Optical fiber connector production line monitoring system based on digital twinning

Abstract

The invention relates to the technical field of industrial digital twinning control, and discloses an optical fiber connector production line monitoring system based on digital twinning, and the system comprises a curing unit which generates a digital accompanying bill representing the hardness trend of a product based on thermal energy integration; the grinding unit reconstructs a targeted pressure and duration process formula according to the bill before the product arrives; according to the system and the method, cross-process feed-forward control logic is constructed, mandatory constraint is carried out on machining parameters by utilizing thermal historical data, error transmission caused by dispersion of a previous process is eliminated, and the system stability is improved.

Description

TECHNICAL FIELD

[0001] The application relates to a fiber connector production line monitoring system based on digital twinning, and belongs to the technical field of industrial digital twinning control. BACKGROUND

[0002] The current fiber connector production process includes glue injection and curing, mechanical assembly and end face grinding and polishing processes. The existing automatic production line adopts an independent control architecture. The curing unit controls the thermal crosslinking of epoxy resin according to a preset time and temperature curve. The grinding unit controls the grinding disc speed, pressure and running time according to a standard process formula. Such a control mode relies on the consistency of the input conditions of each process, and the physical properties of the products delivered by the upstream process are constant.

[0003] The curing process is affected by the temperature field distribution gradient in the oven, the service life change of the glue and the difference in trace glue injection, resulting in dispersion of the hardness of the epoxy resin on the end face of the ferrule in batches or between batches. The clamping and feeding stability is improved by optimizing the mechanical structure of the grinding equipment, for example, the utility model patent with the authorized announcement number CN222078966U discloses a fiber connector grinder. The grinding head side is provided with a contact displacement component and is matched with a limiting and fixing component. The mechanical structure is used to solve the problem of simultaneous grinding of the end faces of multiple fiber connectors to the same level. Although such equipment solves the problem of mechanical fixing of simultaneous processing of multiple products, the control logic still does not deviate from the blind grinding category. The technology simply relies on mechanical rigid limiting or pre-set screw feeding stroke, and ignores the core variable of the fluctuation of the physical properties of the processed objects. The equipment cannot sense whether the incoming epoxy resin is in the under-cured soft state or the over-cured hard state. It can only execute fixed mechanical actions on all products. The existing grinding control strategy is mainly based on fixed time sequence logic or single-dimensional motor load feedback. When the hardness of the incoming material is low, the material removal rate is nonlinearly accelerated. The control system determines the end point according to the standard time or simply the load threshold, causing the fiber end face to be ground excessively, resulting in fiber depression or curvature radius geometric index out of tolerance. If the hardness of the incoming material is high, the grinding is insufficient or needs to be reworked.

[0004] Therefore, the technical problem to be solved by the present application is to reconstruct the data association between processes by using existing hardware resources, and to build a control system with cross-process feedforward compensation, multi-source signal feature decoupling and environmental self-adaptive capability. SUMMARY

[0005] To solve the problems raised in the background art, the technical solution of the present application is as follows: a fiber connector production line monitoring system based on digital twinning, comprising:

[0006] A process state fingerprint generation module is arranged at a front-end curing unit of the production line, configured to collect temperature time-series data of the curing process and perform thermal energy integration operation, and convert the obtained thermal energy accumulation value into a digital accompanying ticket representing the hardness trend of the product material to be processed according to a preset thermal energy and physical property mapping rule.

[0007] A grinding parameter feedforward compensation module is arranged at a control node of a subsequent grinding unit, configured to read the digital accompanying ticket before the product to be processed physically arrives, map the hardness trend data in the digital accompanying ticket to a grinding pressure correction value and a grinding time length correction value for the current batch according to a preset hardness and energy consumption correlation matrix, and reconstruct the grinding process recipe using the grinding pressure correction value and the grinding time length correction value.

[0008] An execution deviation checking module is configured to collect the load current waveform slope of the driving motor in real time during the execution of the grinding process recipe by the grinding unit, calculate the deviation between the load current waveform slope and the theoretical slope estimated based on the digital accompanying ticket, and trigger a process blocking instruction when the deviation exceeds a preset tolerance; the system constructs a cross-process feedforward control logic through the digital accompanying ticket, and implements a mandatory constraint on the machining parameters of the subsequent process using the thermal history data of the front-end process, so as to offset the machining errors caused by fluctuations in the incoming physical properties.

[0009] Preferably, the execution deviation checking module is configured with a transient contact response analysis logic, which is used to implement the following steps to realize multi-dimensional state decoupling under a single signal: capturing the rising edge of the load current of the grinding driving unit at the moment when the grinding head contacts the product to be processed, and calculating the time rate of change characteristic of the rising edge; calling a correlation model of the surface state of the grinding medium and the time rate of change characteristic pre-stored by the system, and evaluating the surface cutting activity of the current grinding medium based on the calculated time rate of change characteristic; before calculating the deviation, the surface cutting activity is used to normalize and correct the theoretical slope, so as to separate the interference of the grinding medium state on the load response characteristic, and ensure that the deviation only reflects the abnormality of the product physical properties.

[0010] Preferably, the control node of the grinding unit is configured with a dynamic reference anchoring logic: the logic is used to control the grinding driving unit to maintain an idle speed during the interval between two adjacent grinding processes, and collect the steady-state no-load current at this time as the zero-point reference of the current period; the grinding parameter feedforward compensation module and the execution deviation checking module are both configured to use the differential value relative to the zero-point reference for operation when processing the load response characteristic; at the same time, the logic compares the zero-point reference with the historical statistical mean value, identifies the mechanical thermal equilibrium state of the grinding equipment based on the comparison result, and dynamically adjusts the response gain parameter of the pressure adjustment instruction according to the mechanical thermal equilibrium state.

[0011] Preferably, the control node of the grinding unit is configured with a consumable state dynamic mapping module configured to perform the following steps to quantify the life consumption of the grinding medium: collecting the instantaneous power of the grinding drive unit in real time , and calculating the effective cutting power consumption equivalent of the current grinding cycle according to the following formula : , wherein, is the no-load reference power of the current cycle, is a preset wear coefficient related to the material of the grinding medium, is a nonlinear weighting factor greater than 1, used to represent the nonlinear acceleration effect of high load working conditions on the life of the medium, represents the effective cutting power after deducting the no-load loss; the grinding parameter feedforward compensation module updates the remaining life index of the grinding medium based on the effective cutting power consumption equivalent, and applies secondary compensation to the grinding process recipe based on the remaining life index in the next grinding cycle.

[0012] Preferably, the grinding unit includes rough grinding, intermediate grinding and fine grinding sub-steps arranged in time sequence, and the system is configured with inter-step state recursive logic: at the end of the rough grinding sub-step, the virtual residual geometric state of the product to be processed is calculated based on the actual executed grinding process recipe and the load response characteristics, and the digital accompanying ticket is updated; the control node of the intermediate grinding sub-step is configured to read the updated digital accompanying ticket, take the virtual residual geometric state as the initial boundary condition, and reconstruct the cutting speed parameter and the pressure holding time parameter of the intermediate grinding sub-step; the inter-step state recursive logic is continuously executed between the rough grinding, intermediate grinding and fine grinding sub-steps until the fine grinding sub-step is completed.

[0013] Preferably, the process state fingerprint generation module also combines the glue injection pressure decay curve characteristics uploaded by the glue injection unit when generating the digital accompanying ticket; the process state fingerprint generation module extracts rheological characteristic values representing the microscopic cross-linking density of the glue by analyzing the glue injection pressure decay rate; the thermal energy and physical property mapping rule includes a multi-dimensional weighting algorithm for fusing the thermal energy accumulation value and the rheological characteristic value to generate a digital accompanying ticket containing viscoelastic modulus prediction information.

[0014] Preferably, the operation of reconstructing the grinding process recipe performed by the grinding parameter feedforward compensation module specifically includes establishing a nonlinear pressure and time complementary mechanism: when the digital accompanying ticket indicates that the product hardness is lower than the preset standard threshold, an instruction is generated to reduce the grinding pressure set value while nonlinearly extending the grinding time set value, so as to maintain the total material removal amount constant and inhibit the generation of end face micro-cracks; when the digital accompanying ticket indicates that the product hardness is higher than the preset standard threshold, an instruction is generated to increase the grinding pressure set value while shortening the grinding time set value, so as to ensure that the material removal is completed within the predetermined production rhythm.

[0015] Preferably, the execution deviation checking module further comprises abnormal trend fusing logic: the logic is used to monitor the time integral value of the deviation in real time, and when the time integral value exceeds a safety threshold within a single polishing cycle, it is determined that the physical property of the incoming material is abnormal or the equipment is mechanically malfunctioning; in response to the determination, the system forces the polishing unit to stop the feeding action and marks the corresponding digital companion ticket as an abnormal lock state, thereby preventing the batch of products from being transferred to the next process until the lock is manually released.

[0016] Preferably, the cross-domain data synchronization unit is built based on the OPCUA unified architecture protocol and is configured with a time-sensitive network scheduling strategy for discrete manufacturing scenarios; the strategy is used to implant a production-tick-based time-to-live stamp in the data packet of the digital companion ticket; if the polishing unit reads the digital companion ticket after the time-to-live stamp expires, the system automatically triggers an expired data warning and forces the polishing unit to request re-verification of the product batch information, so as to prevent data and physical mismatches caused by production line congestion.

[0017] Preferably, the system further comprises a closed-loop model optimization unit for collecting the optical end face geometry parameters output by the final quality inspection process as real true values; the closed-loop model optimization unit calculates the residual error between the real true values and the predicted physical properties in the digital companion ticket, and uses the residual error to reversely correct the thermal energy and physical property mapping rules in the process state fingerprint generation module, so as to eliminate systematic prediction deviation through iterative convergence.

[0018] Compared with the prior art, the present application has the following advantages:

[0019] 1. A digital companion ticket containing thermal energy accumulation characteristics is generated in the solidification process and is transmitted to the polishing control node to build a cross-process parameter feedforward reconstruction mechanism, and the process formula is self-adaptively corrected before the product to be processed contacts the polishing medium, by using the solidification energy integral and the polishing resistance characteristic physical mapping relationship, so as to compensate for the fluctuation of the incoming physical properties in advance from after-feedback to pre-presetting, to eliminate the cross-process error transmission caused by the solidification state dispersion of the previous process in time sequence, to avoid the risk of overcutting or underpolishing of non-standard hardness incoming materials in traditional fixed parameter polishing processing, and to ensure the consistency convergence of the final product geometry parameters under complex incoming material conditions.

[0020] 2. The driving current change rate characteristics of the polishing head contacting the product are used to realize independent decoupling of the polishing medium surface cutting activity and the product physical hardness coupling variable, the contact instantaneous current establishes the speed, the polishing medium surface friction and the occlusion state are represented by the speed, the stable load amplitude represents the product material removal resistance, a double-checking logic is built to separate the interference components caused by polishing pad blockage or glazing under a single current signal source, and it is ensured that the pressure compensation instruction is triggered only for the hardness fluctuation of the product, and the time domain differential characteristic signal separation strategy is used to solve the underdetermined equation solving problem in the process of polishing under the premise of using existing sensing hardware.

[0021] 3. During process intervals, the no-load holding current of the drive unit is collected to establish a dynamic zero-point reference. A relative control logic based on differential characteristics is established. The background resistance drift caused by changes in the viscosity of the bearing grease or mechanical thermal expansion is automatically offset by real-time updates of the no-load reference. This makes the application and monitoring of grinding pressure independent of the absolute thermal equilibrium state of the equipment. The environmental adaptive mechanism ensures that the production line maintains consistent load sensing sensitivity and control accuracy throughout the entire temperature range from cold start to continuous full-load operation, eliminating batch-to-batch quality dispersion caused by the time-varying mechanical characteristics of the equipment. Attached Figure Description

[0022] Fig. 1 This is a schematic diagram of the cross-process feedforward compensation and closed-loop control principle of the digital twin of this invention;

[0023] Fig. 2 This is a comparison diagram of the transient response of load current under different material hardness and dielectric wear conditions of the present invention;

[0024] Fig. 3 This is a time-series interaction diagram of the initialization and real-time data acquisition and update of the digital twin model of this invention. Detailed Implementation

[0025] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0026] This invention provides a digital twin-based monitoring system for a fiber optic connector production line, comprising a process status fingerprint generation module, a cross-domain data synchronization unit, a grinding parameter feedforward compensation module, and an execution deviation verification module. These modules interact via an industrial Ethernet or OPCUA unified architecture protocol, forming a closed-loop control data flow from the acquisition of curing oven thermal history data to the reconstruction of grinding machine mechanical actions. Addressing the issue of discretized physical properties in fiber optic connector production caused by variations in curing oven temperature gradients or resin application amounts, the process status fingerprint generation module is connected to the curing oven's temperature sensor array. It acquires real-time temperature time-series data during the curing process at a preset sampling frequency, which can be set to 1Hz. This module internally runs a thermal energy integration algorithm to execute the following deterministic procedure: setting the threshold temperature for the crosslinking reaction of the epoxy resin. Secondly, the collected temperature curves medium to high The part is integrated over time, and the calculation formula is: ,in, and These are the start and end times of the curing process, respectively, to obtain the cumulative thermal energy value characterizing the current curing process. Finally, the pre-defined thermal energy and physical property mapping rules are invoked, and a lookup table calibrated based on historical experimental data is used to calculate the... The value is mapped to a dimensionless hardness prediction index and encapsulated as a digital accompanying document for the batch of products. This digital accompanying document serves as a unique digital certificate characterizing the hardness trend of the material to be processed and flows downstream with the physical carrier of the product. To solve the problem of over-cutting or under-grinding caused by the hardness fluctuation of incoming materials due to the execution of a fixed process formula in traditional grinding processes, the cross-domain data synchronization unit transmits the digital accompanying document to the grinding unit, and the grinding parameter feedforward compensation module performs process reconstruction. This grinding parameter feedforward compensation module is deployed at the control node of the grinding unit.

[0027] Before the physical carrier carrying the product to be processed arrives at the grinding station, the digital accompanying ticket bound to the carrier ID is read in advance. This module stores a hardness-energy consumption correlation matrix, which defines the correspondence between different hardness prediction indices and the optimal grinding pressure and grinding time. Based on this matrix, the module maps the hardness data in the ticket to a grinding pressure correction value for the current batch. Correction value for grinding time When the digital indicator of the product's hardness is lower than a preset standard threshold, the module generates a negative value. To reduce the grinding pressure setting value while generating a positive value By extending the grinding time setting nonlinearly, a low-pressure, long-grinding strategy is employed to maintain a constant total material removal and suppress micro-cracks on the end face. When the hardness exceeds the standard threshold, an instruction to increase the pressure and shorten the grinding time is generated. This module ultimately outputs a reconstructed grinding process formula containing the aforementioned correction values, which is then overwritten in the grinding machine PLC with the original standard formula. During the grinding process, to prevent quality accidents caused by feedforward model deviations or equipment malfunctions, the deviation verification module collects the load current signal of the drive motor in real time and calculates the slope of the load current waveform at high frequencies such as 100Hz. Simultaneously, based on the hardness index in the digital accompanying document, this module queries a preset theoretical model to obtain the expected theoretical load slope under the current process parameters. The module calculates the deviation between the two. The calculation formula is: When the deviation When the preset tolerance is exceeded, the module triggers a process blocking instruction, forcing the grinder to stop feeding and alerting, thereby achieving a fuse protection for abnormal trends within a single grinding cycle, which can be set to 15%; To further solve the problem of grinding load signal mixed with grinding medium surface state interference, the execution deviation verification module is configured with transient contact response analysis logic, which focuses on analyzing the current characteristics of the grinding head when it contacts the product to be processed. The specific procedures are as follows: the logic monitors the motor current, and marks the time when the current first crosses the no-load reference as the contact zero point ; secondly, the current data in the time window is intercepted, which can be set to , and the time change rate characteristic of the rising edge is calculated ; thirdly, based on the principle of tribology, the surface cutting activity of the grinding medium is evaluated using the pre-set correlation model, where a higher corresponds to a sharp and clean grinding pad surface, and a lower corresponds to a surface with clogging or glazing; finally, the surface cutting activity index is used to normalize the correction of the foregoing theoretical slope , which eliminates the interference of the grinding medium state on the load response characteristic, and ensures that the final calculated deviation only reflects the abnormality of the product physical properties.

[0028] In view of the problem of basic mechanical resistance drift of the grinding equipment in the cold start and continuous hot running state, the control node of the grinding unit is configured with dynamic reference anchoring logic, which uses the gap time between adjacent two grinding processes, which is usually 5 seconds of product loading and unloading gap, to control the grinding drive unit to maintain the idle speed, during which the logic collects the steady-state no-load current and calculates its average value, which is locked as the zero reference of the current grinding cycle , in the grinding process, the grinding parameter feedforward compensation module and the execution deviation verification module use the difference value relative to the when processing any load response characteristic, that is , the logic also compares the measured with the historical statistical average, if the is higher than the average, it means that the cold state and the grease viscosity are large, then the PID response gain parameter of the pressure regulating loop is automatically adjusted to prevent control overshoot; To accurately manage the life of the grinding medium and avoid nonlinear wear errors caused by adaptive variable parameter processing, the grinding unit is configured with a consumable state dynamic mapping module, which executes a life quantization procedure based on effective cutting power consumption. In each grinding cycle, the module real-time collects the instantaneous power of the grinding drive unit , and calculates the effective cutting power equivalent of the current cycle according to the formula , where, a dynamic anchoring idle reference power, a preset wear coefficient related to the material of the grinding media, a non-linear weighting factor greater than 1, used to characterize the non-linear acceleration of the impact of high load conditions on the life of the media, which can take the value 1.5, the module accumulating this value to update the residual life index of the grinding media, which is read by the grinding parameter feedforward compensation module when generating the next batch of grinding recipes, if the index shows that the media is at the end of life, i.e. the cutting force is decreasing, then an additional time length compensation coefficient is automatically superimposed in the recipe.

[0029] ​For multi-pass grinding process including rough grinding, medium grinding and fine grinding, the system is configured with inter-step state recursion logic to realize error convergence across steps. At the end of the rough grinding sub-step, the system records the actual grinding process parameters and load response characteristics during the process, uses the material removal rate model to inversely calculate the virtual residual geometric state of the product end face, such as residual height or curvature radius deviation, and updates the state to the digital accompanying ticket. Before the medium grinding sub-step starts, the control node reads the updated ticket and uses the virtual residual geometric state left by the rough grinding as the initial boundary condition. If the data shows that the rough grinding removal amount is too large, the medium grinding control node automatically reconstructs the cutting speed parameter to reduce the feed speed and reconstructs the pressure holding time parameter to shorten the grinding time, thereby preventing excessive removal. This recursive logic is continuously executed between the rough grinding, medium grinding and fine grinding sub-steps to ensure that the parameters of each process are dynamically adjusted based on the actual output state of the previous process. In addition, to deal with the variables introduced by the glue injection process, the process state fingerprint generation module also combines the glue injection pressure decay curve characteristics uploaded by the glue injection unit. The module analyzes the rate at which the pressure decreases over time during the glue injection process, extracts rheological characteristic values representing the microscopic cross-linking density of the glue, and uses a multi-dimensional weighting algorithm in the thermal energy and physical property mapping rule to fuse the cumulative value of the solidification thermal energy with the rheological characteristic values to generate a digital accompanying ticket containing viscoelastic modulus prediction information, thereby more comprehensively representing the physical processing performance of the incoming material. The system's data transmission is based on the OPCUA unified architecture protocol and is configured with a time-sensitive network scheduling strategy for discrete manufacturing scenarios. This strategy implants a survival timestamp based on the production tempo in the data packet of the digital accompanying ticket. If the grinding unit does not read the digital accompanying ticket until the survival timestamp expires due to production line congestion, the system automatically triggers an expired data warning and forces the grinding unit to request the MES system to re-verify the product batch information, preventing the risk of data and physical mismatch caused by production tempo disorder. The system also includes a closed-loop model optimization unit to ensure continuous control accuracy. This unit collects optical end face geometric parameters such as curvature radius ROC and apex offset ApexOffset output by the final inspection process as true values, calculates the residual error between the true values and the predicted physical properties in the digital accompanying ticket, and uses the residual error to reverse correct the thermal energy and physical property mapping rule in the process state fingerprint generation module. The correction process uses an iterative convergence algorithm to gradually adjust the weight coefficients in the mapping table.

[0030] Embodiment 1: In the mass production scenario of multi-core MPO fiber connector, the local temperature field gradient existing inside the tunnel type curing oven leads to insufficient cross-linking reaction of epoxy resin when part of the batch product passes through the low temperature zone, thus producing a low physical hardness after curing, which is lower than the standard process benchmark value. In the traditional fixed parameter polishing process, such low hardness incoming material withstands the standard polishing pressure, the material removal rate is too fast, causing the fiber to be recessed or the end face to be scratched, which triggers batch quality accidents. The process state fingerprint generation module collects temperature data in real time during the curing process, and calculates the accumulated value of thermal energy in the curing process which is lower than the preset complete curing energy threshold. The system generates a digital accompanying ticket containing the low hardness feature identification and binds it with the physical carrier ID carrying the batch product. Before the carrier arrives at the polishing station, the polishing parameter feedforward compensation module reads the digital accompanying ticket and reconstructs the process according to the hardness and energy consumption correlation matrix. The system calls the compensation strategy for soft material and generates a negative polishing pressure correction value and a positive polishing time correction value , and the original standard high pressure short time process of the polishing machine is reconstructed into a low pressure long time process.

[0031] The execution deviation checking module cooperates with the transient contact response analysis logic. Due to the active reduction of polishing pressure, the steady-state load current amplitude of the driving motor decreases. The system uses the current rising edge feature of the polishing head contacting the product instantaneously to distinguish whether the decrease is due to expected process adjustment or polishing pad slip. The logic extracts the time rate of change feature at the contact instant . If the value remains high, it indicates that the polishing medium surface remains sharp and there is no glazing phenomenon. The system determines that the current low load current is the correct response of the pressure correction strategy and allows the processing to continue. If they decrease synchronously, an abnormality is triggered to block. This mechanism of normalizing the steady-state response using transient features ensures the logical stability of the system under variable parameter control and avoids false judgments due to signal aliasing. The dynamic benchmark anchoring logic is used to ensure the execution accuracy of the small pressure correction. Since the pressure value set by the low pressure polishing strategy is close to the bottom limit of the mechanical friction resistance of the equipment, and this resistance is affected by the change of the viscosity of the lubricating grease, the system collects the steady-state no-load current during the no-load rotation period before processing and locks it as the zero point benchmark . The subsequent pressure control loop is forced to execute feedback regulation based on the difference between the current value and the benchmark. This procedure eliminates the background resistance fluctuations of the equipment due to different engine states, ensuring that the pressure correction instructions generated by the feedforward module are converted into actual contact stress of the polishing interface without loss.

[0032] Example 2: This example builds a test verification platform that includes a complete process closed loop. The technical effectiveness of the fiber connector production line monitoring system based on digital twinning is objectively verified. The test platform simulates real industrial production processes, covering the glue injection and curing unit and the grinding processing unit. The data acquisition and control process strictly follows the industrial field standard. The curing unit is equipped with a high-precision thermocouple array, with a real-time sampling frequency of 1 Hz and a measurement accuracy of ±0.5℃. It is used to capture the temperature time series data of each temperature zone in the curing furnace. The grinding unit integrates a high-frequency current sensor with a sampling frequency of 1000 Hz and a resolution of 0.01 A. It accurately captures the transient and steady-state load response of the driving motor during the grinding process. The interferometer is used as the final quality inspection equipment to quantitatively measure the geometric parameters of the rear surface after grinding. To comprehensively evaluate the effectiveness of the technical solution, a multi-dimensional control test system including three test objects is constructed. Control group A uses the traditional fixed process parameter grinding strategy and does not have a feedforward compensation mechanism. Control group B introduces feedforward compensation based on curing temperature integration, but lacks grinding medium state verification logic based on transient contact response. Test group C fully implements the system, including thermal energy integration feedforward, transient contact response verification, and dynamic reference anchoring. The test introduces a problem intensity gradient by artificially adjusting the curing furnace temperature control parameters to construct three different levels of incoming material states: light under-curing (curing degree about 90%), moderate under-curing (curing degree about 80%), and severe under-curing (curing degree about 70%). The response capability of each scheme to different degrees of process deviation is tested. During the test process, the incoming materials of each group are cured and the thermal history data is recorded. For test group C, the system calculates the thermal energy accumulation value and generates a digital companion ticket based on the thermal history data. Each group of samples enters the grinding process. Control group A always executes the standard pressure (20N) and standard time (30s) grinding formula. Control group B only adjusts the pressure and time based on the thermal energy data, ignoring the state of the grinding pad. Test group C reconstructs the process based on the ticket while real-time monitoring the contact transient current rate of change and correcting the load deviation calculation accordingly. After grinding, the interferometer is used to measure the fiber undercut and radius of curvature (ROC) of each ferrule, and the good product rate of each group is calculated. Table 1 shows the grinding quality comparison data of different test groups under different curing states.

[0033] Table 1: Grinding quality comparison data table

[0034]

[0035] ​Referring to Table 1, the data presents a gradient variation rule. The data of the control group A shows that, with the deepening of the under-curing degree, the fiber recess amount increases non-linearly and sharply under the fixed parameters, and the yield rate drops cliff-like, which confirms that a single fixed process cannot cope with the fluctuation of the physical properties of the incoming material. Although the control group B alleviates the recess problem by reducing the pressure when the under-curing degree is severe, the curvature radius deviates from the optimal range due to the lack of real-time verification of the state of the polishing pad, and the yield rate only rises to 78%, which implies the limitations of a single feedforward mechanism under complex working conditions. The test group C maintains a very high yield rate (>96%) and stable geometric parameters under all curing conditions, with the recess amount controlled within-100 nm and the ROC stabilized at about 19 mm. The data of the test group C proves that the scheme of the present application converges the incoming material with different initial states to the same high-quality endpoint through multi-dimensional collaborative control. Under severe under-curing conditions, the test group C adjusts the pressure and time significantly and combines the transient response verification to ensure that the adjustment is effectively executed.

[0036] Analyzing the process data of the test group C during the polishing process reveals the synergistic mechanism. When processing severe under-curing samples, the contact transient current change rate is always maintained in the preset high-activity interval, verifying that the system correctly distinguishes between low current caused by low polishing pressure and low current caused by polishing pad slipping. This distinguishing ability enables the system to execute significant parameter correction under extreme working conditions, avoiding false alarms triggered by misjudgment, and the dynamic reference anchoring logic ensures that the effective cutting force actually applied to the product is not swallowed by device mechanical resistance fluctuations under low pressure (12N) settings. Through comparative tests and data analysis, the monitoring system based on digital twinning demonstrates its effectiveness in dealing with the curing state dispersion problem in fiber connector production. Through the organic synergy of thermal energy integration feedforward, transient contact response verification, and dynamic reference anchoring, the system effectively eliminates the impact of front-end process fluctuations on the quality of the final product, achieving cross-process quality closed-loop control.

[0037] Embodiment 3: This embodiment combines Figs. 1 to 3 the monitoring system for the fiber connector production line based on digital twinning, as described above. Fig. 1As shown, the data from the curing oven sensor is used as the input source and transmitted to the process status fingerprint generation module. This module performs operations such as collecting temperature time-series data, performing thermal energy integration calculations, and generating digital accompanying documents. The digital accompanying documents are transmitted to the grinding parameter feedforward compensation module via the cross-domain data synchronization unit using the OPCUA unified architecture protocol, embedded survival timestamps, and TSN scheduling strategy. This compensation module reads the documents and reconstructs the grinding process formula based on the hardness and energy consumption correlation matrix. The deviation verification module monitors the load current waveform slope based on the theoretical slope benchmark and runs transient contact response analysis logic and abnormal trend fuse protection. After grinding is completed and transferred to the quality inspection process, the closed-loop model optimization unit collects the optical end face geometric parameters as the true value, calculates the predicted residual, and reversely corrects the thermal energy property mapping rules.

[0038] like Fig. 2 As shown, the horizontal axis represents contact time in milliseconds (ms), and the vertical axis represents load current in amperes (A). The figure depicts current variation curves under four typical conditions: standard hardness incoming material, low hardness incoming material, high hardness incoming material, and abrasive media wear state. Different signals exhibit different rising edge characteristics at the moment of contact (10ms to 12ms). High hardness incoming material produces the highest current amplitude response, while low hardness incoming material produces the lowest response amplitude. The curve under abrasive media wear state shows a slow rise and oscillation characteristics. Fig. 3 As shown, the process begins with the user / operator starting the system. The monitoring system requests configuration information from the database and receives the returned data, which in turn triggers the digital twin engine to initialize the twin model. After the model is loaded, the monitoring system establishes a communication connection with the production line PLC and confirms it, and starts the sensor acquisition command. After the sensor network is ready for acquisition, it enters the data acquisition loop, continuously collecting production data and uploading real-time data to the digital twin engine to update the twin model. At the same time, it stores historical data in the database until the system is running normally.

[0039] Example 4: This example details the system parameter identification and model initialization procedures performed by the fiber optic connector production line monitoring system before mass production to establish benchmarks for key control parameters and algorithm models. Given the inherent physical differences in spindle inertia, motor response characteristics, and transmission chain friction coefficients among different batches of grinding machines, directly applying universal algorithm parameters would lead to discrepancies in control accuracy. Therefore, the system executes a dynamic characteristic calibration program for the grinding drive unit. In this program, the system controls the drive motor to execute a series of step speed commands under no-load conditions, and uses a high-frequency current sensor to collect the response current sequence at a frequency of 1000Hz. The control node of the grinding unit internally runs a parameter identification algorithm based on the least squares method to analyze the sequence and the key features required for transient contact response analysis logic. The system captures the moment of contact. within the last 50 milliseconds discrete sampling points, construct a linear regression model by minimizing the sum of squared residuals to solve the slope with statistical robustness This algorithm path suppresses the interference of single-point current noise on feature extraction.

[0040] After completing the dynamic calibration, the system uses a set of golden samples with known physical properties that meet the standard tolerance to calibrate the mapping rule between the process state fingerprint generation module and the grinding parameter feedforward compensation module. The system records the steady-state load current of the set of samples under the standard grinding formula, calculates the arithmetic mean and standard deviation Based on the statistical principle of normal distribution, the deviation checking module automatically sets the preset tolerance to , ensuring that the abnormal trigger threshold strictly adapts to the mechanical precision limit of the current equipment, avoiding both missed detection due to too wide threshold and false alarms due to too narrow threshold, achieving a balance between control sensitivity and false alarm rate, and establishing a discrete reference feature scanning procedure for the hardness and load slope theoretical correlation model. The Vickers hardness value covers the upper and lower limit gradient standard samples in the process window, and the grinding is carried out under constant speed and pressure process parameters. The sensor records the corresponding load current waveform data of each sample at a frequency of 1000Hz and extracts the steady-state slope characteristic value. The least squares method is used to fit the discrete hardness value and slope characteristic value as a quadratic polynomial function relationship. The function serves as the mathematical basis for calculating the theoretical slope in subsequent production, and the function coefficients are dynamically reset to zero at the grinding medium replacement period. For the nonlinear weighting factor involved in the consumable state dynamic mapping module , the system performs a destructive full-life wear test. In the test, the system uses a single grinding medium to continuously process standard samples until the cutting ability is completely lost. During this period, the cumulative cutting power consumption and the corresponding material removal rate of each processing cycle are recorded in real time. The system fits the obtained data set to a power-law decay model , and solves the inherent wear index of the specific batch of grinding consumables through nonlinear regression operation . This value is fixed in the consumable management algorithm as the core index for calculating the effective cutting power consumption equivalent in subsequent production . This process upgrades the consumable life management from fuzzy empirical estimation to precise quantitative characterization based on material tribological characteristics, ensuring that the system's prediction error of the remaining life of the consumables converges to an acceptable small range under variable parameter processing conditions.

[0041] Example 5: Regarding the initial construction of the thermal energy and physical property mapping rules in the process status fingerprint generation module, the system executes an offline gradient characteristic procedure before formal production. In this procedure, the curing unit is programmed to execute a set of discrete temperature control sequences covering the upper and lower limits of the process window. By adjusting the conveyor belt speed or heating zone power in a stepwise manner, a series of standard samples with controlled curing degree differences are produced. The system synchronously records the thermal energy integral value of each sample during the curing process. And in conjunction with an offline hardness tester, the corresponding true physical hardness value is measured. Thus constructing a collection Sample set of data pairs The system invokes its internally integrated nonlinear regression algorithm to fit the discrete sample set into a continuous functional relationship. Furthermore, this function is solidified into the core mapping rule during production operation to avoid systematic prediction deviations that may arise due to discrepancies between the theoretical model and actual material properties.

[0042] To ensure the applicability of the dynamic reference anchoring logic under different physical equipment or ambient temperatures, the system performs a full-speed-domain mechanical resistance spectrum scanning procedure after initial deployment or major overhaul. The system controls the grinding drive unit to perform linear frequency sweep operation from the lowest to the highest operating speed under no-load conditions, during which steady-state current data of the motor is acquired at high frequency. Based on this, the system plots a reference resistance curve reflecting the current mechanical friction characteristics of the equipment as a function of speed. The curve is stored in non-volatile memory. During subsequent production intervals, the system only needs to collect the no-load current at a single specific speed and compare it with... By comparing the corresponding values, the current global thermal drift coefficient can be calculated, and the entire resistance curve can be shifted or scaled accordingly to achieve adaptive compensation for equipment mechanical aging or changes in ambient temperature. The pressure regulation command response gain parameter of the grinding unit control node adopts a discretization compensation strategy based on the lookup table method. The controller pre-stores the no-load current segmented interval with a step size of 0.1A and the corresponding PID controller proportional coefficient. With integration time constant When the dynamic reference anchoring logic detects that the zero-point reference falls within a specific range, the controller calls the corresponding PID parameter group to rewrite the servo drive control register. A cold start causes the zero-point reference to exceed the rated value, automatically increasing the proportional coefficient. The set value overcomes the high viscosity frictional resistance of low-temperature grease, keeping the pressure response rise time within 50ms.

[0043] Embodiment 6: This embodiment aims to disclose a set of standardization engineering calibration procedures for determining the initial value of the hardness-energy consumption correlation matrix in the grinding parameter feedforward compensation module. In view of the hardness discreteness exhibited by different batches of fiber connectors after glue injection and curing, a systematic experiment is needed to establish the quantitative mapping relationship between the hardness prediction index and the grinding process parameters. A set of standard samples with known curing states covering the full range of the process window, for example, curing degrees from 70% to 100% with an interval of 5%, are selected. The real Vickers hardness values of the ferrule end faces of the samples are measured using a nanoindenter, and they are normalized to dimensionless hardness prediction indexes. For each type of hardness index sample, a full-factor orthogonal test is performed on a standard grinder, and the test variables include grinding pressure (10N to 30N, gradient 2N), grinding disc speed (80rpm to 160rpm, gradient 20rpm), and grinding time (20s to 60s, gradient 5s). In each test, the steady-state power consumption of the drive motor is recorded, and the fiber recess and curvature radius after grinding are measured using an interferometer.

[0044] Based on the above experimental data, the fiber recess is taken as the objective function (target value -100nm, tolerance ±50nm), and the pressure and time combination that can make the objective function optimal under each hardness index is selected. At the same time, the corresponding unit volume material removal energy consumption under the optimal combination is recorded. Through regression analysis, the functional relationship between the hardness prediction index and the optimal grinding pressure and time is established: , Finally, the above discrete optimal parameter points are expanded to a continuous correlation matrix through interpolation algorithm, and are fixed in the storage unit of the grinding parameter feedforward compensation module.

[0045] It is obvious to those skilled in the art that the present application is not limited to the details of the above exemplary embodiments, and can be implemented in other specific forms without departing from the spirit or essential characteristics of the present application.

[0046] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and are not limiting. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced by equivalents without departing from the spirit and scope of the present application.

Claims

1. A digital twin based fiber optic connector production line monitoring system, characterized by, The system comprises: a process state fingerprint generation module deployed at a front-end curing unit of a production line, configured to collect temperature time-series data of a curing process and perform thermal energy integration operation, and convert the obtained thermal energy accumulation value into a digital accompanying ticket representing the hardness trend of the material of the product to be processed according to a preset thermal energy and material property mapping rule; a grinding parameter feedforward compensation module deployed at a control node of a subsequent grinding unit, configured to read the digital accompanying ticket before the physical product to be processed arrives, and map the hardness trend data in the digital accompanying ticket to grinding pressure correction value and grinding time length correction value for the current batch according to a preset hardness and energy consumption correlation matrix, and reconstruct the grinding process recipe using the grinding pressure correction value and the grinding time length correction value; an execution deviation checking module configured to collect the load current waveform slope of the driving motor in real time during the execution of the grinding process recipe by the grinding unit, calculate the deviation between the load current waveform slope and the theoretical slope estimated based on the digital accompanying ticket, and trigger a process blocking instruction when the deviation exceeds a preset tolerance; the system constructs a feedforward control logic across processes through the digital accompanying ticket, and uses the thermal history data of the front-end process to implement mandatory constraints on the mechanical processing parameters of the rear-end process, thereby offsetting the processing errors caused by fluctuations in the physical properties of the incoming material.

2. The digital twin based optical fiber connector production line monitoring system of claim 1, wherein, The execution deviation checking module is configured with transient contact response analysis logic, which is configured to perform the following steps: capture the rising edge of the load current of the grinding drive unit at the moment when the grinding head contacts the product to be processed, and calculate the time rate of change characteristic of the rising edge; invoke a system-preset correlation model of the surface state of the grinding medium and the time rate of change characteristic, and based on the calculated time rate of change characteristic, evaluate the surface cutting activity of the current grinding medium; before calculating the deviation, the surface cutting activity is used to normalize and correct the theoretical slope, so as to eliminate the interference of the state of the grinding medium on the load response characteristic.

3. The digital twin based optical fiber connector production line monitoring system of claim 1, wherein, The control node of the grinding unit is configured with dynamic reference anchoring logic: the logic is configured to control the grinding drive unit to maintain idle rotation at working speed during the interval between two adjacent grinding processes, and collect the steady-state no-load current at this time as the zero-point reference of the current period; the grinding parameter feedforward compensation module and the execution deviation checking module are both configured to use the differential value relative to the zero-point reference for operation when processing the load response characteristic; at the same time, the logic compares the zero-point reference with the historical statistical mean value, identifies the mechanical thermal equilibrium state of the grinding equipment based on the comparison result, and dynamically adjusts the response gain parameter of the pressure regulation instruction according to the mechanical thermal equilibrium state.

4. The digital twin based optical fiber connector production line monitoring system of claim 1, wherein, The control node of the polishing unit is configured with a consumable state dynamic mapping module, which is configured to perform the following steps to quantify the life loss of the polishing medium: collecting the instantaneous power of the polishing drive unit in real time , and calculating the effective cutting power consumption equivalent of the current polishing cycle according to the following formula : , wherein is the no-load reference power of the current cycle, is a preset wear coefficient related to the material of the polishing medium, is a nonlinear weighting factor greater than 1, used to represent the nonlinear acceleration effect of high load working conditions on the life of the medium, represents the effective cutting power after deducting the no-load loss; the polishing parameter feedforward compensation module updates the remaining life index of the polishing medium based on the effective cutting power consumption equivalent, and applies secondary compensation to the polishing process recipe based on the remaining life index in the next polishing cycle.

5. The digital twin based optical fiber connector production line monitoring system of claim 1, wherein, The grinding unit comprises coarse grinding, intermediate grinding and fine grinding sub-steps arranged in sequence, and the system is configured with inter-step state recursion logic: at the end of the coarse grinding sub-step, the virtual residual geometric state of the product to be processed is calculated based on the actually executed grinding process recipe and the load response characteristic, and the digital accompanying ticket is updated; the control node of the intermediate grinding sub-step is configured to read the updated digital accompanying ticket, and use the virtual residual geometric state as the initial boundary condition to reconstruct the cutting-in speed parameter and the pressure maintaining time parameter of the intermediate grinding sub-step; the inter-step state recursion logic is continuously executed between the coarse grinding, intermediate grinding and fine grinding sub-steps until the fine grinding sub-step is completed.

6. The digital twin based optical fiber connector production line monitoring system of claim 1, wherein, The process state fingerprint generation module also combines the glue injection pressure decay curve characteristics uploaded by the glue injection unit when generating the digital accompanying ticket; the process state fingerprint generation module extracts rheological characteristic values representing the microscopic cross-linking density of the glue by analyzing the glue injection pressure decay rate; the thermal energy and physical property mapping rule includes a multi-dimensional weighting algorithm for fusing the thermal energy accumulation value and the rheological characteristic value to generate a digital accompanying ticket containing viscoelastic modulus prediction information.

7. The digital twin based optical fiber connector production line monitoring system of claim 1, wherein, The operation of reconstructing the grinding process recipe performed by the grinding parameter feedforward compensation module specifically includes establishing a nonlinear pressure and time complementary mechanism: when the digital accompanying ticket indicates that the product hardness is lower than the preset standard threshold, an instruction is generated to reduce the grinding pressure set value while nonlinearly extending the grinding time length set value, maintaining the total material removal amount constant and inhibiting the generation of end face micro-cracks; when the digital accompanying ticket indicates that the product hardness is higher than the preset standard threshold, an instruction is generated to increase the grinding pressure set value while shortening the grinding time length set value.

8. The digital twin based optical fiber connector production line monitoring system of claim 1, wherein, The execution deviation checking module also includes abnormal trend fusing logic: the logic is used to monitor the time integral value of the deviation in real time, and when the time integral value exceeds the safety threshold within a single grinding period, it is determined that the incoming material physical property is abnormal or the equipment mechanical failure; the system responds to the determination by forcing the grinding unit to stop the feeding action, and marking the corresponding digital accompanying ticket as an abnormal locked state, prohibiting the batch of products from flowing to the next process until the lock is manually released.

9. The digital twin based optical fiber connector production line monitoring system of claim 1, wherein, The cross-domain data synchronization unit is built based on the OPCUA unified architecture protocol and is configured with a time-sensitive network scheduling strategy for discrete manufacturing scenarios; the strategy is used to implant a survival time stamp based on the production beat in the data packet of the digital accompanying ticket; if the grinding unit reads the digital accompanying ticket after the survival time stamp expires, the system automatically triggers an expired data warning and forces the grinding unit to request re-verification of the product batch information.

10. The digital twin based optical fiber connector production line monitoring system of claim 1, wherein, The system also includes a closed-loop model optimization unit for collecting the optical end face geometry parameters output by the final quality inspection process as real true values; the closed-loop model optimization unit calculates the residual error between the real true values and the predicted physical properties in the digital accompanying ticket, and uses the residual error to reverse correct the thermal energy and physical property mapping rule in the process state fingerprint generation module, eliminating systematic prediction bias through iterative convergence.

Images