CN-121835316-B - Processing design method based on diamond cutter for optical fiber cutting

CN121835316BCN 121835316 BCN121835316 BCN 121835316BCN-121835316-B

Abstract

The invention relates to the field of mechanical engineering, and discloses a processing design method based on a diamond cutter for optical fiber cutting. The method comprises the steps of collecting three-dimensional shapes and optical performance indexes of optical fiber end faces obtained by cutting by diamond cutters with different cutting edge parameters, extracting microscopic geometric characteristics such as cutting edge curvature gradient, crystal orientation, collapse defect density, surface power spectrum density and the like, constructing a physical interpretable neural network model with embedded fracture mechanics and optical waveguide constraint, reversely solving the cutting edge parameter set meeting target optical performance, generating an ultra-precise machining path according to the cutting edge parameter set, and controlling a single-point diamond lathe to custom machine the diamond cutters. The invention improves the stability of the reflectivity of the end face and the consistency of the mode field through closed loop optimization and online fine adjustment, reduces the standard deviation of return loss and meets the requirement of high-speed optical communication.

Inventors

LIU YUHUI
JIA CHUNSHENG
LIU CHANGXUN
WU XIAOPING

Assignees

湖南时光钻石科技有限公司

Dates

Publication Date: 20260512
Application Date: 20260312

Claims (10)

1. The processing design method based on the diamond cutter for optical fiber cutting is characterized by comprising the following steps: collecting end face samples formed by respectively cutting standard single-mode fibers by a plurality of diamond cutters with different cutting edge microscopic geometric characteristics; Scanning the three-dimensional surface morphology of each end surface sample to obtain surface height field data of the end surface sample, and synchronously measuring optical performance indexes of the end surface sample, wherein the optical performance indexes comprise end surface reflectivity, return loss and mode field distortion coefficients; Extracting microscopic geometric characteristic parameters of each diamond cutter blade, wherein the microscopic geometric characteristic parameters comprise curvature radius of a blade line contour, curvature gradient change rate, blade tip atomic level flatness, crystal orientation angles of front and rear surfaces of the blade, microscopic collapse density and depth distribution of the blade along the cutting direction and a power spectrum density function of blade surface roughness; Constructing a multi-input multi-output physical constraint neural network model based on the micro geometrical characteristic parameters and the corresponding end face optical performance indexes, wherein the physical constraint neural network model takes the cutting edge micro geometrical characteristic parameters as input, takes the end face optical performance indexes as output, and embeds a material fracture mechanics equation and an optical waveguide boundary condition as regularization constraint items; Setting a target optical performance index threshold range, taking the target optical performance index threshold range as an inverse solution target of the physical constraint neural network model, and iteratively optimizing microscopic geometric characteristic parameters of an input layer through a gradient back-propagation algorithm until the optical performance index of an output layer falls into the target optical performance index threshold range, so as to obtain a group of cutting edge microscopic geometric parameter solution sets meeting optical performance requirements; Converting the cutting edge microscopic geometric parameter solution set into an ultraprecise machining path instruction, and controlling a nano positioning platform of the single-point diamond lathe to carry out fly cutting machining on the single-crystal diamond blank according to the ultraprecise machining path to form a cutting edge structure with specified microscopic geometric morphology.
2. The method of claim 1, wherein the step of performing three-dimensional surface topography scanning on each end face sample to obtain surface height field data and simultaneously measuring optical performance indexes comprises: scanning a round area with the diameter of 100 micrometers in the central area of the end face of the optical fiber by adopting a white light interferometer to obtain surface height field data; The end surface reflectivity is measured by an optical time domain reflectometer, the return loss is measured by a polarization independent optical return loss tester, a mode field intensity distribution image is obtained by a near field scanning optical microscope, and a mode field distortion coefficient is calculated.
3. The method of designing a diamond cutter for optical fiber cutting according to claim 2, wherein extracting the microscopic geometric characteristic parameter of each diamond cutter edge comprises: Acquiring a cross-section image of the tip of the cutting edge by a focused ion beam scanning electron microscope, and fitting contour points within 10 nanometers of the forefront end of the cutting edge to a secondary curve to calculate the radius of curvature; calculating the change of the curvature radius in a unit distance along the length direction of the cutting edge by adopting a sliding window difference method to serve as the curvature gradient change rate; Observing the surface heights of five atomic layers at the forefront end of the cutting edge by a high-resolution transmission electron microscope, and calculating the standard deviation as the atomic-level flatness; measuring the crystal orientation angles of the front surface and the rear surface of the blade by utilizing an electron back scattering diffraction technology; counting the number of local depressions with the depth of more than 0.5 nanometers on the length of each micrometer cutting edge by an automatic image recognition algorithm to be used as the microscopic collapse density, and calculating the ratio of the maximum depth to the average depth to be used as the depth distribution representation; And performing fast Fourier transform on the cutting edge surface height data, then taking a mode square and normalizing to obtain a power spectrum density function of the surface roughness, and limiting the energy ratio of high-frequency components with spatial frequency greater than 1.5 cycles per micron to be below 15%.
4. The method for designing a diamond cutter for optical fiber cutting according to claim 3, wherein constructing a physical constraint neural network model based on the microscopic geometrical characteristic parameter and the corresponding end surface optical performance index comprises: setting an input layer to contain 7 neurons, wherein the input layer corresponds to a curvature radius, a curvature gradient change rate, an atomic level flatness standard deviation, a front tool face crystal direction deviation angle, a rear tool face crystal direction deviation angle, a microscopic collapse and defect density and a high-frequency power spectrum density ratio respectively; setting an output layer to comprise three neurons, wherein the three neurons correspond to the end surface reflectivity, the return loss and the mode field distortion coefficient respectively; setting two hidden layers, and adopting a correction linear unit for an activation function; and introducing a stress intensity factor calculation module based on the Grignard fracture theory into the loss function as a first physical constraint term, and introducing an electric field tangential component continuity constraint based on the Maxwell equation set boundary condition as a second physical constraint term.
5. The method for designing a diamond cutter for optical fiber cutting according to claim 4, the construction of the first physical constraint item is characterized by comprising the following steps: Calculating equivalent crack length and geometric correction factors according to the geometric features of the cutting edge, and calculating stress intensity factors by combining cutting stress; And mapping the stress intensity factor into a predicted value of the damage depth of the subsurface of the optical fiber end face, and forming a first physical constraint loss term with the square of the difference value of the actually measured damage depth.
6. The method for designing a diamond cutter for optical fiber cutting according to claim 5, the construction of the second physical constraint item is characterized by comprising the following steps: Based on the predicted end surface height field, calculating the normal derivative of the tangential component of the electric field at the air-quartz interface by utilizing finite difference approximation; and constructing electric field tangential component continuity residual errors according to the dielectric constant difference, and constructing a second physical constraint loss term by the square sum of the electric field tangential component continuity residual errors.
7. The method of designing a diamond cutter for cutting optical fibers according to claim 6, wherein setting a target optical performance index threshold range, using the target optical performance index threshold range as an inverse solution target of the physical constraint neural network model, and iteratively optimizing the microscopic geometrical characteristic parameters of the input layer by a gradient back propagation algorithm, comprises: setting a target threshold range of end surface reflectivity, return loss and mode field distortion coefficient; adopting a constrained projection gradient descent algorithm to carry out inverse solution; and after each iteration, performing physical feasibility verification on the microscopic geometric parameters, and eliminating non-feasible solutions beyond the feasible range.
8. The method of claim 7, wherein converting the cutting edge microscopic geometrical parameter solution set into an ultraprecise machining path instruction, comprises: Reconstructing a three-dimensional profile curve of the cutting edge according to the curvature radius and the curvature gradient change rate in the solution set; discretizing the three-dimensional profile curve of the cutting edge into a three-dimensional coordinate sequence; And generating a machining path instruction comprising a three-dimensional coordinate sequence of the tool path, a feeding speed, a cutting depth and a spindle rotating speed.
9. The method of designing a diamond cutter for optical fiber cutting according to claim 8, further comprising the step of, after finishing the diamond cutter, verifying: Cutting a new optical fiber sample by using the diamond knife, and repeating the optical performance measurement flow; if the measured optical performance index deviates from the target threshold by more than 5%, adding new data into the training set, performing online fine tuning on the physical constraint neural network model, and re-executing the inverse solving and processing flow until verification passes.
10. The method for designing a diamond cutter for cutting optical fibers according to claim 9, wherein the online fine tuning adopts a migration learning strategy, freezes weights of the first two layers of the neural network, and adjusts only the last layer and the output layer.

Description

Processing design method based on diamond cutter for optical fiber cutting Technical Field The invention belongs to the field of mechanical engineering, and particularly relates to a processing design method based on a diamond cutter for optical fiber cutting. Background With the rapid development of optical fiber communication, sensing and precision manufacturing technologies, the requirements on the processing quality of the end face of the optical fiber are increasingly stringent. The optical fiber cutting is used as a previous process of packaging and interconnecting optical devices, and the shape of the end face directly determines the transmission efficiency, return loss and connection reliability of optical signals. In this process, the diamond blade acts as a cutting tool, and the edge geometry is the fundamental factor affecting the cutting quality. The design of the traditional diamond cutter mainly depends on process experience and macroscopic mechanical analysis, focuses on mechanical properties such as cutter strength, wear resistance and the like, and fails to establish quantitative mapping relation between microscopic geometrical characteristics of a cutting edge and final optical properties of an optical fiber end face. The limitations of this design paradigm result in the microscopic topography of the cut end face, even if it is in a mechanically acceptable state of "flatness" or "no fringing" under a microscope, possibly with nanoscale undulations, microcracks, or non-ideal fracture surfaces, thereby inducing light scattering, mode mismatch, or fresnel reflection enhancement, resulting in unpredictable fluctuations in optical performance. The diamond-based optical fiber cutting process is essentially a controlled brittle fracture behavior, with the end profile being determined by the local stress field distribution imposed by the tool, the crack initiation location, and the propagation path. In the prior art, although the stress strain response of a cutter-optical fiber contact area can be simulated by a finite element method or the appearance of an end surface can be observed by an experimental means, the methods are isolated from each other, the mechanical simulation cannot predict the optical result, and the optical test is difficult to reversely push the blade shape to optimize the direction. A systematic design framework for modeling the coupling of solid mechanical response and light wave propagation characteristics is lacking, so that tool development is in an inefficient cycle of "trial-and-error-verification" for a long period of time. In the prior art, a diamond cutter design method for performing closed-loop correlation on a mechanical forming mechanism of an optical fiber cutting end face and optical transmission performance is not known. The current scheme has the problems that the cutting edge parameter setting lacks optical target guiding and only meets the feasibility of mechanical cutting, the end face quality assessment is delayed from the machining process, the end face quality assessment depends on offline optical detection and cannot be used for design iteration, and a computable mapping model from edge shape input to optical output is not constructed, so that reverse design taking maximum light transmission efficiency or minimum reflectivity as an optimization target cannot be realized. In application scenarios with high requirements on end face optical consistency, such as high-speed optical interconnection, quantum communication and high-power laser transmission, the defects limit the upper performance limit and the mass manufacturing yield of optical fiber devices, and an intelligent reverse design method integrating multiple physical field simulation is needed to realize accurate regulation and control of microscopic geometry of diamond blade edges and predictable guarantee of optical performance. Disclosure of Invention The invention provides a processing design method based on a diamond cutter for optical fiber cutting, and aims to solve the technical problem that the traditional diamond cutting edge design relies on experience and macroscopic mechanical analysis, and an accurate mapping relation between the cutting edge microscopic geometrical morphology and the optical performance of an optical fiber end face cannot be established. In the prior art, the cutting edge of the diamond cutter is generally simplified into an idealized straight line or circular arc profile, and the machining process mainly sets macroscopic parameters such as a rake angle, a relief angle, a cutting edge radius and the like according to the experience of an operator, and ignores the real three-dimensional morphological characteristics of the cutting edge under the nano scale, including factors such as the gradient of the cutting edge curvature, microscopic collapse, the deviation of the crystal orientation, the surface roughness distribution and the like. In the actual cutting process,