Accurate control method and device for axial grinding thickness of optical fiber

Abstract

The invention relates to the technical field of optical fibers, and discloses a precise control method and device for the axial grinding thickness of an optical fiber, and the method comprises the steps: carrying out the friction coefficient monitoring of the axial grinding process of the end face of the optical fiber, and obtaining a comprehensive friction coefficient value and a material layer recognition signal; the comprehensive friction coefficient value is compared with a coating layer friction coefficient threshold value, and a deceleration grinding execution instruction is generated; comparing the comprehensive friction coefficient value with a cladding friction coefficient threshold value, and generating a fine grinding execution instruction when the material layer identification signal indicates that the material layer enters the cladding and the comprehensive friction coefficient value reaches the cladding friction coefficient threshold value; the comprehensive friction coefficient value is compared with a core layer friction coefficient threshold value, when the material layer identification signal indicates that the material layer enters the core layer and the comprehensive friction coefficient value reaches the core layer friction coefficient threshold value, a target execution termination instruction is generated, triple condition synchronous judgment is established, and smooth transition and accurate control of the grinding process are achieved.

Description

Technical Field

[0001] The present invention relates to the field of optical fiber technology, and in particular to a method and device for accurately controlling the axial grinding thickness of an optical fiber. Background Art

[0002] Currently, nanometer-level processing accuracy is required for core exposure control, cladding thickness adjustment, and coating removal of single-mode optical fibers. Traditional optical fiber end-face polishing processes have technical problems such as insufficient polishing thickness control accuracy, difficulty in identifying material layer boundaries, and a high risk of over-polishing. Traditional polishing methods based on time control or displacement control cannot accurately identify the transition boundaries of the three-layer material structure of the optical fiber, which can easily cause over-polishing of the core layer or residual cladding, seriously affecting the optical performance and mechanical strength of the optical fiber. In addition, existing single-point monitoring technology is easily interfered with by local signal fluctuations and lacks comprehensive monitoring of the entire polishing area, making it difficult to achieve accurate identification and adaptive control of different material layers. Summary of the Invention

[0003] The present invention provides a method and device for accurately controlling the axial grinding thickness of an optical fiber. The present invention establishes a triple condition synchronous judgment to achieve a smooth transition and accurate control of the grinding process.

[0004] In a first aspect, the present invention provides a method for accurately controlling the axial grinding thickness of an optical fiber, the method comprising:

[0005] The friction coefficient of the optical fiber end face is monitored during the axial polishing process to obtain the comprehensive friction coefficient value and material layer identification signal;

[0006] Comparing the comprehensive friction coefficient value with a coating layer friction coefficient threshold value, and generating a deceleration grinding execution instruction when the comprehensive friction coefficient value reaches the coating layer friction coefficient threshold value;

[0007] comparing the comprehensive friction coefficient value with a cladding friction coefficient threshold, and generating a fine grinding execution instruction when the material layer identification signal indicates entering the cladding and the comprehensive friction coefficient value reaches the cladding friction coefficient threshold;

[0008] The comprehensive friction coefficient value is compared with a core layer friction coefficient threshold value, and a target termination execution instruction is generated when the material layer identification signal indicates entering the core layer and the comprehensive friction coefficient value reaches the core layer friction coefficient threshold value.

[0009] In combination with the first aspect, in a first implementation of the first aspect of the present invention, the friction coefficient is monitored during the axial polishing process of the optical fiber end face to obtain a comprehensive friction coefficient value and a material layer identification signal, including:

[0010] The three-axis force sensor is used to collect signals during the axial polishing process of the optical fiber end face to obtain radial friction force component signals and axial positive pressure signals;

[0011] Performing friction coefficient vector synthesis based on the radial friction force component signal and the axial normal pressure signal to obtain an instantaneous friction coefficient value;

[0012] performing noise elimination processing on the instantaneous friction coefficient value to obtain a denoised friction coefficient signal, and performing smoothing processing based on the denoised friction coefficient signal to obtain friction coefficient time series data;

[0013] The piezoelectric sensor is subjected to signal processing based on the friction coefficient time series data to obtain a comprehensive friction coefficient value and a material layer identification signal.

[0014] In combination with the first aspect, in a second implementation of the first aspect of the present invention, the signal processing of the piezoelectric sensor based on the friction coefficient time series data to obtain a comprehensive friction coefficient value and a material layer identification signal includes:

[0015] Through the multi-channel synchronous acquisition of the annular array composed of piezoelectric sensors, a distributed piezoelectric signal matrix is ​​obtained;

[0016] Performing weighted fusion on the friction coefficient time series data and the distributed piezoelectric signal matrix to obtain a comprehensive friction coefficient value, and performing frequency domain decomposition based on the distributed piezoelectric signal matrix to obtain spectrum energy distribution data;

[0017] The characteristic frequency bands of the material layers are identified based on the spectral energy distribution data. When the energy proportion of the first frequency band exceeds the first percentage, it is identified as a coating layer. When the energy proportion of the second frequency band exceeds the second percentage, it is identified as a cladding layer. When the energy proportion of the third frequency band exceeds the third percentage, it is identified as a core layer. A material layer identification signal is generated at the same time.

[0018] In combination with the first aspect, in a third implementation of the first aspect of the present invention, comparing the comprehensive friction coefficient value with a coating layer friction coefficient threshold, and generating a deceleration grinding execution instruction when the comprehensive friction coefficient value reaches the coating layer friction coefficient threshold, includes:

[0019] Performing sliding window statistics on the comprehensive friction coefficient value to obtain a sliding average value and a standard deviation value;

[0020] Calculate the dynamic friction coefficient threshold of the coating layer based on the sliding average value and the standard deviation value;

[0021] Comparing the comprehensive friction coefficient value with the coating layer dynamic friction coefficient threshold value, and generating a threshold trigger signal when the comprehensive friction coefficient value is greater than or equal to the coating layer dynamic friction coefficient threshold value;

[0022] The deceleration grinding parameters are calculated according to the threshold trigger signal, and the deceleration grinding execution instruction is generated by adopting S-curve control.

[0023] In combination with the first aspect, in a fourth implementation of the first aspect of the present invention, calculating the deceleration grinding parameters according to the threshold trigger signal and generating the deceleration grinding execution instruction using an S-shaped curve control include:

[0024] Calculating deceleration grinding parameters based on the threshold trigger signal to obtain a numerical combination of deceleration parameters;

[0025] Performing S-curve trajectory planning on the deceleration parameter value combination to obtain deceleration control trajectory data;

[0026] Performing real-time calculation, monitoring, and processing of the friction coefficient gradient based on the deceleration control trajectory data, calculating the friction coefficient change rate, and generating a material layer conversion warning signal when the friction coefficient gradient of the friction coefficient change rate exceeds a preset gradient threshold;

[0027] A deceleration grinding execution instruction including speed control, pressure control, feed control and early warning monitoring is constructed according to the deceleration control trajectory data and the material layer conversion early warning signal.

[0028] In combination with the first aspect, in a fifth implementation of the first aspect of the present invention, comparing the comprehensive friction coefficient value with a cladding friction coefficient threshold, and generating a fine grinding execution instruction when the material layer identification signal indicates entering the cladding and the comprehensive friction coefficient value reaches the cladding friction coefficient threshold, includes:

[0029] The comprehensive friction coefficient value, the friction coefficient change rate, the grinding depth data and the temperature compensation value are used as input parameters to predict the cladding friction coefficient through a three-layer feedforward neural network to obtain a predicted value of the cladding friction coefficient;

[0030] Performing adaptive compensation based on the predicted value of the cladding friction coefficient to obtain a cladding dynamic friction coefficient threshold;

[0031] Performing a cladding confirmation judgment on the material layer identification signal, and generating a cladding threshold trigger signal when the material layer identification signal indicates entering the cladding and the comprehensive friction coefficient value is greater than or equal to the cladding dynamic friction coefficient threshold;

[0032] Fine grinding parameters are calculated according to the cladding threshold trigger signal and a fine grinding execution instruction is generated in an intermittent working mode.

[0033] In combination with the first aspect, in a sixth implementation of the first aspect of the present invention, the step of calculating fine grinding parameters according to the cladding threshold trigger signal and generating fine grinding execution instructions in an intermittent working mode includes:

[0034] Performing numerical calculation of fine grinding parameters based on the cladding threshold trigger signal to obtain a fine grinding parameter combination;

[0035] Performing intermittent working sequence conversion on the fine grinding parameter combination to obtain an intermittent working control sequence;

[0036] integrating the friction coefficient of the cladding thickness based on the intermittent working control sequence to obtain a cladding thickness monitoring signal;

[0037] Based on the fine grinding parameter combination, the intermittent working control sequence and the cladding thickness monitoring signal, a fine grinding execution instruction including speed control, pressure control, micro-feed control, intermittent timing control and thickness monitoring is constructed.

[0038] In combination with the first aspect, in a seventh implementation of the first aspect of the present invention, comparing the comprehensive friction coefficient value with the core layer friction coefficient threshold, and generating a target termination execution instruction when the material layer identification signal indicates entering the core layer and the comprehensive friction coefficient value reaches the core layer friction coefficient threshold, includes:

[0039] Based on the benchmark friction coefficient of the quartz core layer, a multi-parameter coupling analysis is performed to obtain the threshold value of the dynamic friction coefficient of the core layer.

[0040] Inputting the comprehensive friction coefficient value into a cubic polynomial fitting algorithm for third-order prediction to obtain the second-order derivative of the friction coefficient and a prediction error value;

[0041] Performing a triple-condition synchronous judgment on the material layer identification signal, the comprehensive friction coefficient value, the second-order derivative of the friction coefficient, and the predicted error value, and generating a core layer threshold trigger signal when the material layer identification signal indicates entering the core layer and the comprehensive friction coefficient value is greater than or equal to the core layer dynamic friction coefficient threshold;

[0042] According to the core layer threshold trigger signal, the spindle is instantly stopped, the grinding head is quickly lifted, the coolant is immediately sprayed for cooling, and the triple safety protection mechanism of mechanical limit, electrical interlock and software monitoring is started to obtain the target termination execution instruction.

[0043] In combination with the first aspect, in an eighth implementation of the first aspect of the present invention, inputting the comprehensive friction coefficient value into a cubic polynomial fitting algorithm for third-order prediction to obtain a second-order derivative of the friction coefficient and a prediction error value includes:

[0044] Extracting time series sampling points of the comprehensive friction coefficient value to obtain a friction coefficient fitting data group;

[0045] Inputting the friction coefficient fitting data set into a cubic polynomial fitting algorithm to calculate coefficients, thereby obtaining a polynomial fitting coefficient combination including a cubic term coefficient, a quadratic term coefficient, a linear term coefficient, and a constant term;

[0046] Based on the polynomial fitting coefficient combination, the friction coefficient value is predicted by calculating the cubic polynomial function, and the second-order derivative value of the friction coefficient is obtained by performing a quadratic derivation on the polynomial function;

[0047] The absolute value difference between the predicted friction coefficient value and the comprehensive friction coefficient value is calculated to obtain a predicted error value.

[0048] In a second aspect, the present invention provides a device for precisely controlling the axial grinding thickness of an optical fiber, the device comprising:

[0049] A monitoring module is used to monitor the friction coefficient during the axial polishing process of the optical fiber end face and obtain a comprehensive friction coefficient value and a material layer identification signal;

[0050] a first comparison module, configured to compare the comprehensive friction coefficient value with a coating layer friction coefficient threshold value, and generate a deceleration grinding execution instruction when the comprehensive friction coefficient value reaches the coating layer friction coefficient threshold value;

[0051] a second comparison module, configured to compare the comprehensive friction coefficient value with a cladding friction coefficient threshold value, and generate a fine grinding execution instruction when the material layer identification signal indicates entering the cladding and the comprehensive friction coefficient value reaches the cladding friction coefficient threshold value;

[0052] The third comparison module is used to compare the comprehensive friction coefficient value with the core layer friction coefficient threshold, and generate a target termination execution instruction when the material layer identification signal indicates entering the core layer and the comprehensive friction coefficient value reaches the core layer friction coefficient threshold.

[0053] The technical solution provided by the present invention uses a three-axis force sensor to collect radial friction force components and axial normal pressure in real time, achieving vector synthesis calculation and noise elimination of the friction coefficient. Compared with traditional empirical judgment methods, it can objectively, continuously, and highly accurately reflect the changes in material properties during the grinding process. A distributed acquisition method using a ring array of piezoelectric sensors is used. A weighted fusion algorithm eliminates the influence of local signal fluctuations to obtain a more accurate comprehensive friction coefficient value. Frequency domain decomposition technology is used to automatically identify material layers, overcoming the limitations of single-point monitoring. Three different friction coefficient threshold judgment criteria are established for the coating layer, cladding layer, and core layer, achieving precise identification of material layer transitions and a layered control strategy, avoiding the problems of over-grinding or under-grinding caused by traditional unified control parameters. Dynamic correction of the friction coefficient threshold is achieved through sliding window statistical analysis, an exponentially weighted moving average algorithm, and a multi-parameter coupling model, effectively solving the problem of threshold offset caused by differences in material properties between different optical fiber batches. A three-layer feedforward neural network is used to predict the cladding friction coefficient. Combined with an adaptive compensation mechanism, this method can more accurately predict the timing of material layer transitions compared to simple threshold comparisons, allowing for early adjustment of control strategies. Differentiated control strategies (decelerated grinding, fine grinding, and immediate termination) are employed for different material layers. Through S-curve control and intermittent operation, a smooth transition and precise control of the grinding process are achieved. A triple-condition synchronous judgment (material layer confirmation, friction coefficient threshold, second-order derivative condition, and prediction error) and millisecond-level emergency braking response, combined with a triple safety protection mechanism, achieve the ideal goal of zero over-grinding. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0055] Figure 1 Schematic diagram of the steps of a method for accurately controlling the axial polishing thickness of an optical fiber according to an embodiment of the present invention;

[0056] Figure 2 Schematic diagram of the structure of the device for accurately controlling the axial grinding thickness of an optical fiber in an embodiment of the present invention. DETAILED DESCRIPTION

[0057] Embodiments of the present invention provide a method and apparatus for precisely controlling the axial polishing thickness of an optical fiber. The terms "first," "second," "third," "fourth," and so on (if any) in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or precedence. It should be understood that the numbers used in this way are interchangeable where appropriate, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product, or apparatus that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units that are not explicitly listed or that are inherent to these processes, methods, products, or apparatus.

[0058] For ease of understanding, the specific process of the embodiment of the present invention is described below. Figure 1 An embodiment of the method for accurately controlling the axial polishing thickness of an optical fiber according to an embodiment of the present invention includes:

[0059] Step S1, monitoring the friction coefficient during the axial polishing process of the optical fiber end face to obtain a comprehensive friction coefficient value and a material layer identification signal;

[0060] It is understandable that the execution subject of the present invention can be a precise control device for the axial polishing thickness of an optical fiber, or a terminal or a server, which is not limited here. The embodiment of the present invention is described by taking a server as the execution subject as an example.

[0061] Specifically, during the axial polishing of the optical fiber end face, the mechanical response of the polishing contact area is collected at high frequency. Based on a highly sensitive three-axis force sensor, the friction force component signal F from the radial direction is obtained in real time at a sampling frequency of 1000 Hz. x and F y And the positive pressure signal F in the axial direction zThe triaxial force data are synthesized based on the ratio between friction and normal pressure to obtain the instantaneous friction coefficient value at the corresponding time point. The instantaneous friction coefficient value is subjected to noise reduction to remove environmental noise and system measurement interference from the acquired signal. A high-pass filter is applied to remove low-frequency fluctuations, and a five-point moving average algorithm within a sliding window framework is used to smooth the instantaneous friction coefficient value in the time domain, generating a continuous and stable denoised friction coefficient signal stream. To enhance the friction characteristics' sensitivity to different material layers, the smoothed signal is input into the signal fusion unit of the piezoelectric sensor array. The array consists of 12 highly sensitive piezoelectric sensors evenly distributed around the polishing disc. Each sensor synchronously acquires microscopic vibration and stress fluctuation responses from the polished contact area of ​​the optical fiber end face at a frequency of 2000 Hz, forming a piezoelectric signal sampling matrix. The processing module uses a weighted fusion algorithm to fuse the outputs of each sensor with the real-time friction coefficient signal point by point to obtain a comprehensive friction coefficient value. The fusion weights are set based on the symmetry of the sensor distribution and the stability of the historical response. The fused comprehensive friction coefficient signal is input into the frequency domain analysis module, and its main frequency component is extracted using fast Fourier transform and compared with the material layer frequency feature database for judgment. When the main frequency of the signal is in the range of 300-500Hz and the energy accounts for more than 60%, it is identified as a coating layer; when the main frequency is in the range of 500-800Hz and the energy accounts for more than 70%, it is identified as a cladding layer; when the main frequency is in the range of 800-1200Hz and the energy accounts for more than 80%, it is determined to be a core layer, and finally the corresponding material layer identification signal is output.

[0062] Step S2: comparing the comprehensive friction coefficient value with the coating layer friction coefficient threshold value, and generating a deceleration grinding execution instruction when the comprehensive friction coefficient value reaches the coating layer friction coefficient threshold value;

[0063] Specifically, a continuous time window analysis and processing is performed on the comprehensive friction coefficient value. Within the set sampling period, a sliding window statistical operation is performed on the friction coefficient values ​​within the range of the latest 100 sampling points, and the sliding mean and standard deviation values ​​within the window are calculated in real time to evaluate the stability and fluctuation of the friction signal at the current stage. On this basis, combined with the historical benchmark experience value of the coating friction coefficient, a dynamic threshold adjustment model is constructed, that is, the static benchmark friction value is incrementally corrected using the sliding mean and standard deviation values ​​obtained by current statistics to form a dynamic friction coefficient threshold for the coating. The real-time updated comprehensive friction coefficient value is numerically compared with the dynamic friction coefficient threshold of the coating. When the comprehensive friction coefficient value is greater than or equal to the dynamic threshold, a threshold trigger signal is generated, and the signal is used as the starting condition for the deceleration control process. After the threshold trigger signal is activated, the system enters the deceleration grinding parameter calculation module, which executes a multi-dimensional parameter adjustment strategy based on the current grinding state parameters, including rotation speed, feed speed, grinding pressure and material layer state. To ensure the smoothness of the deceleration process and the stability of the fiber end face morphology, the control system uses an S-curve control method to dynamically interpolate and adjust the various grinding parameters, reducing the original grinding speed from 200rpm to 100rpm in a continuous and smooth manner, while reducing the applied pressure from 8N ​​to 5N, and synchronously adjusting the feed speed to 50% of the initial value. The S-curve adjustment has the characteristics of small acceleration in the initial stage, linear transition in the middle stage, and smooth convergence at the end, which effectively avoids mechanical shock or damage to the fiber end face structure caused by sudden parameter changes. In the deceleration control execution module, continuous grinding parameter adjustment instructions are output in real time according to the S-curve scheduling table, driving the actuator to complete the flexible transition from the normal grinding state to the deceleration grinding state, and continuously monitoring the change trend and gradient information of the comprehensive friction coefficient throughout the process to determine whether it is close to the cladding boundary.

[0064] Step S3: comparing the comprehensive friction coefficient value with the cladding friction coefficient threshold value, and generating a fine grinding execution instruction when the material layer identification signal indicates entering the cladding and the comprehensive friction coefficient value reaches the cladding friction coefficient threshold value;

[0065] Specifically, when the material layer recognition signal indicates the start of the cladding phase, four key physical parameters—the current comprehensive friction coefficient, friction coefficient change rate, grinding depth, and temperature compensation value—are extracted and passed as input vectors to a three-layer feedforward neural network model. This neural network, previously trained using multiple batches of experimental samples, outputs a predicted cladding friction coefficient value for the current operating conditions based on the nonlinear coupling relationship between the input variables. Adaptive compensation is applied based on the predicted cladding friction coefficient value to correct for deviations in the predicted value caused by external disturbances and changes in equipment status. This compensation mechanism makes multi-factor incremental adjustments based on real-time monitored temperature, pressure, and speed deviations to establish a dynamic cladding friction coefficient threshold. This threshold responds to subtle changes in the machining environment in real time within each control cycle, thereby enhancing judgment accuracy and control robustness. The material layer recognition signal is then verified for the cladding phase. When the material layer recognition signal clearly indicates cladding, the current comprehensive friction coefficient value is compared with the dynamic threshold. A cladding threshold trigger signal is generated when the comprehensive friction coefficient value satisfies the condition that it is greater than or equal to the dynamic threshold. After the trigger signal is generated, the fine grinding parameter calculation module is started. By calling the low-speed, low-pressure, and micro-feed strategy corresponding to the cladding stage, the grinding speed is set to 50 rpm, the contact pressure is set to 3N, and the feed amount is strictly controlled within 10 nanometers each time. At the same time, in order to avoid heat accumulation and structural fatigue caused by continuous contact, an intermittent working mode is adopted for control output, that is, a 0.1-second stop interval is added after each 0.5-second grinding cycle, and dynamic adjustment is made according to the local fluctuation of the comprehensive friction coefficient, thereby maintaining the removal rate while ensuring thermal stability and processing uniformity. The fine grinding execution instruction is thus generated and sent to the grinding actuator, which realizes high-precision and stable removal of the cladding material through the alternating operation mode of flexible feeding and periodic pause.

[0066] Step S4: Compare the comprehensive friction coefficient value with the core layer friction coefficient threshold value, and generate a target termination execution instruction when the material layer identification signal indicates entering the core layer and the comprehensive friction coefficient value reaches the core layer friction coefficient threshold value.

[0067] Specifically, a multi-parameter coupling analysis is conducted based on the core layer benchmark friction characteristics of the quartz material, and a comprehensive dynamic correction mechanism is established by combining the vibration frequency, system disturbance and friction fluctuation law in the actual grinding process. In this mechanism, based on the stable friction coefficient of the quartz core layer, a dynamic friction coefficient threshold is constructed by introducing a periodic fluctuation compensation factor and a random disturbance term. The continuously updated comprehensive friction coefficient value is input into the cubic polynomial fitting model, and a local trend function is constructed within the last 30 valid sampling points. Based on this function, a third-order prediction operation of the friction coefficient change trend is completed, and the second-order derivative of the friction coefficient at the current moment and the prediction error value are extracted from it, which are used to reflect the acceleration change characteristics of the friction response and the fitting accuracy between the model prediction and the true value. This prediction mechanism can dynamically track the sudden change trend of the friction signal and effectively compensate for the judgment lag caused by delayed response. The control system performs a triple simultaneous judgment on the material layer identification signal, the comprehensive friction coefficient value, the second-order derivative of the friction coefficient, and the predicted error value: confirming that the material layer identification signal indeed indicates that the core layer area has been entered; verifying that the current comprehensive friction coefficient value has reached or exceeded the core layer dynamic friction coefficient threshold established above; and judging whether the second-order derivative of the friction coefficient is significantly greater than the normal fluctuation level and that the predicted error is within a small range. When these three judgment conditions are met simultaneously, a core layer threshold trigger signal is generated. Based on the core layer threshold trigger signal, the termination control program is initiated. The spindle is instantaneously stopped to prevent continued force from causing microscopic cracking of the core layer. At the same time, the drive module is linked to quickly raise the grinding head 5 mm to quickly break contact with the fiber surface. The cooling system is also synchronously activated to spray coolant into the grinding area to quickly cool the surface, preventing deformation or thermal damage to the fiber end face microstructure due to local heat accumulation. To ensure full control execution in extreme or faulty situations, a triple safety mechanism of mechanical limits, electrical interlocks, and software monitoring is activated. Mechanical limits prevent overshooting, electrical interlocks ensure safe isolation of the equipment's power supply, and the software monitoring module continuously tracks the response status of each step in the termination process, ensuring complete execution within milliseconds and recording the execution log. Through these steps, the target termination execution command is ultimately obtained.

[0068] In a specific embodiment, the process of executing step S1 may specifically include the following steps:

[0069] The three-axis force sensor is used to collect signals during the axial polishing process of the optical fiber end face to obtain radial friction force component signals and axial positive pressure signals;

[0070] The friction coefficient vector is synthesized based on the radial friction force component signal and the axial positive pressure signal to obtain the instantaneous friction coefficient value;

[0071] Performing noise elimination processing on the instantaneous friction coefficient value to obtain a denoised friction coefficient signal, and performing smoothing processing based on the denoised friction coefficient signal to obtain friction coefficient time series data;

[0072] The piezoelectric sensor is processed based on the friction coefficient time series data to obtain the comprehensive friction coefficient value and material layer identification signal.

[0073] Specifically, a triaxial force sensor serves as the core sensing device, located in the contact support module of the polishing execution structure. This sensor senses the two radial friction components generated by the optical fiber end face during polishing, as well as the axial normal pressure signal applied along the main axis. The sampling frequency is set to 1000 Hz to ensure sufficiently detailed dynamic trends are captured. While the polishing disc drive mechanism rotates continuously and the optical fiber workpiece remains firmly clamped, the triaxial force sensor continuously outputs mechanical response signals. The system extracts the friction force components in the X and Y axes, as well as the axial normal pressure signal in the Z axis, and constructs these signals into a synchronized data set in a time series. By vector synthesis of the radial friction force component signals, a resultant force signal reflecting the total tangential sliding resistance of the contact surface is generated. This resultant force signal is then ratioed with the axial normal pressure signal to calculate the instantaneous friction coefficient at each sampling moment. This value directly represents the polishing state changes throughout the entire optical fiber polishing process. It is affected by the material layer type, contact state, feed depth, and temperature rise, exhibiting distinct transition characteristics during material layer transitions. The instantaneous friction coefficient values ​​are de-noised and fed into a digital filtering module. This module incorporates a high-pass filter to suppress low-frequency, slowly varying trends. A limiting filter strategy is also employed to remove sudden interference peaks, ensuring that the signal portion that truly reflects the grinding process is retained. Smoothing is then performed. A sliding average algorithm is used to calculate the mean of the noise-suppressed friction coefficient signal within a window. The window length is set to five consecutive sampling periods based on system response speed and control delay requirements. This local smoothing of the time series improves signal stability and discriminability while preserving the changing trend, generating a smooth and continuous friction coefficient time series data sequence. The friction coefficient time series data is then fed into a piezoelectric sensor array located on the periphery of the grinding structure. The array consists of 12 piezoelectric sensors evenly distributed around the grinding disc to provide spatial coverage of the entire contact surface state. Each piezoelectric sensor independently samples the local micro-vibration response at a frequency of 2000 Hz, generating an independent signal channel. The output is standardized by a conditioning circuit and then aggregated into the data fusion module. In this module, the system uses a channel matching mechanism to align each piezoelectric signal with the current friction coefficient time series data. Then, a weighted fusion strategy is used to perform a time-domain weighted superposition of the 12 signals. Weights are preset or dynamically updated online based on the symmetry of the sensors in the array, historical noise levels, and thermal stability. The fusion output is the combined friction coefficient value for a single path. The fused signal is then subjected to fast Fourier transform analysis to extract its dominant frequency component and calculate the energy distribution within each characteristic frequency band.Based on the friction response frequency characteristics of different optical fiber material layers during polishing, the system sets the characteristic frequency bands of 300–500 Hz for the coating layer, 500–800 Hz for the cladding layer, and 800–1200 Hz for the core layer, with energy thresholds set at 60%, 70%, and 80%, respectively. When the energy percentage of the spectrum of the comprehensive friction coefficient signal exceeds the corresponding threshold in a certain frequency band, the material layer is identified, and the system outputs a material layer identification signal accordingly.

[0074] In a specific embodiment, the step of performing signal processing on the piezoelectric sensor based on the friction coefficient time series data to obtain the comprehensive friction coefficient value and the material layer identification signal may specifically include the following steps:

[0075] Through the multi-channel synchronous acquisition of the annular array composed of piezoelectric sensors, a distributed piezoelectric signal matrix is ​​obtained;

[0076] The friction coefficient time series data and the distributed piezoelectric signal matrix are weightedly fused to obtain the comprehensive friction coefficient value, and the frequency domain decomposition is performed based on the distributed piezoelectric signal matrix to obtain the spectrum energy distribution data;

[0077] The characteristic frequency bands of the material layer are identified based on the spectral energy distribution data. When the energy proportion of the first frequency band exceeds the first percentage, it is identified as a coating layer. When the energy proportion of the second frequency band exceeds the second percentage, it is identified as a cladding layer. When the energy proportion of the third frequency band exceeds the third percentage, it is identified as a core layer. At the same time, a material layer identification signal is generated.

[0078] Specifically, a multi-point synchronous data acquisition platform based on piezoelectric sensors was constructed. This platform utilizes a circular array structure, with 12 highly sensitive piezoelectric sensors evenly distributed around the periphery of the grinding disc. This ensures that the sensors are spatially and angularly spaced, forming a signal acquisition topology that covers the entire grinding contact area. Each sensor has an independent data channel and sampling circuit, and samples the local mechanical vibration signal in real time at a high frequency of 2000 Hz. The resulting signal not only reflects the microscopic energy perturbations in the local grinding zone but also strongly correlates with the friction characteristics generated on the fiber surface during the grinding contact process. The system utilizes a multi-channel synchronization control unit to centrally schedule and timestamp these 12 channels, generating a distributed piezoelectric signal matrix with the data structure S(i,t), where i represents the channel number and t represents the time dimension. Friction coefficient time series data is fused with this piezoelectric signal matrix. A channel-by-channel weighted calculation is performed based on the friction coefficient value at each moment and the corresponding piezoelectric signal. The fusion weights are set and dynamically adjusted based on the spatial symmetry of the sensors in the array, historical response stability, and channel noise levels. After completing the weighted fusion of each channel, a single-channel comprehensive friction coefficient value is formed through accumulation. At the same time, the distributed piezoelectric signal matrix is ​​subjected to frequency domain transformation processing. The fast Fourier transform algorithm is used to perform spectral decomposition on each signal channel to obtain the amplitude spectrum distribution of each channel within the entire sampling window. The spectrum results of all channels are then weighted and accumulated point by point to obtain the integrated spectrum energy distribution data. This spectrum energy data is used to characterize the energy intensity distribution of micro-vibration signals during the grinding process in different frequency ranges. It is related to the structure, hardness, and dielectric properties of the grinding material. Different materials will show energy concentration at characteristic frequencies under the same grinding state. Based on this spectrum energy distribution data, a material layer frequency band characteristic discrimination model is introduced. According to the friction vibration response laws of three typical material layers in the optical fiber multilayer structure, three frequency characteristic segments are delineated: the first frequency segment is 300Hz to 500Hz, the second frequency segment is 500Hz to 800Hz, and the third frequency segment is 800Hz to 1200Hz. Based on the previous offline modeling and actual calibration results, the system sets the first frequency band to be judged as a coating layer when the energy proportion exceeds 60%, the second frequency band to be judged as a cladding layer when the energy proportion exceeds 70%, and the third frequency band to be judged as a core layer when the energy proportion exceeds 80%. In actual implementation, the spectrum energy distribution data is segmented and integrated, and the total energy value in the first, second, and third frequency bands is calculated respectively. The ratio operation is performed with the sum of the energy of all frequency bands to obtain the energy proportion of each frequency band, and then the proportion is compared with the set percentage threshold. If the energy proportion of a certain frequency band currently exceeds the corresponding threshold, it is considered that the characteristics of this frequency band are dominant, and it is judged that the material layer currently being ground belongs to the material type corresponding to this frequency band.When the recognition logic is completed, the system outputs a unique material layer identification signal in real time. The signal value reflects the current layer status, that is, it corresponds to the coating layer, cladding layer or core layer respectively.

[0079] In a specific embodiment, the process of executing step S2 may specifically include the following steps:

[0080] Perform sliding window statistics on the comprehensive friction coefficient value to obtain the sliding mean and standard deviation value;

[0081] The dynamic friction coefficient threshold of the coating layer is calculated based on the sliding mean and standard deviation values;

[0082] Compare the comprehensive friction coefficient value with the dynamic friction coefficient threshold of the coating layer, and generate a threshold trigger signal when the comprehensive friction coefficient value is greater than or equal to the dynamic friction coefficient threshold of the coating layer;

[0083] The deceleration grinding parameters are calculated according to the threshold trigger signal, and the deceleration grinding execution instruction is generated by adopting S-curve control.

[0084] Specifically, the comprehensive friction coefficient value is subjected to sliding window statistical processing to extract in real time the statistical characteristics that reflect the fluctuation trend of the current contact state. A sliding time window of fixed length is used for data collection, and the window is set to contain 100 continuous sampling points. In each sliding cycle, the sliding average value in the window is calculated to characterize the central trend of the friction coefficient under the current grinding state. At the same time, the standard deviation value of the window data is obtained to characterize the fluctuation intensity and variation amplitude of the friction signal. Based on the above sliding average and standard deviation values, a dynamic threshold model is constructed according to the friction behavior characteristics of the coating material, that is, a real-time statistical correction term is introduced on the basis of the coating baseline friction coefficient, and the standard deviation value is superimposed on the sliding mean as a dynamic disturbance amplitude indicator to form a coating dynamic friction coefficient threshold with immediate adaptability. This construction method can effectively offset the short-period friction coefficient offset caused by external factors such as material non-uniformity, micro-wear of grinding tools, and temperature fluctuations. The coating layer's dynamic threshold is compared with the comprehensive friction coefficient on a cycle-by-cycle basis. When the comparison indicates that the current comprehensive friction coefficient is greater than or equal to the dynamic threshold, the system generates a threshold trigger signal, indicating that the coating layer is approaching or entering a critical state, initiating a deceleration control procedure to prevent over-grinding or end-face structural disturbance. This threshold trigger signal serves as a key input to the control system. Within the next control cycle after the signal is generated, the system immediately discontinues the existing grinding state parameter maintenance strategy and transitions to the deceleration parameter calculation process. During this deceleration parameter calculation, the system utilizes a pre-set coating layer deceleration control reference template from the grinding process database. This template defines the speed change curve, pressure change curve, and feed rate adjustment curve from the current standard grinding state to the target deceleration state. To avoid mechanical shock or system oscillation caused by sudden parameter changes, the control system uses an S-shaped curve as the core variation model. Its numerical characteristics of "slow start and stop, linear mid-section" ensure a continuous and smooth deceleration transition. The grinding spindle speed is gradually reduced from an initial 200 rpm to 100 rpm, the contact pressure is gradually reduced from 8 N to 5 N, and the feed rate is simultaneously reduced to 50% of its original value. The entire parameter change process is divided into the initial acceleration section, the uniform speed adjustment section and the terminal deceleration section, which correspond to the three control intervals of the S-shaped curve respectively. Each control point is time-calibrated and closed-loop verified with the physical feedback signal to ensure that the equipment response is not delayed or out of step. In order to enhance the safety and robustness of the deceleration process, the system introduces an online correction mechanism in the S-shaped curve generation stage, that is, to make slight offset corrections to the target control curve based on the real-time change trend of the friction coefficient gradient in the current grinding process. For example, when it is detected that the gradient of the friction coefficient change is greater than 0.01 / s, the system actively lengthens the adjustment interval of the S-shaped curve to adapt to the sudden fluctuation, and dynamically updates the adjustment amplitude according to the first derivative information in the curve deduction process to achieve control of the system behavior under critical state.The generated S-shaped control curve is decomposed into a set of continuous and discrete parameter execution instructions, which are transmitted to the spindle motor, pressure controller and feed drive through the servo control module respectively, completing the parameter slow-changing scheduling while ensuring the coordinated operation of the entire system.

[0085] In a specific embodiment, the execution step calculates the deceleration grinding parameters according to the threshold trigger signal, and the process of generating the deceleration grinding execution instruction using the S-curve control can specifically include the following steps:

[0086] Calculate the deceleration grinding parameters based on the threshold trigger signal to obtain a numerical combination of the deceleration parameters;

[0087] Perform S-curve trajectory planning on the deceleration parameter value combination to obtain deceleration control trajectory data;

[0088] Based on the deceleration control trajectory data, the friction coefficient gradient is calculated and monitored in real time to calculate the friction coefficient change rate. When the friction coefficient gradient of the friction coefficient change rate exceeds a preset gradient threshold, a material layer conversion warning signal is generated;

[0089] According to the deceleration control trajectory data and the material layer conversion warning signal, a deceleration grinding execution instruction including speed control, pressure control, feed control and warning monitoring is constructed.

[0090] Specifically, the system calculates deceleration polishing parameters based on a threshold trigger signal to determine the optimal parameter combination for the final coating or material transition phase. This parameter combination encompasses three core variables: spindle speed, contact pressure, and feed rate. The system compares the current machine operating state with historical data templates and references typical deceleration parameter solutions for different fiber materials in a process library. The system then outputs a set of deceleration parameter values, including a reduction in spindle speed from 200 rpm to 100 rpm, contact pressure from 8 N to 5 N, and feed rate from its original value to 50%. Each parameter transition must be flexible without compromising the machine's mechanical stability, ensuring that the polishing head does not impact or erroneously wear the endface structure before entering the cladding. This deceleration parameter combination is then fed into the trajectory planning module, where an S-curve generator is used to generate trajectory models. This ensures that the timeline evolution of each control variable follows a velocity profile that follows a "slow start, steady middle, slow end" pattern. By setting the starting state, target state, and time span, the S-curve trajectory planner smoothly interpolates parameter changes using second-order derivative continuity, thus avoiding physical shock and control distortion caused by sudden or sudden jumps. During the trajectory planning process, the system sets separate S-shaped transition curves for the spindle speed, pressure output, and feed drive, and performs synchronous alignment processing on the time axis to form a multi-channel deceleration control trajectory data set for joint control. While executing the deceleration trajectory, in order to monitor the evolution trend of the material layer state in real time, the system performs high-frequency gradient calculation processing on the comprehensive friction coefficient signal, synchronizes the friction coefficient time series data with the trajectory time node, and performs dynamic calculation of the friction coefficient change rate at each trajectory sampling point. This change rate, as a derivative indicator of the friction coefficient, can capture sudden changes in friction behavior caused by differences in structure and physical properties between material layers. The system continuously calculates the difference in friction coefficient between adjacent time points and divides it by the sampling interval to form a friction coefficient change rate flow, and sets a preset gradient threshold, such as 0.01 / s, to determine whether there is a significant gradient jump in the current state. When the rate of change of the friction coefficient exceeds the threshold within the deceleration control trajectory, it indicates that the material layer is undergoing a transition from the coating layer to the cladding layer. The system generates a material layer transition warning signal and binds the signal to the current trajectory state. On this basis, the system integrates the deceleration control trajectory data with the material layer transition warning signal to construct a set of deceleration grinding execution instructions with joint control and forward-looking response characteristics. The instruction structure consists of four parts: a speed control sub-instruction, which defines the time series scheduling value of the spindle speed during the deceleration process; a pressure control sub-instruction, which is used to adjust the output voltage or pneumatic value of the downforce during the deceleration phase; a feed control sub-instruction, which specifies the pulse drive logic for the feed mechanism to gradually decelerate along an S-shaped trajectory; and a warning monitoring sub-instruction, which continuously detects whether the friction coefficient gradient meets the transition conditions during execution. Once a warning signal is generated, the system enters the data buffering and process preparation state.

[0091] In a specific embodiment, the process of executing step S3 may specifically include the following steps:

[0092] The comprehensive friction coefficient value, friction coefficient change rate, grinding depth data and temperature compensation value are used as input parameters to predict the cladding friction coefficient through a three-layer feedforward neural network to obtain the predicted value of the cladding friction coefficient.

[0093] Adaptive compensation is performed based on the predicted value of the cladding friction coefficient to obtain the cladding dynamic friction coefficient threshold;

[0094] Performing cladding confirmation judgment on the material layer identification signal, generating a cladding threshold trigger signal when the material layer identification signal indicates entering the cladding and the comprehensive friction coefficient value is greater than or equal to the cladding dynamic friction coefficient threshold;

[0095] Fine grinding parameters are calculated according to the cladding threshold trigger signal and fine grinding execution instructions are generated in an intermittent working mode.

[0096] Specifically, the control unit feeds multi-source data into a pre-trained three-layer feedforward neural network to intelligently predict the evolution of the cladding's friction characteristics. This neural network uses four key input variables as its foundation: a real-time, updated composite friction coefficient value, the friction coefficient change rate, the current grinding depth data, and a temperature compensation value. These parameters reflect the material contact state, friction fluctuation trends, feed control history, and thermodynamic interference factors, respectively. Together, they form a multidimensional description vector of the cladding's friction state, which is fed into the neural network's input layer nodes. The neural network's hidden layer consists of two nonlinear mapping layers, each containing several neuron nodes. The weights and activation functions between these nodes are trained and optimized using historical process data sets to ensure a high degree of accuracy in predicting the cladding's friction characteristics under varying environmental conditions. The output layer consists of a single node, which outputs the system's current prediction of the cladding's friction coefficient. This value structurally combines the ability to perceive the current friction state with the ability to predict short-term future evolution trends. Based on the predicted value of the cladding friction coefficient, adaptive compensation processing is performed in combination with the current process disturbance information to form a more adaptable cladding dynamic friction coefficient threshold. This compensation process uses temperature deviation, pressure deviation and feed speed deviation as input factors, which correspond to the difference between the reference value and the feedback information of the temperature sensor, pressure regulating device and feed controller respectively. The system performs weighted superposition of these deviation terms according to the multiple coefficient weights set in the compensation model, and adds them to the predicted value to form a dynamic threshold. The final confirmation and judgment of the cladding state is made in combination with the material layer identification signal. The material layer identification signal is derived from the energy analysis conclusion of the piezoelectric array signal by the frequency domain feature discrimination module. When the identification result clearly indicates that the grinding head has entered the cladding stage, the system compares the current comprehensive friction coefficient value with the aforementioned constructed cladding dynamic friction coefficient threshold cycle by cycle, and generates a cladding threshold trigger signal when the friction coefficient value is greater than or equal to the threshold. According to the cladding threshold trigger signal, the system calls the fine grinding parameter calculation module, which sets a set of stable parameter values ​​for low speed, low pressure, and micro-feed based on the cladding material's sensitivity to thermal shock, structural integrity, and thickness uniformity. The grinding speed is controlled below 50 rpm, the contact pressure is adjusted to 3 N, and the feed rate is limited to a micro-step value of 10 nm. At the same time, in order to avoid changes in the structural performance of the material layer due to heat accumulation under continuous contact conditions, the system introduces an intermittent working mode for control rhythm scheduling. After every 0.5 seconds of continuous work, the grinding is forced to be interrupted for 0.1 seconds to achieve heat dissipation buffering and structural release. The intermittent rhythm design reduces the local heat accumulation rate and slows down the thermal diffusion rate, thereby improving the local stability of the cladding material and the final thickness consistency.The system encapsulates the execution structure consisting of the above-mentioned fine grinding parameters and working rhythm into fine grinding execution instructions, and sends them to the spindle drive, voltage and pressure regulator and feed system through the controller respectively. At the same time, a feedback closed-loop adjustment mechanism based on the material layer status and friction trend is set in the control logic.

[0097] In a specific embodiment, the execution step of calculating fine grinding parameters according to the cladding threshold trigger signal and generating fine grinding execution instructions in an intermittent working mode may specifically include the following steps:

[0098] Numerical calculation of fine grinding parameters is performed based on the cladding threshold trigger signal to obtain a fine grinding parameter combination;

[0099] Perform intermittent working sequence conversion on the fine grinding parameter combination to obtain an intermittent working control sequence;

[0100] Based on the intermittent working control sequence, the friction coefficient of the cladding thickness is integrated to obtain the cladding thickness monitoring signal;

[0101] Based on the fine grinding parameter combination, intermittent working control sequence and cladding thickness monitoring signal, a fine grinding execution instruction including speed control, pressure control, micro feed control, intermittent timing control and thickness monitoring is constructed.

[0102] Specifically, a fine grinding parameter numerical calculation module is activated based on the current material identification status and a historical process parameter library. This module comprehensively considers multiple indicators, including the stability level of the current comprehensive friction coefficient value, the grinding depth range, material physical properties, and the temperature compensation coefficient. Using a multi-factor mapping relationship or parameter matching model, it outputs the optimal fine grinding parameter combination, which includes factors such as spindle speed, contact pressure, microfeed increment, and cycle time. In a typical setup, the system sets the spindle speed to below 50 rpm, adjusts the contact pressure to 3 N, and fixes the microfeed increment to 10 nm. A basic process rhythm is established with a continuous operation sequence of 0.5 seconds and a 0.1 second pause to ensure effective heat release and suppress local stress concentrations caused by continuous loading. This set of static parameters is converted into an intermittent operation control sequence with a time-rhythm control structure. This sequence includes the time scheduling order of the fine grinding parameters and the startup control logic. It also integrates a synchronous execution logic correction mechanism between subsystems to ensure precise temporal synchronization of spindle deceleration, pressure ramping, feed positioning, and intermittent stops during execution. In this control sequence, each periodic segment is divided into a startup phase, a stable phase, and an interrupt phase. The system generates a slowly varying speed and pressure curve for the startup phase using an S-shaped curve or linear interpolation method. The stable phase maintains constant parameter output, while the interrupt phase uses control logic to force all mechanical actuators into a pause or hold state. In intermittent operation, the system synchronizes the real-time acquired comprehensive friction coefficient value with the current control rhythm and integrates the friction coefficient value and the duration of each effective grinding period to continuously monitor the cladding thickness. This friction coefficient integration method, based on a calibrated empirical conversion coefficient, converts friction behavior per unit time into a corresponding amount of material removal. Through continuous integration, a cladding thickness evolution trajectory is constructed. The system updates the accumulated thickness information once during each complete intermittent cycle, generating a cladding thickness monitoring signal. This signal dynamically evolves as the grinding process progresses, numerically reflecting the current remaining or removed cladding thickness. The fine grinding parameter combination, intermittent working control sequence and cladding thickness monitoring signal are jointly input into the instruction construction module, in which the five-dimensional fine grinding execution instruction encapsulation including speed control, pressure control, micro-feed control, intermittent timing control and thickness monitoring is completed.The speed control part defines the slow-changing curve of the spindle speed in the startup section, the constant value in the stable section, and the stagnation logic in the interruption section; the pressure control part sets the downward force parameters that the grinding head needs to maintain in different stages, and introduces adjustment gain to cope with the deviation of the real-time feedback signal; the micro-feed control part adopts a stepping drive method to limit the feed displacement to the order of 10nm, and ensures that the positioning error is maintained within the range of ±1nm through closed-loop feedback; the intermittent timing control part uses the hardware timer and software control logic to ensure that the grinding execution action and the pause action alternate at a rhythm of 0.5 seconds and 0.1 seconds; and the thickness monitoring part continuously receives the friction integral accumulation signal, dynamically evaluates the cladding thickness change trend, and issues a process warning signal before approaching the core layer boundary.

[0103] In a specific embodiment, the process of executing step S4 may specifically include the following steps:

[0104] Based on the benchmark friction coefficient of the quartz core layer, a multi-parameter coupling analysis is performed to obtain the threshold value of the dynamic friction coefficient of the core layer.

[0105] The comprehensive friction coefficient value is input into the cubic polynomial fitting algorithm for third-order prediction, and the second-order derivative of the friction coefficient and the prediction error value are obtained;

[0106] Perform triple conditional synchronous judgment on the material layer identification signal, comprehensive friction coefficient value, second-order derivative of friction coefficient and prediction error value. When the material layer identification signal indicates entering the core layer and the comprehensive friction coefficient value is greater than or equal to the core layer dynamic friction coefficient threshold, a core layer threshold trigger signal is generated.

[0107] According to the core layer threshold trigger signal, the spindle is instantly stopped, the grinding head is quickly lifted, and the coolant is immediately sprayed for cooling. The triple safety protection mechanism of mechanical limit, electrical interlock, and software monitoring is activated to obtain the target termination execution instruction.

[0108] Specifically, the system continuously tracks the evolution of the comprehensive friction coefficient at the end of the cladding phase. Based on the physical properties of the quartz core, it activates a multi-parameter coupling analysis module to construct a dynamic threshold generation mechanism for core identification. This mechanism uses the friction stability, processing response sensitivity, thermal expansion characteristics, and high-frequency micro-vibration behavior of the quartz material as reference indicators. The baseline friction coefficient is used as a central constant. A periodic perturbation compensation factor and a random perturbation term corresponding to high-frequency noise within the system are introduced. A dynamic evolution function is established within the parameter space, resulting in a dynamic core friction coefficient threshold that is adjusted in real time with time and system state changes. This threshold is a dynamic function output based on the current system sampling frequency, the periodic characteristics of the friction signal, and the number of sampling points, supported by a specific mathematical model. This function aims to adapt to process sensitivity drift caused by different fiber batches, varying ambient temperatures, and changes in mechanical precision, thereby enhancing the robustness and real-time performance of core identification. The system also performs a cubic polynomial fit on the comprehensive friction coefficient values ​​over the most recent period. The system selects the most recent 30 valid sampling points as a fitting window, constructing a cubic curve with time as the independent variable and friction coefficient as the dependent variable. The curve is fitted using the least squares method, yielding the derivative information of the fitted curve. The system directly calculates the second-order derivative of this cubic curve, using it as a quantitative representation of the acceleration of the friction coefficient trend. The system also calculates the absolute difference between the real-time friction coefficient data point and the theoretical value of the fitted curve at the current moment, generating a prediction error index. These two values, when used together, determine whether the system is experiencing a rapid change trend and how closely the fitted model approximates the real data. A significantly positive second-order derivative indicates an upward bend in the friction coefficient curve, meaning the system is approaching or has already entered the core layer. Extremely small prediction errors indicate that the data trend is accurately captured by the model, indicating high system stability and enabling deterministic judgment. The system then simultaneously matches and judges the four parameters—the material layer identification signal, the comprehensive friction coefficient value, the second-order derivative of the friction coefficient, and the prediction error—using a triple-conditional approach. The system confirms whether the material layer identification signal has marked the current state as the core layer. The signal is based on the frequency domain analysis module to identify the high-frequency energy ratio of the piezoelectric array signal. When the energy ratio of the 800Hz to 1200Hz frequency band exceeds 80%, it is marked as the core layer stage; judge whether the comprehensive friction coefficient value has reached or exceeded the core layer friction coefficient threshold dynamically generated above, so as to confirm whether the friction behavior of the material surface conforms to the core layer characteristic performance; check whether the second-order derivative of the friction coefficient is greater than the set threshold (such as 0.05 / s 2), and at the same time, whether the prediction error is lower than the tolerance limit (such as 0.001). When all three of the above conditions are met at the same time, a core layer threshold trigger signal is generated, indicating that the system has fully identified the entry into the core layer area and the friction behavior is highly consistent with expectations. The termination control strategy needs to be immediately activated to avoid over-grinding behavior beyond the design thickness range. According to the core layer threshold trigger signal, the system enters the emergency control stage and immediately issues a termination execution instruction. In terms of control logic, this instruction is composed of a coordinated response structure composed of spindle control, feed mechanism control, cooling system control and multiple protection mechanisms. The system issues an instantaneous stop instruction through the spindle driver, requiring the spindle to completely reduce the current speed to 0 within no more than 0.1 seconds, and adjusts the braking force through the inertia compensation algorithm to avoid end face micro-cracks caused by high inertia impact. The system also issues a lifting instruction to the actuator, driving the grinding head to quickly lift 5mm within 10ms to achieve complete separation between the grinding medium and the fiber end face, thereby preventing mechanical residual pressure from forming additional scratches or particle embedding on the material surface. At the same time, the cooling system immediately activates the liquid injection module, spraying coolant into the grinding area through an independent circuit. This rapidly reduces the end-face temperature, suppresses the thermal expansion of the fiber structure caused by heat diffusion, and ensures that the material stress state at the termination point remains frozen within a stable range, thereby improving the repeatability of the final thickness position. To ensure that termination control is completed in a very short time with high safety and fault self-recovery, the system simultaneously activates a triple safety protection mechanism. The first is a mechanical limit protection mechanism, which sets an upper limit on the grinding head travel and forcibly disconnects the power supply if any lift command exceeds a safe position. The second is an electrical interlock mechanism. If the spindle is not stopped or the feed is not retracted, the system prohibits other modules from continuing to execute, avoiding command conflicts caused by cross-commands. The third is a software monitoring module, which continuously compares the response time, action status, and execution path of each execution unit in real time in the background. If any delay, error, or conflict occurs, the fault handling procedure is immediately activated, logging the log and switching to a protection state. Through these steps, the target termination execution instruction is finally obtained.

[0109] In a specific embodiment, the step of inputting the comprehensive friction coefficient value into a cubic polynomial fitting algorithm for third-order prediction to obtain the second-order derivative of the friction coefficient and the prediction error value may specifically include the following steps:

[0110] Extract the time series sampling points of the comprehensive friction coefficient value to obtain the friction coefficient fitting data set;

[0111] Inputting the friction coefficient fitting data set into a cubic polynomial fitting algorithm to calculate coefficients, and obtaining a polynomial fitting coefficient combination including cubic term coefficients, quadratic term coefficients, linear term coefficients and a constant term;

[0112] Based on the combination of polynomial fitting coefficients, the friction coefficient value is predicted by calculating the cubic polynomial function, and the second-order derivative value of the friction coefficient is obtained by taking the second derivative of the polynomial function;

[0113] The absolute difference between the predicted friction coefficient value and the comprehensive friction coefficient value is calculated to obtain the prediction error value.

[0114] Specifically, based on a set time window parameter, a fixed-length sequence of sampling points is extracted from the original time series of the comprehensive friction coefficient to form a friction coefficient fitting data set. This data set contains the most recent 30 valid sampling points, arranged in chronological order. The friction coefficient value corresponding to each time point serves as the dependent variable, and the corresponding sampling time is used as the independent variable to form a data pair. The friction coefficient fitting data set is then input into a cubic polynomial fitting algorithm module to calculate the fitting coefficients. This algorithm uses the least squares method as an optimization criterion to construct an objective function. The optimization objective is the sum of the squared errors between the fitted curve and the actual data. The algorithm numerically solves for the four unknown parameters: the cubic term, the quadratic term, the linear term, and the constant term. By establishing a fitting matrix for the sampling points and performing linear algebraic operations, a set of polynomial fitting coefficients is output, including the cubic, quadratic, linear, and constant terms. Based on these polynomial fitting coefficients, a cubic polynomial function is used to calculate the predicted friction coefficient value. In this function structure, the time variable is used as the input, and the independent variable interval is set to the coverage of the fitting data set. The system calculates the polynomial function value for each target time point, outputting the corresponding predicted friction coefficient value. This predicted value reflects the evolution trajectory of the friction coefficient based on the existing trend and provides a reference for dynamic approximation of the termination control point. Furthermore, to obtain the acceleration characteristics of the current friction state—that is, the rate of change of the friction coefficient—the constructed cubic polynomial function is quadratically differentiated. The derivative function is solved analytically and the current time value is input to obtain the second-order derivative of the friction coefficient, reflecting the "curvature" of the friction change. When the value exceeds a certain threshold, it indicates that the friction coefficient is about to increase significantly. The predicted friction coefficient value is subtracted from the actual comprehensive friction coefficient value at the same time, and absolute value processing is used to eliminate deviations caused by positive and negative directions to obtain the prediction error. This error value measures the accuracy of the fitting model, namely, the accuracy of the currently constructed cubic polynomial in representing the actual evolution trend of the friction coefficient. When the error value is sufficiently small, for example, below the error tolerance of 0.001, it indicates that the model reliably reflects the current state characteristics and the prediction result is credible, which can be used to trigger the control strategy. Conversely, if the error is large, the system updates the fitting data set and recalculates the coefficients to ensure sufficient engineering reliability of the prediction mechanism.

[0115] In this embodiment, it also includes a particle swarm optimization multi-parameter collaborative adjustment and early warning prediction processing step: the deceleration grinding execution instruction, the fine grinding execution instruction and the target termination execution instruction are subjected to parameter extraction and integration processing, and a 12-key parameter combination including a three-level friction coefficient threshold, three deceleration ratio coefficients, three feed speed parameters and three pressure adjustment coefficients is constructed, and a multi-objective optimization function is established based on the accuracy index, efficiency index and safety index; the 12 key parameter combinations are input into the improved particle swarm optimization algorithm for global optimization processing, the particle motion trajectory is adjusted by the adaptive inertia weight and the dynamic learning factor, and the multi-objective optimization function is used for adaptive optimization. The system performs a degree evaluation to obtain the global optimal control parameter configuration; based on the global optimal control parameter configuration, the long short-term memory network is trained and processed, a friction coefficient gradient change rate prediction model is constructed, and the friction coefficient gradient of the next 10 sampling points is predicted and calculated to obtain gradient change rate prediction data; a three-level warning judgment processing is performed based on the gradient change rate prediction data, and a first-level warning signal is generated when the gradient change rate exceeds 0.01 / s, a second-level warning signal is generated when the gradient change rate exceeds 0.05 / s, and a third-level warning signal is generated when the gradient change rate exceeds 0.1 / s, and the emergency termination preparation is started 0.2 seconds in advance to obtain a warning control instruction.

[0116] The above describes the precise control method of the axial grinding thickness of the optical fiber in the embodiment of the present invention. The following describes the precise control device of the axial grinding thickness of the optical fiber in the embodiment of the present invention. Figure 2 In one embodiment of the present invention, a device for accurately controlling the axial polishing thickness of an optical fiber includes:

[0117] A monitoring module is used to monitor the friction coefficient during the axial polishing process of the optical fiber end face and obtain a comprehensive friction coefficient value and a material layer identification signal;

[0118] A first comparison module is used to compare the comprehensive friction coefficient value with the coating layer friction coefficient threshold value, and generate a deceleration grinding execution instruction when the comprehensive friction coefficient value reaches the coating layer friction coefficient threshold value;

[0119] A second comparison module is configured to compare the comprehensive friction coefficient value with a cladding friction coefficient threshold value, and generate a fine grinding execution instruction when the material layer identification signal indicates that the material layer has entered the cladding and the comprehensive friction coefficient value reaches the cladding friction coefficient threshold value;

[0120] The third comparison module is used to compare the comprehensive friction coefficient value with the core layer friction coefficient threshold, and generate a target termination execution instruction when the material layer identification signal indicates entering the core layer and the comprehensive friction coefficient value reaches the core layer friction coefficient threshold.

[0121] Through the collaborative efforts of these components, a triaxial force sensor collects radial friction and axial normal pressure in real time, enabling vector synthesis calculation and noise elimination of the friction coefficient. Compared to traditional empirical methods, this approach objectively, continuously, and accurately reflects changes in material properties during the polishing process. A distributed acquisition method utilizing a circular array of piezoelectric sensors uses a weighted fusion algorithm to eliminate the influence of local signal fluctuations, resulting in a more accurate composite friction coefficient value. Frequency domain decomposition technology enables automatic identification of material layers, overcoming the limitations of single-point monitoring. Establishing three distinct friction coefficient thresholds for the coating, cladding, and core layers enables precise identification of material layer transitions and a layered control strategy, avoiding the over- or under-polishing issues associated with traditional unified control parameters. Dynamic correction of the friction coefficient threshold is achieved through sliding window statistical analysis, an exponentially weighted moving average algorithm, and a multi-parameter coupling model, effectively addressing threshold drift caused by material property differences between fiber batches. A three-layer feedforward neural network is used to predict the cladding friction coefficient, combined with an adaptive compensation mechanism. Compared to simple threshold comparison, this method more accurately predicts material layer transitions and enables proactive control strategy adjustments. Differentiated control strategies (decelerated grinding, fine grinding, and immediate termination) are employed for different material layers. Through S-curve control and intermittent operation, a smooth transition and precise control of the grinding process are achieved. A triple-condition synchronous judgment (material layer confirmation, friction coefficient threshold, second-order derivative condition, and prediction error) and millisecond-level emergency braking response, combined with a triple safety protection mechanism, achieve the ideal goal of zero over-grinding.

[0122] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0123] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), magnetic disk or optical disk, etc., various media that can store program code.

[0124] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that the technical solutions described in the above embodiments can still be modified, or some of the technical features thereof can be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for accurately controlling the axial polishing thickness of an optical fiber, characterized in that: include: The friction coefficient of the optical fiber end face is monitored during the axial polishing process to obtain the comprehensive friction coefficient value and material layer identification signal; Comparing the comprehensive friction coefficient value with a coating layer friction coefficient threshold value, and generating a deceleration grinding execution instruction when the comprehensive friction coefficient value reaches the coating layer friction coefficient threshold value; comparing the comprehensive friction coefficient value with a cladding friction coefficient threshold, and generating a fine grinding execution instruction when the material layer identification signal indicates entering the cladding and the comprehensive friction coefficient value reaches the cladding friction coefficient threshold; The comprehensive friction coefficient value is compared with a core layer friction coefficient threshold value, and a target termination execution instruction is generated when the material layer identification signal indicates entering the core layer and the comprehensive friction coefficient value reaches the core layer friction coefficient threshold value.

2. The method for accurately controlling the axial polishing thickness of an optical fiber according to claim 1, characterized in that: The friction coefficient is monitored during the axial polishing process of the optical fiber end face to obtain a comprehensive friction coefficient value and a material layer identification signal, including: The three-axis force sensor is used to collect signals during the axial polishing process of the optical fiber end face to obtain radial friction force component signals and axial positive pressure signals; Performing friction coefficient vector synthesis based on the radial friction force component signal and the axial normal pressure signal to obtain an instantaneous friction coefficient value; performing noise elimination processing on the instantaneous friction coefficient value to obtain a denoised friction coefficient signal, and performing smoothing processing based on the denoised friction coefficient signal to obtain friction coefficient time series data; The piezoelectric sensor is subjected to signal processing based on the friction coefficient time series data to obtain a comprehensive friction coefficient value and a material layer identification signal.

3. The method for accurately controlling the axial polishing thickness of an optical fiber according to claim 2, characterized in that: The signal processing of the piezoelectric sensor based on the friction coefficient time series data to obtain a comprehensive friction coefficient value and a material layer identification signal includes: Through the multi-channel synchronous acquisition of the annular array composed of piezoelectric sensors, a distributed piezoelectric signal matrix is ​​obtained; Performing weighted fusion on the friction coefficient time series data and the distributed piezoelectric signal matrix to obtain a comprehensive friction coefficient value, and performing frequency domain decomposition based on the distributed piezoelectric signal matrix to obtain spectrum energy distribution data; The characteristic frequency bands of the material layers are identified based on the spectral energy distribution data. When the energy proportion of the first frequency band exceeds the first percentage, it is identified as a coating layer. When the energy proportion of the second frequency band exceeds the second percentage, it is identified as a cladding layer. When the energy proportion of the third frequency band exceeds the third percentage, it is identified as a core layer. A material layer identification signal is generated at the same time.

4. The method for accurately controlling the axial polishing thickness of an optical fiber according to claim 1, wherein: The step of comparing the comprehensive friction coefficient value with a coating layer friction coefficient threshold value and generating a deceleration grinding execution instruction when the comprehensive friction coefficient value reaches the coating layer friction coefficient threshold value includes: Performing sliding window statistics on the comprehensive friction coefficient value to obtain a sliding average value and a standard deviation value; Calculate the dynamic friction coefficient threshold of the coating layer based on the sliding average value and the standard deviation value; Comparing the comprehensive friction coefficient value with the coating layer dynamic friction coefficient threshold value, and generating a threshold trigger signal when the comprehensive friction coefficient value is greater than or equal to the coating layer dynamic friction coefficient threshold value; The deceleration grinding parameters are calculated according to the threshold trigger signal, and the deceleration grinding execution instruction is generated by adopting S-curve control.

5. The method for accurately controlling the axial polishing thickness of an optical fiber according to claim 4, characterized in that: The calculating of the deceleration grinding parameters according to the threshold trigger signal and generating the deceleration grinding execution instruction by adopting S-curve control include: Calculating deceleration grinding parameters based on the threshold trigger signal to obtain a numerical combination of deceleration parameters; Performing S-curve trajectory planning on the deceleration parameter value combination to obtain deceleration control trajectory data; Performing real-time calculation, monitoring, and processing of the friction coefficient gradient based on the deceleration control trajectory data, calculating the friction coefficient change rate, and generating a material layer conversion warning signal when the friction coefficient gradient of the friction coefficient change rate exceeds a preset gradient threshold; A deceleration grinding execution instruction including speed control, pressure control, feed control and early warning monitoring is constructed according to the deceleration control trajectory data and the material layer conversion early warning signal.

6. The method for accurately controlling the axial polishing thickness of an optical fiber according to claim 1, characterized in that: The step of comparing the comprehensive friction coefficient value with a cladding friction coefficient threshold value, and generating a fine grinding execution instruction when the material layer identification signal indicates entering the cladding and the comprehensive friction coefficient value reaches the cladding friction coefficient threshold value, comprises: The comprehensive friction coefficient value, the friction coefficient change rate, the grinding depth data and the temperature compensation value are used as input parameters to predict the cladding friction coefficient through a three-layer feedforward neural network to obtain a predicted value of the cladding friction coefficient; Performing adaptive compensation based on the predicted value of the cladding friction coefficient to obtain a cladding dynamic friction coefficient threshold; Performing a cladding confirmation judgment on the material layer identification signal, and generating a cladding threshold trigger signal when the material layer identification signal indicates entering the cladding and the comprehensive friction coefficient value is greater than or equal to the cladding dynamic friction coefficient threshold; Fine grinding parameters are calculated according to the cladding threshold trigger signal and a fine grinding execution instruction is generated in an intermittent working mode.

7. The method for accurately controlling the axial polishing thickness of an optical fiber according to claim 6, characterized in that: The step of calculating fine grinding parameters according to the cladding threshold trigger signal and generating fine grinding execution instructions in an intermittent working mode includes: Performing numerical calculation of fine grinding parameters based on the cladding threshold trigger signal to obtain a fine grinding parameter combination; Performing intermittent working sequence conversion on the fine grinding parameter combination to obtain an intermittent working control sequence; integrating the friction coefficient of the cladding thickness based on the intermittent working control sequence to obtain a cladding thickness monitoring signal; Based on the fine grinding parameter combination, the intermittent working control sequence and the cladding thickness monitoring signal, a fine grinding execution instruction including speed control, pressure control, micro-feed control, intermittent timing control and thickness monitoring is constructed.

8. The method for accurately controlling the axial polishing thickness of an optical fiber according to claim 1, wherein: The step of comparing the comprehensive friction coefficient value with a core layer friction coefficient threshold value, and generating a target termination execution instruction when the material layer identification signal indicates entering the core layer and the comprehensive friction coefficient value reaches the core layer friction coefficient threshold value, comprises: Based on the benchmark friction coefficient of the quartz core layer, a multi-parameter coupling analysis is performed to obtain the threshold value of the dynamic friction coefficient of the core layer. Inputting the comprehensive friction coefficient value into a cubic polynomial fitting algorithm for third-order prediction to obtain the second-order derivative of the friction coefficient and a prediction error value; Performing a triple-condition synchronous judgment on the material layer identification signal, the comprehensive friction coefficient value, the second-order derivative of the friction coefficient, and the predicted error value, and generating a core layer threshold trigger signal when the material layer identification signal indicates entering the core layer and the comprehensive friction coefficient value is greater than or equal to the core layer dynamic friction coefficient threshold; According to the core layer threshold trigger signal, the spindle is instantly stopped, the grinding head is quickly lifted, the coolant is immediately sprayed for cooling, and the triple safety protection mechanism of mechanical limit, electrical interlock and software monitoring is started to obtain the target termination execution instruction.

9. The method for accurately controlling the axial polishing thickness of an optical fiber according to claim 8, characterized in that: The method of inputting the comprehensive friction coefficient value into a cubic polynomial fitting algorithm for third-order prediction to obtain the second-order derivative of the friction coefficient and the prediction error value includes: Extracting time series sampling points of the comprehensive friction coefficient value to obtain a friction coefficient fitting data group; Inputting the friction coefficient fitting data set into a cubic polynomial fitting algorithm to calculate coefficients, thereby obtaining a polynomial fitting coefficient combination including a cubic term coefficient, a quadratic term coefficient, a linear term coefficient, and a constant term; Based on the polynomial fitting coefficient combination, the friction coefficient value is predicted by calculating the cubic polynomial function, and the second-order derivative value of the friction coefficient is obtained by performing a quadratic derivation on the polynomial function; The absolute value difference between the predicted friction coefficient value and the comprehensive friction coefficient value is calculated to obtain a predicted error value.

10. A device for precisely controlling the axial grinding thickness of an optical fiber, characterized in that: A method for accurately controlling the axial grinding thickness of an optical fiber according to any one of claims 1 to 9, wherein the device for accurately controlling the axial grinding thickness of the optical fiber comprises: A monitoring module is used to monitor the friction coefficient during the axial polishing process of the optical fiber end face and obtain a comprehensive friction coefficient value and a material layer identification signal; a first comparison module, configured to compare the comprehensive friction coefficient value with a coating layer friction coefficient threshold value, and generate a deceleration grinding execution instruction when the comprehensive friction coefficient value reaches the coating layer friction coefficient threshold value; a second comparison module, configured to compare the comprehensive friction coefficient value with a cladding friction coefficient threshold value, and generate a fine grinding execution instruction when the material layer identification signal indicates entering the cladding and the comprehensive friction coefficient value reaches the cladding friction coefficient threshold value; The third comparison module is used to compare the comprehensive friction coefficient value with the core layer friction coefficient threshold, and generate a target termination execution instruction when the material layer identification signal indicates entering the core layer and the comprehensive friction coefficient value reaches the core layer friction coefficient threshold.

Images