturing porous glass matrix, acquiring the image data may include the following steps: acquiring the image data by photographing the deposition surface and the burner flame with a camera, and capturing the deposition surface and the burner flame in the same frame.

[0008] In any of the methods for manufacturing porous glass matrix, acquiring the image data may include the following steps: taking pictures of the deposition surface and the burner flame using two cameras that are shooting from completely opposite directions; and acquiring the image data projected onto the same coordinate plane by inverting one of the images of the deposition surface and the burner flame.

[0009] In any of the manufacturing methods of the porous glass matrix, calculating the characteristic quantities of the shape of the deposition surface may include the following steps: detecting the boundary points between the deposition surface and the background by performing image processing on the image data; calculating at least one of the characteristic quantities of the front center position of the deposition surface, the characteristic quantity of the inclination of the deposition surface, the characteristic quantity of the outer diameter shape of the deposition surface, and the characteristic quantity of the deformation degree of the deposition surface based on the coordinate data of the detected boundary points; and including the at least one characteristic quantity in the characteristic quantities of the shape of the deposition surface.

[0010] In any of the methods for manufacturing porous glass substrates, calculating the characteristic quantities of the shape of the burner flame may include the following steps: detecting the boundary points between the burner flame and the background by performing image processing on the image data; calculating each of a plurality of measurement positions on the extension line of the central axis of the burner as the coordinates of the central position of the burner flame based on the coordinate data of the detected boundary points; thereby calculating an approximate straight line characterizing the center line of the burner flame and the angle of the burner flame; and including the approximate straight line in the characteristic quantities of the shape of the burner flame.

[0011] The manufacturing method of any of the porous glass substrates may further include the following steps: detecting the boundary points between the deposition surface and the background by performing image processing on the image data; and calculating the feature quantity of the front center position of the deposition surface as a feature quantity of the shape of the deposition surface based on the coordinate data of the detected boundary points. The manufacturing method of any of the porous glass substrates may further include the following steps: detecting the boundary points between the burner flame and the background by performing image processing on the image data; calculating each of a plurality of measurement positions on the extension line of the central axis of the burner as the coordinates of the center position of the burner flame based on the coordinate data of the detected boundary points; and calculating an approximate straight line characterizing the center line of the burner flame as a feature quantity of the shape of the burner flame. The manufacturing method of any of the porous glass substrates may further include the following steps: using the feature quantity of the front center position of the deposition surface and the approximate straight line, calculating the horizontal distance between the front center position of the deposition surface and the center line of the burner flame as a feature quantity of the spatial relationship.

[0012] The manufacturing method of any of the porous glass substrates may further include the following steps: detecting the boundary points between the deposition surface and the background by performing image processing on the image data; and calculating the characteristic quantity of the inclination of the deposition surface as a characteristic quantity of the shape of the deposition surface based on the coordinate data of the detected boundary points. The manufacturing method of any of the porous glass substrates may further include the following steps: detecting the boundary points between the burner flame and the background by performing image processing on the image data; calculating each of multiple measurement positions on the extension line of the burner's central axis as the coordinates of the center position of the burner flame based on the coordinate data of the detected boundary points; and calculating an approximate straight line characterizing the angle of the burner flame as a characteristic quantity of the shape of the burner flame. The manufacturing method of any of the porous glass substrates may further include the following steps: using the characteristic quantity of the inclination of the deposition surface and the approximate straight line, calculating the angle between the centerline of the deposition surface and the centerline of the burner flame as a characteristic quantity of the spatial relationship.

[0013] In a second aspect of the present invention, a method for manufacturing a transparent glass substrate is provided. The method for manufacturing a transparent glass substrate may include the following steps: dehydrating the porous glass substrate obtained by any of the methods for manufacturing porous glass substrates using a heating furnace to make it transparent and vitrified, thereby producing a transparent glass substrate. The method for manufacturing a transparent glass substrate may include the following steps: measuring the refractive index distribution in the radial direction at multiple locations along the length of the transparent glass substrate. The method for manufacturing a transparent glass substrate may include the following steps: calculating characteristic quantities of the refractive index distribution using a computer based on the refractive index distribution.

[0014] In any of the methods for manufacturing transparent glass parent materials, the relative refractive index difference Δ(r) at a position of radius r in the core layer of the optical fiber is expressed as Δ(r) = 100 × (n(r) - n) / n(r) with the refractive index n of the cladding of the optical fiber as a reference. When the relative refractive index difference at the center of the core (r = 0) is set as Δ1, and the maximum value of the relative refractive index difference near the interface between the core layer and the cladding of the optical fiber is set as Δ2, the characteristic quantity of the refractive index distribution can be at least one of Δ1, Δ2, the ratio of Δ1 to Δ2 calculated based on the refractive index distribution measured for the transparent glass parent material, the core diameter, and the inferred value of the optical properties of the optical fiber inferred from the refractive index distribution measured for the transparent glass parent material. The inferred value of the optical properties of the optical fiber is at least one of the cutoff wavelength, mode field diameter, and zero dispersion wavelength.

[0015] The manufacturing method of any of the transparent glass substrates may further include the following steps: constructing a database of the characteristic quantities of the spatial relationship and the characteristic quantities of the refractive index distribution. The manufacturing method of any of the transparent glass substrates may further include the following steps: using data stored in the database, including characteristic quantities of the shape of the deposition surface, characteristic quantities of the shape of the burner flame, and characteristic quantities of the spatial relationship as explanatory variables, and characteristic quantities of the refractive index distribution as target variables, using the explanatory variables and the target variables to enable the learning model to learn, thereby constructing a learned model.

[0016] The manufacturing method of any of the transparent glass parent materials may further include the following steps: Expressing the relative refractive index difference Δ(r) at a position of radius r in the fiber core layer as Δ(r) = 100 × (n(r) - n) / n(r), setting the relative refractive index difference at the core center (r = 0) as Δ1, and setting the maximum value of the relative refractive index difference near the interface between the core layer and the cladding as Δ2, using the learned model, based on the characteristic quantities of the shape of the deposition surface, the shape of the burner flame, and the spatial relationship calculated from the image data during the manufacturing process of the porous glass parent material, inferring at least one of the following as characteristic quantities of the refractive index distribution: Δ1, Δ2, the ratio of Δ1 to Δ2, the core diameter, and at least one of the cutoff wavelength, mode field diameter, and zero-dispersion wavelength. The manufacturing method of any of the transparent glass parent materials may further include the following step: determining whether the inferred value deviates from the target value. The method for manufacturing any of the transparent glass preforms may further include the following steps: if the inferred value deviates from the target value, decide to take at least one of the following measures: stop manufacturing the porous glass preform; extend the manufacturing time of the porous glass preform; adjust the conditions of the gas used in manufacturing the porous glass preform; adjust the position of the burner that injects the burner flame; and maintain the burner.

[0017] In a third aspect of the invention, a method for manufacturing a transparent glass matrix is ​​provided. This method involves depositing glass microparticles generated by a burner flame onto the front end of a starting material rotating about a vertical axis, and then lifting the starting material upwards to form a porous glass matrix. The porous glass matrix is ​​then dehydrated in a furnace to achieve transparent vitrification, thereby forming a transparent glass matrix. The method includes the following steps: capturing images of the deposition surface of the glass microparticles in the starting material and the burner flame using one or more cameras, and projecting the images of the deposition surface and the burner flame onto the same coordinate plane to obtain image data; and inferring the values ​​of the optical properties by inputting the newly acquired image data into a learned model, wherein the learned model has learned the relationship between the image data and the values ​​of the optical properties of the optical fiber during the manufacturing process of the porous glass matrix.

[0018] In the manufacturing method of the transparent glass matrix, the relative refractive index difference Δ(r) at a position of radius r in the core layer of the optical fiber is expressed as Δ(r) = 100 × (n(r) - n) / n(r) with the refractive index n of the cladding of the optical fiber as a reference. When the relative refractive index difference at the center of the core (r = 0) is set as Δ1, and the maximum value of the relative refractive index difference near the interface between the core layer and the cladding of the optical fiber is set as Δ2, the optical properties of the optical fiber can be at least one of Δ1, Δ2, the ratio of Δ1 to Δ2, the core diameter, the cutoff wavelength, the mode field diameter, and the zero-dispersion wavelength.

[0019] The method for manufacturing any of the transparent glass preforms may further include the following steps: if the inferred value deviates from the target value, decide to take at least one of the following measures: stop manufacturing the porous glass preform; extend the manufacturing time of the porous glass preform; adjust the conditions of the gas used in manufacturing the porous glass preform; adjust the position of the burner that injects the burner flame; and maintain the burner.

[0020] In a fourth aspect of the present invention, an apparatus for manufacturing a porous glass matrix is ​​provided, wherein a porous glass matrix is ​​formed by depositing glass microparticles generated by a burner flame onto the front end of a starting material rotating about a vertical axis, and then lifting the starting material upward. The apparatus comprises: one or more cameras for capturing images of the deposition surface of the glass microparticles in the starting material and the burner flame; and a feature quantity calculation unit for projecting images of the deposition surface and the burner flame captured by the one or more cameras onto the same coordinate plane, acquiring the obtained image data, and performing image processing on the image data to calculate a feature quantity of the spatial relationship between the deposition surface and the burner flame.

[0021] Furthermore, the description of the invention does not list all the features of the invention. Additionally, sub-combinations of these feature groups may also constitute an invention. Attached Figure Description

[0022] Figure 1 An example of a transparent glass master material manufacturing apparatus 11 according to one embodiment. Figure 2 This is a flowchart illustrating an example of a method for manufacturing a transparent glass substrate according to one embodiment. Figure 3 This is an example of extracting the boundary points between the deposition surface of the porous glass substrate 1 and the background, and the boundary points between the flame of the core-forming burner 40 and the background through image processing according to one embodiment, and plotting them on the xy plane. Figure 4 An example illustrating the refractive index distribution of a transparent glass substrate according to one embodiment. Figure 5 An example representing the change over time within a batch of the relative distance d between the front center of the deposition surface and the center of the flame. Figure 6 This is an example illustrating the change in the sediment surface shape fitting coefficient a2 over time within a batch. Figure 7 It is a graph showing the correlation between the measured and predicted values ​​of the target variable in the training and test data when learning using a model implemented in one way. Figure 8 It is a graph showing the correlation between the measured and predicted values ​​of the target variable in the training and test data when learning using the model with comparative examples. Detailed Implementation

[0023] The present invention will now be described through embodiments thereof, but these embodiments do not limit the invention as described in the claims. Furthermore, not all combinations of features described in the embodiments are essential to the solution of the invention.

[0024] Figure 1 This illustrates an example of a transparent glass masterbatch manufacturing apparatus 11 according to one embodiment. The transparent glass masterbatch manufacturing apparatus 11 of this embodiment includes a porous glass masterbatch manufacturing apparatus 10, a sintering apparatus 200, a measuring apparatus 300, and an anomaly detection apparatus 400. Figure 1 In the diagram, the direction of signal flow is indicated by a thin arrow, and the conveying direction of the porous glass substrate 1 and the transparent glass substrate is indicated by a thick arrow. In the following description, the manufacturing apparatus 11 for the transparent glass substrate will sometimes be referred to simply as manufacturing apparatus 11, and similarly, the manufacturing apparatus 10 for the porous glass substrate 1 will sometimes be referred to simply as manufacturing apparatus 10.

[0025] The manufacturing apparatus 10 for the porous glass substrate 1 of this embodiment includes a chamber 20, a shaft 30, a core-forming burner 40, cladding-forming burners 50 and 60, CCD cameras 71 and 72, a light source 80, and a lifting and rotating device 90. The manufacturing apparatus 10 of this embodiment also includes an MFC (Mass Flow Controller) 100, an image processing device 110, an image analysis device 120, and a control device 130.

[0026] The apparatus 10 for manufacturing porous glass matrix 1 forms a porous glass matrix by depositing glass microparticles generated by a burner flame onto the front end of a starting material rotating around a vertical axis, and then lifting the starting material upwards. More specifically, the apparatus 10 deposits glass microparticles generated in a chamber 20 using the VAD (Vapor-phase Axial Deposition) method onto the axial front end of the starting material mounted on a shaft 30, which is suspended from above the chamber 20 and rises while rotating around a vertical axis. The VAD method involves introducing a combustible gas and an oxidizing gas into a burner to generate an oxyhydrogen flame, and then introducing a raw material gas such as silicon tetrachloride or germanium tetrachloride into the flame to cause a hydrolysis reaction, thereby generating glass microparticles. The apparatus 10 forms the porous glass matrix 1 by growing the deposited glass microparticles in a cylindrical shape on the front end of the starting material.

[0027] The core-forming burner 40 and the cladding-forming burners 50 and 60 are the burners used in the VAD method and are located below the chamber 20. The core-forming burner 40 is a burner for generating glass microparticles that form the core of the optical fiber. SiCl4 is supplied from a gas supply source to the core-forming burner 40 as a glass feed gas, GeCl4 is supplied as a dopant gas to increase the refractive index, H2 and O2 are supplied as flame-forming gases, and Ar or N2 is supplied as a sealing gas. Similarly, SiCl4 is supplied from a gas supply source to the cladding-forming burners 50 and 60 as a glass feed gas, H2 and O2 are supplied as flame-forming gases, and Ar or N2, air, etc., are supplied as sealing gases.

[0028] CCD cameras 71 and 72 are examples of one or more cameras used to capture the deposition surface of glass microparticles in the starting material and the burner flame. CCD camera 71 is disposed outside chamber 20, captures the deposition surface of glass microparticles in the starting material, and outputs the image data to image processing apparatus 110. CCD camera 72 is disposed outside chamber 20, captures the burner flame ejected from the core-forming burner 40, and outputs the image data to image processing apparatus 110. In the following description, the burner flame will sometimes be simply referred to as a flame. Alternatively, one or more CMOS cameras may be used instead of CCD cameras 71 and 72 in the manufacturing apparatus 10.

[0029] A light source 80 is disposed outside the chamber 20 and illuminates the porous glass substrate 1. By illuminating the porous glass substrate 1 with light captured by CCD cameras 71 and 72, the light source 80 facilitates image recognition or image processing in the image processing apparatus 110 and the like. Considering the wavelength sensitivity of the CCD cameras 71 and 72, the light source 80 can, for example, illuminate light in the wavelength range of 400 to 700 nm. As the light source 80, a white LED lamp can be used, for example, which has the highest light intensity in the wavelength range of 460 to 470 nm and also has a peak energy intensity in the wavelength range of 550 to 600 nm. As an example, the light source 80 is configured to illuminate at least the deposition surface opposite to the deposition surface in contact with the flame of the core forming burner 40.

[0030] The lifting and rotating device 90 raises and rotates the shaft 30. More specifically, the lifting and rotating device 90 causes the shaft 30 to rise while rotating about a vertical axis, suspended from above the chamber 20 inside the chamber 20. The MFC 100 controls the flow rate of the raw material gas supplied from the gas supply source to the core forming burner 40 and the cladding forming burners 50 and 60.

[0031] Image processing device 110 causes CCD cameras 71 and 72 to capture images, and performs image processing on the image data input from CCD cameras 71 and 72 using methods described in detail below. Image processing device 110 outputs the image-processed image data to image analysis device 120. Image analysis device 120 analyzes the image data input from image processing device 110. Image analysis device 120 can use programming languages ​​such as Python (registered trademark) to sequentially process image data acquired during the deposition of glass microparticles on the front end of the starting material. Image analysis device 120 transmits the analysis results to anomaly detection device 400 via wired or wireless communication.

[0032] The control device 130 controls the operation of the CCD cameras 71 and 72, the lifting and rotating device 90, the image processing device 110, and the image analysis device 120. As an example, when manufacturing the porous glass substrate 1, the control device 130 adjusts the lifting speed of the lifting and rotating device 90 relative to the shaft 30 to keep the front end position of the porous glass substrate 1 fixed in the vertical direction. This lifting speed adjustment can use known methods, such as the method disclosed in Japanese Patent No. 3199642.

[0033] The porous glass substrate 1 produced by the manufacturing apparatus 10 is transferred into the sintering apparatus 200, whereby the sintering apparatus 200 produces a transparent glass substrate from the porous glass substrate 1. More specifically, the sintering apparatus 200 uses a heating furnace to dehydrate the porous glass substrate 1 formed by the manufacturing apparatus 10, causing it to become transparent and vitrified, thereby forming a transparent glass substrate.

[0034] The transparent glass substrate manufacturing apparatus 11 can also produce optical fiber substrates by adjusting the cladding thickness of the transparent glass substrate in the sintering apparatus 200 or in a device outside the sintering apparatus 200. The manufacturing apparatus 11 can also produce optical fibers by spinning the optical fiber substrate into a predetermined diameter. The manufacturing apparatus 11 can also transfer the transparent glass substrate into the measuring device 300 before adjusting the cladding thickness.

[0035] With the transparent glass substrate produced in the sintering apparatus 200 already in place, the measuring device 300, for example using a preform analyzer, measures the refractive index distribution of the transparent glass substrate in the radial direction at multiple locations along its length. More specifically, the measuring device 300 can scan the transparent glass substrate in-plane by incident laser light from its side along a cross-section perpendicular to the axis of the transparent glass substrate, and measure the change in the angle of refraction of the emitted light in-plane, thereby determining the refractive index distribution. The measuring device 300 transmits the measurement results to the anomaly detection device 400 via wired or wireless communication.

[0036] The anomaly detection device 400 receives measurement results from the measuring device 300. Based on the refractive index distribution shown in the measurement results, the anomaly detection device 400 infers the optical properties of the optical fiber made from the transparent glass substrate with the measured refractive index distribution, i.e., the target optical fiber. The optical properties of the optical fiber may be, for example, at least one of the cutoff wavelength, mode field diameter, and zero-dispersion wavelength.

[0037] The anomaly detection device 400 can use the finite element method to infer the optical properties of the optical fiber based on the refractive index distribution. More specifically, the inferred values ​​of the optical properties of the optical fiber can be calculated by applying the refractive index distribution to the dimensions of the target optical fiber and solving Maxwell's equations, etc.

[0038] The anomaly detection device 400 may receive the analysis results from the image analysis device 120 of the manufacturing apparatus 10 instead of receiving the measurement results from the measurement device 300, or it may receive the analysis results from the image analysis device 120 of the manufacturing apparatus 10 in addition to receiving the measurement results from the measurement device 300. The anomaly detection device 400 may also use the learned model described in detail below to infer the values ​​of the optical properties of the optical fiber based on the analysis results of the image analysis device 120.

[0039] The anomaly detection device 400 determines whether the inferred value deviates from a pre-defined target value. If the inferred value deviates from the target value, the anomaly detection device 400 decides to take pre-defined measures. This target value is, for example, a value based on a standard (ITU-T G.652 recommendation, etc.). For example, if the inferred value is within ± a few percent of the target value, the anomaly detection device 400 may determine that the inferred value has not deviated from the target value.

[0040] When the anomaly detection device 400 decides to take a pre-defined measure, it instructs the manufacturing device 10 to take the determined measure. Alternatively, the anomaly detection device 400 may also transmit the intention to take the measure to the control device 130 of the manufacturing device 10 via wired or wireless communication.

[0041] Figure 2 This is a flowchart illustrating an example of a method for manufacturing a transparent glass masterbatch according to one embodiment. Using Figure 2 The flowchart illustrates an example of a method for manufacturing a transparent glass substrate using the transparent glass substrate manufacturing apparatus 11 of this embodiment, which has been described in a general manner above.

[0042] In the manufacturing apparatus 10, the image processing apparatus 110 projects the images of the deposition surface captured by the CCD cameras 71 and 72 and the image of the burner flame onto the same coordinate plane, acquires the obtained image data, and performs image processing on the image data. The image analysis apparatus 120 calculates the characteristic quantities of the spatial relationship between the deposition surface and the burner flame based on the image data processed by the image processing apparatus 110 (step S101).

[0043] The spatial relationship characteristics can refer to the characteristics of relative configuration or the characteristics of relative positional relationship. Examples of spatial relationship characteristics include, for instance, the six degrees of freedom relationship between the deposition surface and the burner flame, relative angles, and relative positions. In step S101, the image analysis device 120 of this embodiment further calculates the characteristic quantities of the burner flame shape and the characteristic quantities of the deposition surface shape based on the image data processed by the image processing device 110.

[0044] The details of the method for calculating the various feature quantities based on image data will be described below. Furthermore, the combination of the image processing device 110 and the image analysis device 120 is an example of a feature quantity calculation unit that projects images of the deposition surface and the burner flame captured by one or more cameras onto the same coordinate plane, acquires the resulting image data, performs image processing on the image data, and thereby calculates the feature quantities of the spatial relationship between the deposition surface and the burner flame.

[0045] Here, as described above, the CCD camera 71 captures the deposition surface of glass microparticles in the starting material and outputs it as image data via the image processing device 110. The CCD camera 72 captures the burner flame ejected from the core forming burner 40 and outputs it as image data via the image processing device 110.

[0046] The CCD camera 71 can also photograph the deposition surface of glass microparticles in the starting material from an angle that is not parallel to the central axis of the starting material, thereby enabling the assessment of the tilt of the axis of the porous glass matrix 1 formed from the starting material. The optical axis of the CCD camera 71 can have an angle of 45 degrees or more but less than 90 degrees relative to the central axis of the porous glass matrix 1, specifically, an angle of 60 degrees or more but less than 90 degrees, and more specifically, an angle of 80 degrees or more but less than 90 degrees.

[0047] Similarly, the CCD camera 72 can also photograph the flame of the core-forming burner 40 from an angle not parallel to the central axis of the core-forming burner 40, thereby enabling a lateral assessment of the shape of the flame ejected from the core-forming burner 40. The CCD camera 72 can also photograph the flame of the core-forming burner 40 from an angle not parallel to both the central axis of the core-forming burner 40 and the central axis of the porous glass substrate 1, thereby enabling an assessment of the position or direction at which the flame ejected from the core-forming burner 40 contacts the deposition surface of the porous glass substrate 1. The optical axis of the CCD camera 72 can have an angle of 45 degrees or more but less than 90 degrees relative to the central axis of the core-forming burner 40; specifically, it can have an angle of 60 degrees or more but less than 90 degrees; more specifically, it can have an angle of 80 degrees or more but less than 90 degrees. Furthermore, the optical axis of the CCD camera 72 can have an angle of 45 degrees or more but less than 90 degrees relative to the central axis of the porous glass matrix 1, or an angle of 60 degrees or more but less than 90 degrees, or an angle of 80 degrees or more but less than 90 degrees.

[0048] The angle θ (degrees) between the optical axis of CCD camera 71 and the optical axis of CCD camera 72 can be 0 degrees or more and 20 degrees or 160 degrees or more and 180 degrees. Specifically, it can be 0 degrees or more and 10 degrees or 170 degrees or more and 180 degrees. More specifically, it can be 0 degrees or more and 5 degrees or 175 degrees or more and 180 degrees. In this case, the image analysis device 120, which analyzes the image data from CCD cameras 71 and 72, can easily grasp the relative relationship between the shape of the deposition surface of the porous glass substrate 1 and the shape of the flame ejected from the core forming burner 40.

[0049] Alternatively, the optical axis of CCD camera 71 can be made parallel to the optical axis of CCD camera 72, meaning the angle between them is 0 degrees or 180 degrees. Therefore, a single CCD camera with a wide-angle lens can be used instead of both CCD camera 71 and CCD camera 72 to capture the deposition surface of glass microparticles in the starting material and the flame ejected from the core-forming burner 40 in the same frame. In this case, the relative positional relationship between the deposition surface and the flame can be evaluated in the same frame.

[0050] Alternatively, CCD cameras 71 and 72 can also capture images of the deposition surface and the flame from completely opposite directions. By inverting one of the images of the deposition surface and the flame, the two images are placed on the same coordinate plane, meaning that image data projected onto the same coordinate plane can be acquired. In this case, there are advantages: it is easy to simultaneously capture images from an illumination position with increased contrast between the porous glass substrate 1 and the background, and to capture images via a wavelength filter. Capturing images via a wavelength filter makes it easy to observe the shape of the flame stream ejected from the burner 40 used to form the fiber core. Furthermore, in the case of using a single camera, for example, the camera used to capture the flame can be an infrared camera, and the spectral sensitivity characteristics of the two cameras can also be set differently. Additionally, the manufacturing apparatus 10 can use multiple cameras to capture images of the deposition surface and the flame from various angles. In this case, the manufacturing apparatus 10 can increase the amount of information acquired, thereby improving the accuracy of the regression described below.

[0051] The CCD cameras 71 and 72 can capture images at synchronized time points. In this case, when the rotational speed of axis 30 is set to r (revolutions per minute), the time difference t (seconds) between the two cameras can be set as [Equation 1]. Or [Equation 2] In this case, it is easy to evaluate the relative positional relationship between the front end of the porous glass matrix 1 and the flame ejected from the burner 40 for core forming on the same screen. Alternatively, it can be set as [Equation 3], for example. Or [Equation 4] More specifically, it can be approximated by [Equation 5]. Or [Equation 6]

[0052] Alternatively, CCD camera 71 and CCD camera 72 can be used to capture and temporarily store images in advance via video recording, and still images can be extracted from the continuous data at the time points described above.

[0053] As modules for CCD cameras 71 and 72, both monochrome and color camera modules can be used. Regarding the spectral sensitivity characteristics of the CCD camera, the peak sensitivity is between 450 and 600 nm, and the sensitivity at 850 nm can be more than 10% of the peak sensitivity. Furthermore, the more effective pixels a CCD camera has, the higher the resolution of the acquired data, but the larger the data volume also is. Therefore, cameras with an effective pixel count of, for example, 768 × 494 pixels can be used for CCD cameras 71 and 72.

[0054] In the manufacturing apparatus 10 of this embodiment, as an example, a characteristic quantity of the spatial relationship between the deposition surface and the burner flame is calculated based on image data obtained by projecting an image of the deposition surface and an image of the burner flame onto the same coordinate plane. The horizontal distance between the front center position of the deposition surface and the center line of the burner flame can be calculated based on the image data as the characteristic quantity.

[0055] More specifically, the manufacturing apparatus 10 can detect boundary points between the deposition surface and the background by performing image processing on the image data, and calculate a feature quantity of the center position of the front end of the deposition surface based on the coordinate data of the detected boundary points. The feature quantity of the center position of the front end of the deposition surface is one example of a feature quantity of the shape of the deposition surface. In addition, or alternatively, the manufacturing apparatus 10 can also calculate at least one of the following feature quantities based on the coordinate data: a feature quantity of the inclination of the deposition surface, a feature quantity of the outer diameter shape of the deposition surface, and a feature quantity of the degree of deformation of the deposition surface. At least one of these feature quantities can also be included in the feature quantity of the shape of the deposition surface.

[0056] The manufacturing apparatus 10 can also detect boundary points between the burner flame and the background by performing image processing on the image data. Based on the coordinate data of the detected boundary points, it calculates the coordinates of each of multiple measurement positions on the extension line of the burner's central axis as the center position of the burner flame, and calculates an approximate straight line characterizing the center line of the burner flame. This approximate straight line is an example of a characteristic quantity of the burner flame's shape. Furthermore, this approximate straight line can also characterize the angle of the burner flame. In addition to this approximate straight line, other characteristic quantities of the burner flame's shape may include, for example, the width of the burner flame.

[0057] The manufacturing apparatus 10 may also use a characteristic quantity of the center position of the front end of the deposition surface and an approximate straight line to calculate the horizontal distance between the center position of the front end of the deposition surface and the centerline of the burner flame as a characteristic quantity of the spatial relationship. Alternatively, the manufacturing apparatus 10 may also use a characteristic quantity of the inclination of the deposition surface and an approximate straight line to calculate the angle between the centerline of the deposition surface and the centerline of the burner flame as a characteristic quantity of the spatial relationship.

[0058] When detecting the coordinates of the boundary points between the deposition surface of the porous glass substrate 1 and the background based on the image data, as an example, the manufacturing apparatus 10 can focus on the brightness component of the image to determine the boundary points with large brightness variations between adjacent points. The manufacturing apparatus 10 can, for example, use the Canney edge detection method. When using the Canney edge detection method, the manufacturing apparatus 10 can perform a process to convert the image into a grayscale image consisting of 256 shades of gray from 0 (black) to 255 (white), or it can omit this process. Furthermore, if the manufacturing apparatus 10 uses a monochrome camera module as the CCD camera 71, since grayscale conversion has already been performed, there is no need to perform the grayscale image conversion operation. If the manufacturing apparatus 10 uses a color camera module as the CCD camera 71, it can also use the conversion formula y = 0.114B + 0.587G + 0.299R (ITU-R Rec BT.601 standard) to convert it into 256 shades of gray. Besides this, other methods for grayscale conversion include averaging the B, G, and R values, averaging the values ​​and adding gamma correction, using the conversion formula of the CIE XYZ standard, or extracting only the R channel. The appropriate grayscale method can be selected based on the obtained image data.

[0059] As an example of using the Canney edge detection method, the manufacturing apparatus 10 first sets two thresholds: a large threshold and a small threshold. If the derivative of a pixel value is above the large threshold, it is considered a candidate for the boundary point; if it is below the small threshold, it is considered not a boundary point and is excluded. If the derivative of a pixel value is between these two thresholds, the distinction is made based on the adjacency relationship between a pixel considered a candidate for a boundary point and a pixel considered outside the candidate boundary point. If a pixel is adjacent to a pixel considered a candidate boundary point, it is considered a candidate boundary point; otherwise, it is considered a non-boundary point. Candidate boundary points are obtained using the method described above.

[0060] Then, from these candidates, the following candidate coordinates are selected as boundary points. These candidates are compared with a range of boundary point coordinates that may be predetermined by pre-determining an object with the same shape as the porous glass substrate 1, which may be captured by CCD camera 71, and the candidate is selected if it falls within that range. Furthermore, a group of adjacent boundary point sequences is selected, and if the brightness change of the sequence exceeds a pre-set specified value, the group is deleted. Through this two-stage process, only the boundary points between the deposition surface and the background are detected, and these are output as coordinates on the xy plane.

[0061] When the manufacturing apparatus 10 calculates the characteristic quantities of the deposition surface based on the coordinate data of the boundary points between the deposition surface and the background of the obtained porous glass substrate 1, for example, based on the coordinate data of multiple boundary points... [Formula 7](x0,y0),(x1,y1),(x2,y2),...,(x n y n ), The coordinates of the center position of the deposition surface are calculated on multiple straight lines orthogonal to the central axis of the porous glass matrix 1, and the center line of the deposition surface is calculated from the average point. At this time, the coordinates of the intersection of the lower end of the deposition surface and the center line of the deposition surface are taken as the center position of the front end of the deposition surface, that is, [Equation 8] (b0, a0).

[0062] The manufacturing apparatus 10 further calculates the coefficients of a polynomial of degree 3 or higher that approximates the coordinate data of multiple boundary points. This approximation can be performed using the least squares method. [Formula 9]y=a0+a1(x-b0)+a2(x-b0) 2 +a3(x-b0) 3 +...

[0063] In this way, the manufacturing apparatus 10 can calculate the characteristic quantity of the front center position of the deposition surface as [Equation 10](b0, a0), the characteristic quantity of the inclination of the deposition surface as [Equation 11]a1, the characteristic quantity of the outer diameter shape of the deposition surface as [Equation 12]a2, the characteristic quantity of the degree of deformation of the deposition surface as [Equation 13]a3, and so on. The manufacturing apparatus 10 can also approximate it using a polynomial containing higher-order components. For higher-order components, since the coefficient values ​​are small, they can be scaled as needed. As described above, the manufacturing apparatus 10 can select an appropriate grayscale method based on the obtained image data, for example, several different grayscale methods can be used. Specifically, the manufacturing apparatus 10 can detect the coordinates of the boundary points in advance using several different grayscale methods, and calculate the characteristic quantity of the deposition surface by performing the operation on the boundary points obtained from these different grayscales (scale 1, scale 2, ..., etc.), for example, obtaining individual characteristic quantities such as a1 (scale 1), a1 (scale 2), ... for a1. Individual feature quantities obtained using several different grayscale methods can be used for learning the learning model described in detail below.

[0064] The manufacturing apparatus 10 can quantify the differences in complex deposition surface shapes into several characteristic quantities, for example, by acquiring and calculating these characteristic quantities at fixed time intervals during the manufacturing process, and storing them as data. The manufacturing apparatus 10 can also acquire and calculate characteristic quantities for each manufacturing batch and store them as data. Furthermore, when using multiple VAD manufacturing apparatuses, the manufacturing apparatus 10 can acquire and calculate characteristic quantities for each manufacturing apparatus and store them as data. In this case, there is an advantage: the stored manufacturing characteristic quantity data of these porous glass substrates 1 can be compared with the refractive index distribution data of the subsequently obtained transparent glass substrate. Similarly, the manufacturing apparatus 10 can also acquire and calculate the characteristic quantities of the burner flame at fixed time intervals, for example, and store them as data. The manufacturing apparatus 10 can also use the temporal variations of the characteristic quantities of the deposition surface shape and the characteristic quantities of the burner flame shape, such as the rate of change of each shape or flame fluctuations, as new characteristic quantities for calculating the spatial relationships.

[0065] In the manufacturing apparatus 10, when detecting the boundary points between the flame and the background based on the image data and calculating the characteristic quantities of the flame based on the coordinate data of the detected boundary points, as an example, multiple measurement positions are set on the image along the extension line of the burner's central axis from the flame ejection tip of the core-forming burner 40 towards the deposition surface of the porous glass substrate 1. At each measurement position, the manufacturing apparatus 10 sets a straight line in a direction that crosses the extension line of the burner's central axis and calculates the brightness distribution on these straight lines of the image data.

[0066] As an example of calculating this brightness distribution, the manufacturing apparatus 10 can first define the area where the flame is captured, and within this area, convert it to a 256-level grayscale value using the conversion formula y = 0.114B + 0.587G + 0.299R (ITU-R Rec BT.601 standard). Other grayscale conversion methods include averaging the B, G, and R values, adding gamma correction to the averaging method, using the conversion formula of the CIE XYZ standard, or extracting only the R channel. An appropriate grayscale method can be selected based on the obtained image data.

[0067] The manufacturing apparatus 10 performs a first-order derivative on the brightness distribution calculated along each straight line to calculate the brightness distribution derivative. The manufacturing apparatus 10 defines the maximum and minimum values ​​of this brightness distribution derivative as the boundary points between the flame and the background. Regarding the flame, unlike images of the deposited surface, there may be images with blurred flame outlines, or images where bright stripes appear in the flame due to raw material reactions. If the maximum and minimum values ​​of the derivative are taken, the boundary points between the flame and the background may not be accurately captured. In such cases, the manufacturing apparatus 10 can calculate the cross-sectional area of ​​the brightness distribution along each straight line, determine a threshold corresponding to the ratio of this cross-sectional area (e.g., 0.001), and define the ends of regions where the brightness exceeds this value as the boundary points between the flame and the background. In this case, when the brightness distribution characteristics differ between the right and left ends of the flame, the definition of the boundary points between the flame and the background can also be different at the left and right ends.

[0068] The manufacturing apparatus 10 further uses the boundary point data of the flame and the background obtained from multiple measurement positions on the extension of the central axis of the burner 40 for core forming to determine the coordinates of the central position of the flame, and arranges them on the same coordinate plane as the deposition surface before evaluation, and calculates the following approximate straight line characterizing the center of the flame and the angle of the flame, namely the center line of the flame. [Formula 14]y=kx+m The angle of the flame is represented by k, and these are used as characteristic quantities of the flame.

[0069] Figure 3 This is an example of extracting the boundary points between the deposition surface of the porous glass substrate 1 and the background, and the boundary points between the flame of the core-forming burner 40 and the background through image processing according to one embodiment, and plotting them on the xy plane. Figure 3 The xy plane shown is an example of the same coordinate plane as the image of the projected deposition surface and the image of the burner flame described above.

[0070] The coordinates where the previously calculated centerline of the flame intersects at the height of the front end of the deposition surface [Equation 15] (y = a0) are [Equation 16]. Therefore, the horizontal relative distance d between the center position of the front end of the deposition face and the center line of the flame is calculated by the following formula. [Formula 17] This becomes a characteristic quantity of the relative position of the deposition surface and the flame. By calculating the relative position of the deposition surface and the flame, the manufacturing apparatus 10 can make comparisons even if the deposition center position or the flame center position changes between manufacturing batches or manufacturing apparatuses.

[0071] When comparing the measured value of the refractive index distribution of the transparent glass substrate with the characteristic quantity obtained from the image of the deposition process, the manufacturing apparatus 11 can photograph the deposition surface of the porous glass substrate 1 at intervals of 2000 times or less, specifically 200 times or less, and more specifically 20 times or less, the spatial resolution of the preform analyzer used to measure the refractive index distribution of the transparent glass substrate. In other words, when using a preform analyzer with a spatial resolution of 0.05 mm, the deposition surface of the porous glass substrate 1 can be photographed at intervals of 50 mm or less, specifically 5 mm or less, and more specifically 1 mm or less from the porous glass substrate 1.

[0072] When the porous glass substrate 1 prepared by the VAD method is sintered to make it transparent, the axial length is shortened to about half. Therefore, when the rise speed of the axis 30 during the deposition process is set to v (mm / min), the porous glass substrate 1 obtained by depositing glass microparticles for 1 minute is approximately equivalent to v / 2 (mm) of the sintered transparent glass substrate. Therefore, for example, when the rise speed is 1 mm / min, when the deposition surface is photographed at a time interval of 2 minutes, it is possible to compare the results with those obtained by measuring the refractive index distribution of the transparent glass substrate at a spacing of 1 mm.

[0073] The image data results are largely consistent with the refractive index distribution measurements obtained at the minimum measurement interval of the preform analyzer. Regarding the interval for photographing the sedimentary surface, images can be taken at intervals of 1 minute, 30 seconds, 10 seconds, or 1 second. Shorter intervals increase the likelihood of identifying transient changes during the deposition process, such as flame turbulence or changes in sedimentary surface shape. However, excessively short intervals require processing large amounts of image data, thus increasing image processing time.

[0074] The image processing of the image data can be performed sequentially during the deposition of glass microparticles, thereby constructing a manufacturing database that stores manufacturing data containing feature quantities obtained from image processing. Furthermore, a characteristic database of the transparent glass substrate can be constructed, storing feature quantities of the refractive index distribution along the length of the transparent glass substrate. A learning model can also be created by merging the data from the two databases based on their longitudinal positional relationships, or by calculating averages and merging them.

[0075] In the manufacturing apparatus 11, for example, the anomaly detection device 400 can calculate characteristic quantities of the refractive index distribution based on the refractive index distribution of the transparent glass substrate measured by the measuring device 300. Here, the relative refractive index difference Δ(r) at a position of radius r in the fiber core layer is expressed as Δ(r) = 100 × (n(r) - n) / n(r) with the refractive index n of the fiber cladding as a reference. The relative refractive index difference at the center of the core (r = 0) is defined as Δ1, and the maximum value of the relative refractive index difference near the interface between the core layer and the cladding of the fiber is defined as Δ2. Under such definition, the characteristic quantities of the refractive index distribution calculated based on the measured refractive index distribution of the transparent glass substrate can be, for example, at least one of Δ1, Δ2, the ratio of Δ1 to Δ2 calculated based on the refractive index distribution measured for the transparent glass substrate, the core diameter, and the inferred value of the optical properties of the fiber inferred from the refractive index distribution measured for the transparent glass substrate. The inferred optical properties of an optical fiber can be at least one of the cutoff wavelength, mode field diameter, and zero-dispersion wavelength.

[0076] The anomaly detection device 400 can construct a database of characteristic quantities of the spatial relationship and the refractive index distribution. The anomaly detection device 400 can also use data stored in this database, including characteristic quantities of the shape of the deposition surface, the shape of the burner flame, and the spatial relationship as explanatory variables, and characteristic quantities of the refractive index distribution as target variables, to enable the learning model to learn, thereby constructing a learned model. The learned model can be constructed by a device external to the anomaly detection device 400, or the anomaly detection device 400 can obtain the learned model from an external device. Similarly, the database can be constructed by a device external to the anomaly detection device 400, or the anomaly detection device 400 can obtain the database from an external device.

[0077] Following Figure 2After step S101, the anomaly detection device 400 uses the learned model to infer at least one of the following as a characteristic quantity of the refractive index distribution based on the characteristic quantity of the shape of the deposition surface, the characteristic quantity of the shape of the burner flame, and the characteristic quantity of the spatial relationship calculated from the image data of the manufacturing process of the porous glass substrate 1: Δ1, Δ2, the ratio of Δ1 to Δ2, the core diameter, and at least one of the cutoff wavelength, the mode field diameter, and the zero-dispersion wavelength (step S103).

[0078] The anomaly detection device 400 determines whether the inferred value deviates from a predetermined target value (step S105). If the inferred value deviates from the target value (step S105: Yes), the anomaly detection device 400 decides to take predetermined measures, instructs the manufacturing device 10 on the determined measures (step S107), and ends the process. If, in step S105, the anomaly detection device 400 does not deviate from the target value (step S105: No), it decides not to take the measures (step S109), and ends the process.

[0079] The pre-defined measures may be, for example, at least one of the following: stopping the manufacture of porous glass substrate 1 using the manufacturing apparatus 10; extending the manufacture time of porous glass substrate 1 in the manufacturing apparatus 10; adjusting the conditions of the gas used in the manufacture of porous glass substrate 1 in the manufacturing apparatus 10; adjusting the position of any burner in the manufacturing apparatus 10; and maintaining any burner in the manufacturing apparatus 10. Furthermore, maintaining the burner may refer to at least one of cleaning the burner, replacing the burner, cleaning the burner cover, and replacing the burner cover.

[0080] In this embodiment, the transparent glass substrate manufacturing apparatus 11, under predetermined conditions, infers the optical properties of the target optical fiber for the porous glass substrate 1 produced by the manufacturing apparatus 10, and determines whether the measures need to be taken based on the inferred values.

[0081] For example, the manufacturing apparatus 11 can analyze the image data using the image analysis device 120 after a predetermined time has elapsed since the manufacturing apparatus 10 began depositing glass microparticles onto the front end of the starting material, or whenever a predetermined time has elapsed, and determine whether the aforementioned measures need to be taken based on the analysis results. In this case, at the point in time when the determination of whether the aforementioned measures need to be taken, the fabrication of the porous glass substrate 1 is not yet complete, and the refractive index distribution is not measured using the measuring device 300. In this case, the manufacturing apparatus 11 can decide to take the aforementioned measures midway through the deposition of glass microparticles onto the starting material in the manufacturing apparatus 10, that is, anomaly detection can be performed early and feedback can be immediately fed back into the manufacturing conditions, thereby stabilizing the optical properties of the target optical fiber made from the starting material.

[0082] For example, the manufacturing apparatus 11 may determine whether the aforementioned measures are needed each time a porous glass substrate 1 is produced by the manufacturing apparatus 10. For example, the manufacturing apparatus 11 may also determine whether the aforementioned measures are needed each time a predetermined number of porous glass substrates 1 are produced by the manufacturing apparatus 10. For example, the manufacturing apparatus 11 may also determine whether the aforementioned measures are needed whenever the operating time of the manufacturing apparatus 10 meets a predetermined time. For example, the manufacturing apparatus 11 may also determine whether the aforementioned measures are needed if one or both of the following conditions are met: a predetermined number of porous glass substrates 1 have been produced by the manufacturing apparatus 10; and the operating time of the manufacturing apparatus 10 has met a predetermined time. In these cases, the manufacturing apparatus 11 may decide to take the aforementioned measures before the next or subsequent production of porous glass substrates 1 in the manufacturing apparatus 10, that is, to immediately feed back into the manufacturing conditions, thereby stabilizing the optical properties of the target optical fiber produced from the next or subsequent porous glass substrates 1.

[0083] Figure 4 An example of the refractive index distribution of a transparent glass matrix according to one embodiment is shown. The refractive index distribution of the transparent glass matrix includes the relative refractive index difference of the core layer with a high refractive index, such as germanium, and the relative refractive index difference of the cladding. The relative refractive index difference Δ(r) at a position of radius r is expressed as follows, with the refractive index n of the cladding as a reference, as Δ(r) = 100 × (n(r) - n) / n(r). Figure 4 The vertical axis refers to the relative refractive index difference [%).

[0084] The relative refractive index difference at the core center (r=0) is set as Δ1, and the maximum relative refractive index difference near the interface between the core layer and the cladding is set as Δ2. In this case, as described above, except... Figure 4 In addition to Δ1, Δ2, the ratio of Δ1 to Δ2 (i.e., Δ2 / Δ1), and the core diameter, the inferred values ​​of optical fiber optical properties such as cutoff wavelength, mode field diameter, and zero-dispersion wavelength, which are calculated theoretically using the fiber dimensions, can also be used as characteristic quantities of the refractive index distribution obtained from the measurement results of the refractive index distribution.

[0085] The variables can also be used to train the learning model to construct a learned model. As the learning model algorithm, multiple regression, Lasso regression, Ridge regression, Support Vector regression, Random Forest regression, etc., can be used, or the algorithm with the highest analytical accuracy can be used. Alternatively, a combination of these methods can be used as the learning model algorithm; for example, random forest regression can be used after selecting features using Lasso regression, the average of the results of the various regressions exemplified here can be taken, or a majority vote of multiple regressions can be used.

[0086] The manufacturing apparatus 11 can also prepare a learned model created in this way in advance, and use the learned model to directly infer the characteristic quantities or optical properties of the refractive index distribution from the image data during the manufacturing process. That is to say, the manufacturing apparatus 11 can also omit the image analysis device 120 and the measuring device 300, and omit the process of calculating characteristic quantities of spatial relationships, etc.

[0087] In this scenario, for example, the image processing device 110 uses one or more cameras to capture images of the deposition surface of glass microparticles in the starting material and the burner flame, and projects the images of the deposition surface and the burner flame onto the same coordinate plane to obtain the resulting image data. The anomaly detection device 400 infers the values ​​of the optical properties of the transparent glass substrate by inputting the newly acquired image data into a learned model, which has learned the relationship between the image data acquired by the image processing device 110 during the manufacturing process of the porous glass substrate 1 and the values ​​of the optical properties of the optical fiber. The values ​​of the optical properties of the optical fiber can be values ​​calculated based on measurements of the refractive index distribution in the radial direction measured at multiple locations along the length of the optical fiber. These optical properties can be, for example, at least one of Δ1, Δ2, the ratio of Δ1 to Δ2, the core diameter, the cutoff wavelength, the mode field diameter, and the zero-dispersion wavelength.

[0088] In one embodiment of the method for manufacturing the transparent glass substrate according to this method, during the VAD deposition process, SiCl4 and GeCl4 are supplied as glass raw materials to the core forming burner 40, and SiCl4 is supplied as glass raw material to the cladding forming burners 50 and 60. The lifting speed of the lifting and rotating device 90 relative to the shaft 30 is adjusted to keep the front end position of the porous glass substrate 1 in the vertical direction fixed.

[0089] The front end of the porous glass substrate 1 is photographed using a CCD camera 71. The image analysis device 120, which analyzes the image data output by the image processing device 110, quantifies the following during the deposition process: the coordinate data of the boundary point between the deposition surface of the porous glass substrate 1 and the background, the center position of the front end of the deposition surface, the coefficients a1 to a6 obtained by fitting the deposition surface with a sixth-order function, the coordinate data of the boundary point between the flame of the core forming burner 40 and the background, the center position of the flame, and the relative distance d (unit: pixels) between the center position of the front end of the deposition surface and the center of the flame.

[0090] Then, the porous glass substrate 1 deposited by the VAD method is heated in the electric furnace of the sintering apparatus 200 to dehydrate it and vitrify it to produce a transparent glass substrate. In the measuring apparatus 300, the refractive index distribution along the length of the transparent glass substrate is measured, and optical properties such as the cutoff wavelength or mode field diameter and zero dispersion wavelength are calculated and inferred.

[0091] exist Figure 5 The diagram shows an example of how the relative distance d between the front center of the deposition surface and the flame center changes over time within a batch. Figure 5 The horizontal axis represents manufacturing time (minutes), and the vertical axis represents relative distance d (pixels). Figure 6 The figure shows an example of how the sediment surface shape fitting coefficient a2 changes over time within a batch. Figure 6 The horizontal axis represents manufacturing time (minutes), and the vertical axis represents the deposition surface shape fitting coefficient a2. Figure 5 and Figure 6 In the graph, a thick solid line represents the curve of abnormal batch A, a thicker dashed line with larger gaps represents the curve of abnormal batch B, and a thinner dashed line with smaller gaps represents the curve of normal batch (average). Table 1 below shows the core inference values ​​(average value in the length direction) of zero dispersion wavelength for abnormal batch A, abnormal batch B, and normal batch (average), as well as Δ1, Δ2, and Δ2 / Δ1. [Table 1]

[0092] As shown in Table 1, the zero-dispersion wavelength of the normal batch is 1308.1 nm. In contrast, the zero-dispersion wavelength of abnormal batch A is larger, at 1312.5 nm, and that of abnormal batch B is also larger, at 1310.7 nm. Furthermore, differences were observed in Δ1, Δ2, and Δ2 / Δ1 among abnormal batch A, abnormal batch B, and the normal batch, indicating a change in the refractive index distribution shape. Abnormal batches A and B, compared to other normal batches (target batches), show... Figure 5 As shown, the relative distance d between the center of the front end of the deposition surface and the center of the flame is relatively large, such as... Figure 6As shown, the shape fitting coefficient a2 of the deposition surface is relatively large. In particular, a2 is very large for anomalous batches A and B in the early stage of deposition. That is to say, compared with normal batches, the flame center of the core-forming burner 40 and the center of the porous glass matrix 1 of anomalous batches A and B move away from each other, the front end shape of the porous glass matrix 1 is deposited into a steeper shape, and a2 becomes larger.

[0093] As the porous glass substrate 1 becomes steeper, the refractive index distribution changes, leading to an increase in the zero-dispersion wavelength. One reason for the increased relative distance d between the front center of the deposition surface and the flame center is the deterioration of the burner cap front end over time. GeO2 or SiO2 from the raw materials may adhere to the burner cap front end, solidifying to form glass beads. When glass beads form, the flame state changes near the burner cap front end, altering the flame orientation and consequently the relative distance between the flame and ash particles. In this case, the flame position can be adjusted by replacing the burner cap, cleaning the burner cap, or adjusting the burner position to approximate a normal batch, thus ensuring a normal zero-dispersion wavelength. Therefore, by quantifying and analyzing the deposition surface shape and flame shape and position in real time during the glass microparticle deposition process using the manufacturing apparatus 11, abnormal batches A or B can be detected early, allowing for timely intervention and stabilizing characteristics within or between batches.

[0094] Using a manufacturing database containing manufacturing data with features obtained from image processing and a characteristic database of transparent glass substrate, the explanatory variables include features such as the front center position of the deposition surface, the tilt of the deposition surface, the outer diameter shape of the deposition surface, the degree of deformation of the deposition surface, the flame centerline, the flame angle, and the horizontal relative distance between the front center of the porous glass substrate and the center of the burner flame. The target variable includes the ratio of the relative refractive index difference Δ1 to Δ2. A learning model is created using random forest.

[0095] Figure 7 It is a graph showing the correlation between the measured and predicted values ​​of the target variable in the training and test data when learning using the model described in one embodiment. Figure 7 The horizontal axis represents the measured value of the target variable Δ2 / Δ1, and the vertical axis represents the predicted value of the target variable Δ2 / Δ1. Figure 7 In the diagram, training data is plotted using circles, and test data is plotted using triangles. The ratio of training data to test data is set to 75:25. Figure 7As shown, a significant correlation was observed between the measured and predicted values ​​of the target variable. By creating this predictive learning model, the manufacturing apparatus 11 is able to predict inferred values ​​of characteristic quantities or optical properties of the refractive index distribution based on the results obtained from image processing.

[0096] Secondly, as a comparative example with this embodiment, a learning model is created using random forest, which includes the characteristic quantity of the front center position of the deposition surface, the characteristic quantity of the inclination of the deposition surface, the characteristic quantity of the outer diameter shape of the deposition surface, and the characteristic quantity of the deformation degree of the deposition surface as explanatory variables, and the characteristic quantity of the ratio of the relative refractive index difference Δ1 to Δ2 as the target variable.

[0097] Figure 8 It is a graph showing the correlation between the measured and predicted values ​​of the target variable in the training and test data when learning using the model described in the comparative example. Figure 8 The horizontal axis represents the measured value of the target variable Δ2 / Δ1, and the vertical axis represents the predicted value of the target variable Δ2 / Δ1. Figure 8 In the diagram, training data is plotted using circles, and test data is plotted using triangles. For example... Figure 8 As shown, no significant correlation was observed between the measured and predicted values ​​of the target variable. Therefore, it can be concluded that a more accurate model can be created by using features that include not only the shape of the deposition surface but also the shape of the burner flame and the spatial relationship between the deposition surface and the burner flame as explanatory variables.

[0098] The present invention has been described above using embodiments, but the technical scope of the present invention is not limited to the scope described in the embodiments. It will be apparent to those skilled in the art that various modifications or improvements can be made to the embodiments. It is evident from the claims that forms obtained by applying the modifications or improvements described above may also be included within the protection scope of the present invention.

[0099] It should be noted that the execution order of actions, steps, procedures, and stages in the apparatus, systems, programs, and methods shown in the claims, description, and drawings can be implemented in any order, unless specifically stated as "before" or "prior to," and the output of the previous process is not used in the subsequent process. Even if terms such as "firstly" or "secondly" are used to describe the flow of actions in the claims, description, and drawings for convenience, it does not mean that the actions must be performed in that order. [Explanation of reference numerals in the attached figures]

[0100] 1: Porous glass matrix 10: Manufacturing equipment 20: Chamber 30: Axis 40: Burner for fiber core formation 50, 60: Burners for cladding formation 71, 72: CCD camera 80: Light source 90: Lifting and rotating device 100: MFC 110: Image processing device 120: Image Analysis Device 130: Control device 200: Sintering apparatus 300: Measuring device 400: Anomaly Detection Device

Claims

1. A method for producing a porous glass base material, characterized by, A porous glass preform is formed by depositing glass fine particles generated by a burner flame formed by a burner on the leading end of a starting material rotating around a vertical axis, and pulling the starting material upward. The manufacturing method includes the steps of: capturing the deposition surface of the glass fine particles in the starting material and the burner flame with one or more cameras, and projecting the image of the deposition surface and the image of the burner flame onto the same coordinate plane to obtain image data; and calculating a characteristic quantity of the spatial relationship between the deposition surface and the burner flame by image processing of the image data.

2. The method for producing a porous glass mother material according to claim 1, wherein Further including the step of calculating a characteristic quantity of the shape of the burner flame and a characteristic quantity of the shape of the deposition surface by image processing of the image data.

3. The method of producing a porous glass precursor material according to claim 1 or 2, characterized in that, In the one or more cameras, the optical axis of the camera that captures the deposition surface has an angle of 45 degrees or more and 90 degrees or less with respect to the central axis of the porous glass preform, specifically, an angle of 60 degrees or more and 90 degrees or less, more specifically, an angle of 80 degrees or more and 90 degrees or less, and the optical axis of the camera that captures the burner flame has an angle of 45 degrees or more and 90 degrees or less with respect to the central axis of the burner, specifically, an angle of 60 degrees or more and 90 degrees or less, more specifically, an angle of 80 degrees or more and 90 degrees or less.

4. The method of producing a porous glass precursor material according to claim 1 or 2, wherein In the one or more cameras, the angle between the optical axis of the camera that captures the deposition surface and the optical axis of the camera that captures the burner flame is 0 degrees or more and 20 degrees or less or 160 degrees or more and 180 degrees or less, specifically, 0 degrees or more and 10 degrees or less or 170 degrees or more and 180 degrees or less, more specifically, 0 degrees or more and 5 degrees or less or 175 degrees or more and 180 degrees or less.

5. The manufacturing method of the porous glass preform according to claim 1 or 2, wherein: The image data is obtained by capturing the deposition surface and the burner flame with one camera to capture the deposition surface and the burner flame onto the same screen.

6. The manufacturing method of the porous glass preform according to claim 1 or 2, wherein: The image data is obtained by capturing the deposition surface and the burner flame with two cameras that capture from completely opposite directions, and inverting one of the image of the deposition surface and the image of the burner flame to obtain the image data projected onto the same coordinate plane.

7. The method for producing a porous glass mother material according to claim 2, wherein The characteristic quantity of the shape of the deposition surface includes the steps of detecting a boundary point of the deposition surface and the background by image processing of the image data, calculating at least one of a characteristic quantity of the center position of the leading end of the deposition surface, a characteristic quantity of the inclination of the deposition surface, a characteristic quantity of the outer diameter shape of the deposition surface, and a characteristic quantity of the degree of deformation of the deposition surface from the coordinate data of the detected boundary point, and including the at least one characteristic quantity in the characteristic quantity of the shape of the deposition surface.

8. The method for producing a porous glass mother material according to Claim 2, wherein The feature quantity of the shape of the burner flame includes the following steps: detecting boundary points of the burner flame and the background by image processing of the image data, calculating each of a plurality of measurement positions on the extension line of the center axis of the burner as coordinates of the central position of the burner flame from coordinate data of the detected boundary points, thereby calculating an approximate straight line representing the center line of the burner flame and the angle of the burner flame, and including the approximate straight line in the feature quantity of the shape of the burner flame.

9. The method for producing a porous glass mother material according to Claim 2, wherein Further comprising the following steps: detecting boundary points of the deposition surface and the background by image processing of the image data, calculating a feature quantity of the front end center position of the deposition surface as a feature quantity of the shape of the deposition surface from coordinate data of the detected boundary points; detecting boundary points of the burner flame and the background by image processing of the image data, calculating each of a plurality of measurement positions on the extension line of the center axis of the burner as coordinates of the central position of the burner flame from coordinate data of the detected boundary points, and calculating an approximate straight line representing the center line of the burner flame as a feature quantity of the shape of the burner flame; and calculating a horizontal distance between the front end center position of the deposition surface and the center line of the burner flame as a feature quantity of the spatial relationship using the feature quantity of the front end center position of the deposition surface and the approximate straight line.

10. The method of producing a porous glass precursor material according to claim 2, wherein Further comprising the following steps: detecting boundary points of the deposition surface and the background by image processing of the image data, calculating a feature quantity of the inclination of the deposition surface as a feature quantity of the shape of the deposition surface from coordinate data of the detected boundary points; detecting boundary points of the burner flame and the background by image processing of the image data, calculating each of a plurality of measurement positions on the extension line of the center axis of the burner as coordinates of the central position of the burner flame from coordinate data of the detected boundary points, and calculating an approximate straight line representing the angle of the burner flame as a feature quantity of the shape of the burner flame; and calculating an angle formed by the center line of the deposition surface and the center line of the burner flame as a feature quantity of the spatial relationship using the feature quantity of the inclination of the deposition surface and the approximate straight line.

11. A method of manufacturing a transparent glass parent material, characterized by, Further comprising the following steps: dehydration of the porous glass base material obtained by the method for manufacturing a porous glass base material according to claim 2 using a heating furnace, transparent vitrification thereof, thereby manufacturing a transparent glass base material; determining a radial refractive index distribution at a plurality of positions in the length direction of the transparent glass base material; and calculating a feature quantity of the refractive index distribution by the computer from the refractive index distribution.

12. The method of producing a transparent glass base material according to claim 11, wherein Δ(r) = 100 x (n(r) - n) / n(r) is represented with respect to the refractive index n of the cladding layer of the optical fiber, Δ1 is set as the relative refractive index difference at the center of the core (r = 0), and Δ2 is set as the maximum value of the relative refractive index difference near the interface between the core layer and the cladding layer of the optical fiber, The characteristic quantity of the refractive index distribution is at least one of Δ1, Δ2, a ratio of Δ1 to Δ2, a core diameter, and at least one of a cutoff wavelength, a mode field diameter, and a zero dispersion wavelength, which are calculated based on the refractive index distribution measured for the transparent glass parent material, and which are inferred values of optical characteristics of an optical fiber inferred from the refractive index distribution measured for the transparent glass parent material.

13. The method of producing a transparent glass base material according to claim 11, wherein Further comprising the steps of: constructing a database of the characteristic quantity of the spatial relationship and the characteristic quantity of the refractive index distribution; and using the data stored in the database, using the characteristic quantity of the deposited surface shape, the characteristic quantity of the burner flame shape, and the characteristic quantity of the spatial relationship as explanatory variables, using the characteristic quantity of the refractive index distribution as a target variable, and using the explanatory variables and the target variable to cause a learning model to learn, thereby constructing a learned model.

14. The method of producing a transparent glass base material according to claim 13, wherein Further comprising the steps of: Δ(r) = 100 x (n(r) - n) / n(r) is represented with respect to the refractive index n of the cladding layer of the optical fiber, Δ1 is set as the relative refractive index difference at the center of the core (r = 0), and Δ2 is set as the maximum value of the relative refractive index difference near the interface between the core layer and the cladding layer of the optical fiber, using the learned model, inferring at least one value of Δ1, Δ2, a ratio of Δ1 to Δ2, a core diameter, and at least one of a cutoff wavelength, a mode field diameter, and a zero dispersion wavelength, as the characteristic quantity of the refractive index distribution, from the characteristic quantity of the deposited surface shape, the characteristic quantity of the burner flame shape, and the characteristic quantity of the spatial relationship calculated from the image data during the manufacturing process of the porous glass parent material; determining whether the inferred value deviates from a target value; and in the case where the inferred value deviates from the target value, deciding to take at least one of the following measures: stopping the manufacturing of the porous glass parent material; extending the manufacturing time of the porous glass parent material; adjusting the conditions of the gas used when manufacturing the porous glass parent material; adjusting the position of the burner that ejects the burner flame; and maintaining the burner.

15. A method of manufacturing a transparent glass parent material, characterized by, a porous glass parent material is formed by depositing glass fine particles generated by a burner flame on the leading end of a starting material that is rotated around a plumb axis and pulling the starting material upward, and the porous glass parent material is dehydrated by a heating furnace to be transparent glass, thereby forming a transparent glass parent material, and the manufacturing method includes the steps of: acquiring image data by projecting an image of the deposition surface and an image of the burner flame onto the same coordinate plane; and inference of the value of the optical characteristic by inputting the newly acquired image data into a learned model that has learned a relationship between the image data and the value of the optical characteristic of the optical fiber during the manufacturing process of the porous glass preform.

16. The method of manufacturing a transparent glass preform according to claim 15, characterized in that: the relative refractive index difference Δ(r) at a position of radius r in the core layer of the optical fiber is expressed as Δ(r) = 100 x (n(r) - n) / n(r) with the refractive index n of the cladding layer of the optical fiber, the relative refractive index difference at the core center (r = 0) is set as Δ1, and the maximum value of the relative refractive index difference in the vicinity of the interface between the core layer and the cladding layer of the optical fiber is set as Δ2, the value of the optical characteristic of the optical fiber is at least one of Δ1, Δ2, the ratio of Δ1 to Δ2, the core diameter, the cutoff wavelength, the mode field diameter, and the zero dispersion wavelength.

17. The method of producing a transparent glass mother material according to claim 15 or 16, characterized in that, the step of deciding to take at least one of the following measures in the case where the inferred value deviates from a target value: stopping the manufacturing of the porous glass preform; extending the manufacturing time of the porous glass preform; adjusting the conditions of the gas used during the manufacturing of the porous glass preform; adjusting the position of the burner that ejects the burner flame; and maintaining the burner.

18. An apparatus for manufacturing a porous glass matrix, characterized in that, a porous glass preform is formed by depositing glass particles generated by a burner flame on the leading end of a starting material that is rotated around a plumb axis and pulling the starting material upward, and the manufacturing apparatus is provided with: one or more cameras that capture the deposition surface of the glass particles in the starting material and the burner flame; and a feature quantity calculation section that acquires image data by projecting an image of the deposition surface and an image of the burner flame onto the same coordinate plane and performs image processing on the image data, thereby calculating a feature quantity of the spatial relationship between the deposition surface and the burner flame.

Images