A carbon slide plate full life cycle polishing method and system based on life prediction

Abstract

This invention relates to a method and system for grinding carbon slide plates throughout their entire lifecycle based on lifespan prediction. The invention collects and preprocesses visual and historical lifecycle data of carbon slide plates, and configures a training dataset based on real lifespans. It then uses this training dataset to train a model of the remaining lifespan of the carbon slide plate. For any carbon slide plate to be processed, its data is obtained in the form of the training dataset, and the trained model is used to predict its current predicted remaining lifespan. Grinding settings are then configured based on the predicted remaining lifespan and the three-dimensional digital model of the carbon slide plate, including: selecting a grinding strategy based on the predicted remaining lifespan that meets a confidence threshold; performing mesh analysis based on the selected grinding strategy and the three-dimensional digital model to obtain the cutting depth and desired cutting force setting value of the grinding tool on the planned path; and finally, grinding is performed on the carbon slide plate according to the determined grinding strategy, cutting path, depth, and desired cutting force setting value.

Description

Technical Field

[0001] This invention relates to the field of maintenance technology for carbon sliding plates of pantographs in rail transit, and in particular to a grinding method and system for the entire life cycle of carbon sliding plates based on life prediction. Background Technology

[0002] Carbon contact plates are a core component of the pantograph in electric locomotives and urban rail vehicles, collecting electrical energy through sliding friction with the overhead contact line. During operation, they continuously rub against the contact wire and are subjected to multiple effects, including electrical arcing, mechanical impact, and environmental corrosion, inevitably resulting in defects such as wear, dents, cracks, skew wear, and unevenness. These defects not only exacerbate vibration and offline arcing between the pantograph and the contact wire, leading to deterioration of current collection quality and increased energy loss, but can also scratch or even break the contact wire, causing serious traffic safety accidents. Therefore, regular condition inspection and restorative grinding of carbon contact plates are necessary measures to ensure the safety and efficiency of rail transit operations.

[0003] Existing carbon skateboard maintenance and polishing technologies mainly rely on planned maintenance at fixed intervals or mileages, and employ preset procedures during the polishing process. This presents several significant technical bottlenecks: Current maintenance decisions are mostly based on fixed time intervals (e.g., every X kilometers) or simple thickness measurement thresholds. This approach fails to accurately reflect the actual health and degradation trend of individual carbon skateboards. A carbon skateboard that appears uniformly worn but has internal microcracks may continue to be used before reaching the thickness threshold, posing a safety hazard; conversely, a carbon skateboard in good condition may be prematurely and excessively worn, resulting in material waste and a shortened lifespan. This "one-size-fits-all" maintenance model leads to both under-maintenance and over-maintenance.

[0004] Carbon slide plates, as a carbon-based composite material, inherently exhibit non-uniformity in internal material density, resulting in extremely complex wear profiles during operation. Traditional automated grinding equipment typically relies on pre-set, fixed toolpaths and feed rates. This open-loop control method cannot make real-time adjustments when encountering changes in material hardness or abrupt changes in geometric profile. The direct consequences are: in softer or concave areas, the actual cutting amount is less than expected, leading to "under-grinding" and incomplete defect elimination; in harder or convex areas, the cutting force increases sharply, resulting in "over-grinding" or "over-cutting," which not only damages the carbon slide plate matrix but may also cause tool chipping or even equipment vibration. This results in poor surface quality consistency after grinding, making it difficult to achieve the ideal flow-receiving surface morphology. Summary of the Invention

[0005] In order to solve the above-mentioned technical problems, or at least partially solve the above-mentioned technical problems, and in view of the above-mentioned shortcomings, the present invention provides a carbon skateboard full life cycle grinding method and system based on life prediction.

[0006] In a first aspect, the present invention provides a carbon skateboard full life cycle polishing method based on lifespan prediction, comprising: Visual and full life-cycle historical data collection and collaborative aggregation of carbon skateboards; personalized data cleaning and preprocessing for each carbon skateboard based on internal standard scores; alignment and reconstruction of the cleaned high-quality feature parameters and working condition data in strict chronological order, and configuration of real lifespan for corresponding time points to construct a regular training dataset suitable for input to time series analysis models; training of the carbon skateboard remaining lifespan model using the training dataset. For any carbon skateboard to be processed, its data is obtained in the form of a training dataset, and its current predicted remaining lifespan is predicted using the trained carbon skateboard remaining lifespan model. Automated grinding settings are achieved by combining the predicted remaining service life with a 3D digital model of the carbon slide plate. This includes: selecting a grinding strategy based on the predicted remaining service life that meets a confidence threshold; performing mesh analysis based on the selected grinding strategy and the 3D digital model to calculate the cutting depth of the grinding tool along the planned path; inputting the cutting depth into a pre-calibrated carbon slide plate material removal model to calculate a constant desired cutting force setting; and establishing a quantitative relationship between cutting force, feed rate, cutting depth, and material removal rate using the carbon slide plate material removal model. Carbon slide plate cutting is performed according to the grinding strategy, cutting path, depth, and expected cutting force settings determined by the grinding settings.

[0007] Furthermore, industrial cameras and 3D laser scanners are used to acquire images and 3D point cloud data of the carbon skateboard. The carbon skateboard images are used to identify macroscopic surface defect features. The 3D point cloud data is used to accurately reconstruct the 3D digital model of the carbon skateboard and accurately determine the current geometric feature parameters of the carbon board, including the current contour, current wear depth, and current wear distribution. The data interface is used to obtain the full life cycle historical data of the carbon skateboard from the enterprise maintenance management system, including but not limited to: each grinding record, including grinding time and grinding amount; cumulative running mileage; route information, including route slope, curve conditions and cable conditions; and operating current.

[0008] Furthermore, personalized data cleaning and preprocessing are performed on each carbon skateboard based on the internal standard score, including: for each individual carbon skateboard, for any characteristic parameter of the carbon skateboard, calculate the internal standard score of that characteristic parameter; set a score threshold, and when the absolute value of the internal standard score of a data point exceeds the corresponding score threshold, it is determined to be an outlier that deviates from its own historical normal level and is removed.

[0009] Furthermore, the remaining useful life prediction model includes the following structure: The input to the remaining useful life prediction model is a multidimensional feature sequence within a fixed time window, which includes cleaned, time-ordered defects, feature parameters, and historical data throughout the entire life cycle. The front end of the remaining useful life prediction model uses a convolutional layer specifically designed to process the feature data at each time step; the convolutional operation can automatically extract local correlations and spatial patterns between features; and reduce dimensionality to enhance the model's feature representation capabilities. The high-dimensional feature sequences extracted by the convolutional layer are fed into the LSTM layer. The LSTM layer, with its gating mechanism, captures and memorizes the long-term dependencies in the time series, and learns the dynamic laws and trends of the carbon skateboard degradation process. The final hidden state output of the LSTM layer is passed to the fully connected layer, which maps the learned high-level spatiotemporal fusion features to a predicted remaining lifespan of a carbon skateboard. A confidence interval for the predicted remaining lifespan is output by introducing a probability output or Monte Carlo Dropout technique into the fully connected layer.

[0010] Furthermore, the mean absolute error, root mean square error, and coefficient of determination of the predicted remaining useful life are used to quantitatively evaluate and validate the remaining useful life prediction model.

[0011] Furthermore, the selection of polishing strategies based on the predicted remaining service life that meets the confidence threshold includes: Select the target remaining useful life whose confidence interval meets the confidence threshold; If the predicted remaining service life of the target is much greater than the next planned maintenance cycle, a "conservative polishing" strategy will be adopted. The goal of the "conservative polishing" strategy is only to remove the fatigue layer such as surface oxide layer and microcracks, as well as the most uneven wear peaks, so as to maximize the preservation of the carbon skateboard's healthy material. If the predicted remaining service life of the target is close to or less than the next planned maintenance cycle, the "restorative polishing" strategy will be initiated. The goal of the "restorative polishing" strategy is to restore the working profile of the carbon slide to a near-standard geometry as much as possible to ensure that it has excellent flow collection performance during its remaining service life.

[0012] Furthermore, according to the determined grinding strategy, cutting path, depth, and desired cutting force setting, the carbon slide plate cutting includes: The grinding path, cutting depth, and calculated desired cutting force setting value are sent to the PLC control cabinet of the cutting execution equipment. After parsing the instructions, the PLC control cabinet controls the grinding equipment to execute them; The grinding equipment feeds back the actual cutting force to the PLC control cabinet; In each control cycle, the PLC calculates the force deviation between the desired cutting force setpoint and the actual cutting force in real time. The force deviation is immediately input into an optimized PID controller, or an adaptive controller or fuzzy PID controller. The controller calculates the corresponding control compensation signal based on the magnitude, historical accumulation and trend of the force deviation. The compensation signal is then applied to the grinding execution equipment in real time.

[0013] Furthermore, when the force deviation value is greater than 0, it indicates that the actual cutting force is too small. The controller will output a positive compensation signal to increase the feed rate at the end of the execution, thereby increasing the instantaneous cutting amount and causing the actual cutting force to rise back to near the desired cutting force setting value. When the force deviation value is less than 0, it indicates that the actual cutting force is too large. The controller will output a negative compensation signal to reduce the feed rate or perform an instantaneous tool lifting action to reduce the cutting depth, prevent overcutting and protect the tool, and cause the actual cutting force to fall back to the desired cutting force setting value.

[0014] Furthermore, the actual wear data after each carbon skateboard polishing operation is collected as a new training dataset sample for regular incremental training and fine-tuning of the carbon skateboard remaining service life model. This allows the carbon skateboard remaining service life model to adapt to individual differences and changes in operating conditions of different carbon skateboards, thereby continuously improving prediction accuracy.

[0015] Secondly, the present invention provides a life-prediction-based carbon skateboard life-cycle polishing system for implementing a life-cycle polishing method for carbon skateboards, comprising: The data processing module enables the collection and collaborative aggregation of visual and full life-cycle historical data of carbon skateboards; it performs personalized data cleaning and preprocessing for each carbon skateboard based on internal standard scores; it aligns and reconstructs the cleaned high-quality feature parameters and working condition data in strict chronological order, and configures the actual lifespan for the corresponding time points to construct a regular training dataset suitable for the input of time series analysis models. The model training module uses the training dataset to train a model for the remaining lifespan of carbon skateboards. The remaining service life prediction module uses the carbon slide plate remaining service life model to predict the remaining service life. The adaptive grinding execution module, comprising a decision unit, PLC control cabinet, six-dimensional force sensor, servo drive system, and grinding execution mechanism, is used to automatically set grinding parameters based on the predicted remaining service life and a three-dimensional digital model of the carbon slide plate. This includes: selecting a grinding strategy based on the predicted remaining service life that meets a confidence threshold; performing mesh analysis based on the selected grinding strategy and the three-dimensional digital model to calculate the cutting depth of the grinding tool on the planned path; inputting the cutting depth into a pre-calibrated carbon slide plate material removal model to calculate a constant desired cutting force setting; establishing a quantitative relationship between cutting force, feed rate, cutting depth, and material removal rate in the carbon slide plate material removal model; and performing carbon slide plate cutting according to the determined grinding strategy, cutting path, depth, and desired cutting force setting; thus achieving adaptive force-controlled grinding. The full-cycle data management module is used to store all associated data for any carbon skateboard and supports incremental learning of the carbon skateboard's remaining service life model.

[0016] The technical solutions provided in the embodiments of the present invention have the following advantages compared with the prior art: This invention transforms carbon skateboard maintenance decisions from fixed-cycle "planned maintenance" to health-state-based "predictive maintenance" by introducing a data-driven remaining useful life prediction model. This effectively avoids safety hazards caused by insufficient maintenance and resource waste caused by over-maintenance, significantly improving the availability and life-cycle value of carbon skateboards.

[0017] This application configures the grinding process based on remaining service life prediction results. By accurately predicting service life and optimizing maintenance plans, this invention reduces unnecessary disassembly and replacement. Adaptive grinding reduces material waste and scrap rates, while high-quality grinding improves current collection quality, reduces energy consumption and contact wire wear, ultimately resulting in significant overall cost reduction and improved operational safety. Furthermore, it employs a closed-loop control strategy based on real-time force feedback to achieve "constant force grinding," enabling the system to handle uncertainties such as material unevenness and irregular contours. This not only completely eliminates over-cutting and under-grinding, ensuring extremely high grinding quality consistency and surface morphology accuracy, but also protects grinding tools and improves operational safety.

[0018] Specifically, this invention provides a data-driven integrated solution for predictive maintenance and adaptive precision operation, used to optimize the maintenance cycle of conductive carbon sliding plates, improve maintenance quality, and extend their service life. Through multi-source data fusion, advanced algorithm models, and real-time closed-loop control, it achieves accurate prediction of the remaining service life of the carbon sliding plate and executes personalized, high-precision adaptive constant-force grinding accordingly. Ultimately, this achieves the goals of extending the service life of the carbon sliding plate, ensuring current collection quality and operational safety, and reducing the total lifecycle maintenance cost. Attached Figure Description

[0019] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 A flowchart of a carbon skateboard full life cycle polishing method based on life prediction provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of a grinding execution device provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of a carbon skateboard full life cycle polishing system based on life prediction, provided as an embodiment of the present invention. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0024] Example 1 like Figure 1 As shown, this invention provides a carbon skateboard lifecycle polishing method based on lifespan prediction. The specific process includes: Visual and full life-cycle historical data collection and collaborative aggregation of carbon skateboards.

[0025] In the application of carbon skateboards, a two-dimensional high-resolution industrial camera and a three-dimensional line laser scanner are used to acquire high-definition images of the carbon skateboard surface and high-precision three-dimensional point cloud data, respectively; and to acquire historical life cycle data of the carbon skateboard.

[0026] Images and 3D point cloud data of the carbon fiber slide plate are acquired using a vision system (including high-resolution industrial cameras and 3D laser scanners). The carbon fiber slide plate images are used to identify macroscopic surface defects such as cracks and chipping; the 3D point cloud data is used to accurately reconstruct the 3D digital model of the carbon fiber slide plate, precisely determining its current geometric parameters, including current contour, current wear depth, and current wear distribution. Historical data of the carbon fiber slide plate's entire lifecycle is obtained from the enterprise maintenance management system via a data interface, including but not limited to: records of each grinding operation (grinding time, grinding amount), cumulative mileage, route information (such as route slope, curve conditions, cable conditions), and operating current. Different routes, varying route slopes, curve conditions, and cable conditions all affect the normal wear and tear of the carbon fiber slide plate, influencing its lifespan degradation rate.

[0027] All these multimodal and multi-temporal-scale data are uniformly aggregated into a central database, providing comprehensive information support for subsequent analysis.

[0028] Data-driven cleaning and alignment reconstruction.

[0029] Personalized data cleaning and preprocessing are performed on each carbon skateboard based on internal standard scores. This includes: considering the individual differences of each carbon skateboard, for any characteristic parameter of the carbon skateboard obtained from S1, such as the currently measured maximum wear depth among the geometric characteristic parameters, the internal standard score of that characteristic parameter is calculated. The calculation formula is as follows: ;in, The characteristic value of this characteristic parameter of the carbon skateboard was measured in this study. σ represents the average value of this characteristic parameter in the historical data of this specific carbon skateboard, indicating its normal wear level; σ is the standard deviation of the corresponding historical data, reflecting its normal fluctuation range. A score threshold is set; when the absolute value of the internal standard score of a data point exceeds the corresponding score threshold, it is determined to be an outlier that significantly deviates from its historical normal level and is removed. Anomalies may originate from measurement noise, temporary attachments, etc.; compared to using a global threshold, this method can more accurately identify the true anomalies of individual carbon skateboards.

[0030] The cleaned high-quality feature parameters and operating condition data are aligned and reconstructed in strict chronological order, and the actual lifespan is configured for the corresponding time points to construct a well-organized, high-quality training dataset suitable for input to time series analysis models.

[0031] A model for predicting the remaining lifespan of carbon skateboards is trained using a training dataset. This invention designs a model for predicting the remaining lifespan of carbon skateboards that integrates convolutional neural networks and long short-term memory networks. The carbon skateboard remaining lifespan prediction model includes the following structure: The input to the carbon skateboard remaining service life prediction model is a multidimensional feature sequence within a fixed time window, which includes cleaned, time-ordered defects, feature parameters, and full life cycle historical data.

[0032] The front end of the carbon skateboard remaining life prediction model uses a convolutional layer specifically designed to process the feature data at each time step. Convolutional operations can automatically extract local correlations and spatial patterns between features (such as the correlation of wear at different locations), effectively reducing dimensionality and enhancing the model's feature representation capabilities.

[0033] The high-dimensional feature sequences extracted by the convolutional layers are fed into the LSTM layer. The LSTM layer, with its internal gating mechanism, effectively captures and remembers long-term dependencies in the time series, learning the dynamic patterns and trends of carbon skateboard degradation. The LSTM layer includes an input gate, a forget gate, and an output gate.

[0034] The final hidden state output of the LSTM layer is passed to a fully connected layer. The fully connected layer maps the learned high-level spatiotemporal fusion features to a predicted remaining lifespan of a carbon skateboard, which can be expressed as remaining kilometers or remaining days. Furthermore, by introducing a probabilistic output or Monte Carlo Dropout technique into the fully connected layer to output a confidence interval for the predicted remaining lifespan, an uncertainty measure is provided for decision-making.

[0035] The mean absolute error, root mean square error, and coefficient of determination between the remaining useful life and the actual useful life are used to quantitatively evaluate and validate the remaining useful life prediction model.

[0036] ; ; ; in, This represents the predicted lifetime of the j-th sample. This represents the true lifetime of sample j. The mean absolute error (MAE) represents the average value of the actual lifespan, and N represents the number of test samples. A smaller MAE value corresponds to a smaller RMSE value, and a smaller coefficient of determination (R²) corresponds to a smaller root mean square error (RMSE). 2 The larger the value, the better the prediction results of the remaining useful life prediction model.

[0037] For any carbon skateboard to be processed, its data is obtained in the form of a training dataset. The current predicted remaining lifespan is then predicted using the trained carbon skateboard remaining lifespan model. The process involves automating the grinding settings by combining the remaining service life with a 3D digital model of the carbon skateboard; this process serves as an intelligent decision-making bridge connecting service life prediction and grinding execution.

[0038] For any carbon skateboard to be processed, receive the remaining useful life prediction results from the remaining useful life prediction model and the reconstructed three-dimensional digital model of the carbon skateboard.

[0039] Grinding strategy selection is based on the predicted remaining service life that meets the confidence threshold: Select a target remaining service life that meets the confidence threshold. If the predicted target remaining service life is ample (e.g., much longer than the next planned maintenance cycle), a "conservative polishing" strategy is adopted. The goal of this strategy is solely to remove fatigue layers such as surface oxide layers and microcracks, as well as the most uneven wear peaks, maximizing the preservation of healthy material. If the predicted target remaining service life is nearing its end (e.g., close to or less than the next planned maintenance cycle), a "restorative polishing" strategy is initiated. The goal of this strategy is to restore the working profile of the carbon slide to a near-standard geometry as much as possible to ensure excellent flow collection performance throughout its remaining service life.

[0040] Based on the selected grinding strategy, combined with the three-dimensional digital model, a mesh analysis is performed to calculate the material depth (cutting depth) that the grinding tool needs to remove at each point on the planned path.

[0041] Subsequently, based on the carbon slide material removal model obtained through extensive pre-experiment calibration (which establishes a quantitative relationship between cutting force, feed rate, depth of cut, and material removal rate), the calculated depth value is converted into a constant desired cutting force setpoint. . The settings are designed to achieve efficient and smooth material removal, while ensuring a smooth cutting process and avoiding tool overload or chatter.

[0042] Control signal issuance and initialization: setting the grinding path, cutting depth, and calculated desired cutting force. Instructions are sent to the PLC control cabinet of the cutting equipment via industrial Ethernet or fieldbus. After parsing the instructions, the PLC control cabinet controls the grinding equipment to execute them. The servo driver of the grinding equipment drives the servo motor, which, through a precision transmission screw or linkage mechanism, drives a six-axis industrial robot or gantry equipped with a force-controlled grinding head to move along the planned path. At the same time, the PLC sends instructions to start the spindle motor (driving the grinding tool head to rotate) and the dust removal equipment.

[0043] like Figure 2 As shown, the grinding execution equipment consists of a worktable 1, a pneumatic clamping device 2, a grinding actuator 3, a dust collection device 4, a vision system 5, a compressor 6, and a PLC control cabinet 7. The worktable 1 provides a platform for positioning and fixing the carbon slide plate during processing; the pneumatic clamping device 2 includes a chuck, a cylinder, and a fixing table to clamp and fix the carbon slide plate; the compressor 6 provides an air source for the clamping device; the grinding actuator 3 meets the surface grinding requirements of the carbon slide plate and can achieve X-direction (carbon slide plate length direction), Y-direction (carbon slide plate thickness direction), and cutter head tilt angle adjustment on the worktable (used for processing the chamfers on both sides of the carbon slide plate); the dust collection device 4 is used to promptly clean up the powder generated during processing.

[0044] A high-bandwidth, high-precision six-dimensional force / torque sensor is directly mounted between the grinding head of the grinding actuator 3 and the end flange of the robot. The six-dimensional force / torque sensor collects triaxial force (Fx, Fy, Fz) and triaxial torque data in real time during the grinding process at a frequency higher than 1000Hz. The normal force perpendicular to the grinding surface of the carbon slide plate (usually Fz) is extracted as a reflection of the actual cutting force. The main control signal. The actual cutting force. After the signal is processed by digital signal processing techniques such as low-pass filtering to suppress noise, it is transmitted in real time and at high speed to the input module of the PLC in PLC control cabinet 7.

[0045] The PLC calculates the force deviation between the desired cutting force setpoint and the actual cutting force in real time within each control cycle (typically 1 ms or less): e = - The force deviation value e is immediately input into a tuned PID (proportional-integral-derivative) controller, or a more advanced adaptive controller or fuzzy PID controller. Based on the magnitude of the force deviation value e (proportional term), its historical accumulation (integral term), and its changing trend (derivative term), the controller calculates the corresponding control compensation signal (usually an adjustment in speed or position). This compensation signal is applied to the servo drive system in real time, dynamically and continuously adjusting the motion trajectory or feed rate of the actuator end, thereby achieving precise and continuous control of the grinding force under complex actual working conditions, ensuring that the actual cutting force closely follows the desired cutting force setpoint.

[0046] When e > 0, it indicates that the actual cutting force is too low, possibly due to encountering pits or areas with softer material. The controller will output a positive compensation signal, instructing the servo system to appropriately increase the feed rate at the end of the process, thereby increasing the instantaneous cutting depth and bringing the actual cutting force back to near the desired cutting force setting. When e < 0, it indicates that the actual cutting force is too high, possibly due to encountering protrusions, hard spots, or areas with high material density. The controller will output a negative compensation signal, instructing the servo system to reduce the feed rate or perform a small, instantaneous tool lift-off action to reduce the depth of cut, prevent overcutting, and protect the tool, bringing the actual cutting force back to the desired cutting force setting.

[0047] Throughout the grinding process, the real-time curves of the desired cutting force setpoint and the filtered actual cutting force are simultaneously displayed on the human-machine interface for operator monitoring. For a high-performance constant force control system, the two curves should essentially overlap, with only minor, rapid fluctuations under high dynamic disturbances. All force control process data, including force curves, PID outputs, servo adjustment records, and alarm events, are uniquely associated with this operation of the carbon slide plate and are automatically uploaded to the full-cycle data management subsystem for permanent archiving. This detailed process data provides data support for subsequent process parameter optimization, tool life prediction, and system fault diagnosis.

[0048] The actual wear data after each polishing operation is collected and used as new training samples for regular incremental training and fine-tuning of the carbon skateboard remaining service life model. This enables the carbon skateboard remaining service life model to adapt to individual differences and changes in operating conditions of different carbon skateboards, thereby continuously improving prediction accuracy.

[0049] Example 2 like Figure 3 As shown, this invention provides a carbon skateboard lifecycle grinding system based on lifecycle prediction for implementing a carbon skateboard lifecycle grinding method, comprising: The data processing module enables the collection and collaborative aggregation of visual and full life-cycle historical data of carbon skateboards; it performs personalized data cleaning and preprocessing for each carbon skateboard based on internal standard scores; it aligns and reconstructs the cleaned high-quality feature parameters and working condition data in strict chronological order, and configures the actual lifespan for the corresponding time points to construct a regular training dataset suitable for the input of time series analysis models. The model training module uses the training dataset to train a model for the remaining lifespan of carbon skateboards. The remaining service life prediction module uses the carbon slide plate remaining service life model to predict the remaining service life. The adaptive grinding execution module, comprising a decision unit, PLC control cabinet, six-dimensional force sensor, servo drive system, and grinding execution mechanism, is used to automatically set grinding parameters based on the predicted remaining service life and a three-dimensional digital model of the carbon slide plate. This includes: selecting a grinding strategy based on the predicted remaining service life that meets a confidence threshold; performing mesh analysis based on the selected grinding strategy and the three-dimensional digital model to calculate the cutting depth of the grinding tool on the planned path; inputting the cutting depth into a pre-calibrated carbon slide plate material removal model to calculate a constant desired cutting force setting; establishing a quantitative relationship between cutting force, feed rate, cutting depth, and material removal rate in the carbon slide plate material removal model; and performing carbon slide plate cutting according to the determined grinding strategy, cutting path, depth, and desired cutting force setting; thus achieving adaptive force-controlled grinding. The full-cycle data management module stores all relevant data for any carbon skateboard and supports incremental learning of the remaining service life model. This module assigns a unique electronic ID to each carbon skateboard and is built upon an industrial real-time database and a relational database. It centrally stores all process data, model parameters, and operation records, and provides version management and data service support for incremental model learning.

[0050] In the embodiments provided by this invention, it should be understood that the disclosed structures and methods can be implemented in other ways. For example, the structural embodiments described above are merely illustrative. For instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection through some interfaces, structures, or units, and may be electrical, mechanical, or other forms.

[0051] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0052] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0053] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features claimed herein.

Claims

1. A method for polishing carbon skateboards throughout their entire lifecycle based on lifespan prediction, characterized in that, include: The process involves the visual and full lifecycle historical data acquisition and collaborative aggregation of carbon fiber skateboards. Specifically, industrial cameras and 3D laser scanners are used to acquire images and 3D point cloud data of the carbon fiber skateboards. The images are used to identify macroscopic surface defects. The 3D point cloud data is used to reconstruct the 3D digital model of the carbon fiber skateboard, determining its current geometric parameters, including current contour, current wear depth, and current wear distribution. The entire lifecycle historical data of the carbon fiber skateboards is obtained from the enterprise maintenance management system via a data interface. This data includes: records of each polishing operation, including polishing time and amount; cumulative mileage; route information, including route gradient, curve conditions, and cable conditions; and operating current. Personalized data cleaning and preprocessing are performed on each carbon skateboard based on internal standard scores, including: for each individual carbon skateboard, for any characteristic parameter of the carbon skateboard, the internal standard score of that characteristic parameter is calculated; a score threshold is set, and when the absolute value of the internal standard score of a data point exceeds the corresponding score threshold, it is determined to be an outlier that deviates from its own historical normal level and is removed. The cleaned high-quality feature parameters and operating condition data are aligned and reconstructed in strict chronological order, and the actual lifespan is configured for the corresponding time points to construct a regular training dataset suitable for the input of the time series analysis model; the training dataset is used to train the carbon skateboard remaining lifespan model. For any carbon skateboard to be processed, its data is obtained in the form of a training dataset, and its current predicted remaining lifespan is predicted using a trained carbon skateboard remaining lifespan prediction model; wherein, the remaining lifespan prediction model includes the following structure: The input to the remaining useful life prediction model is a multidimensional feature sequence within a fixed time window, which includes cleaned, time-ordered defects, feature parameters, and historical data throughout the entire life cycle. The front end of the remaining useful life prediction model uses a convolutional layer specifically designed to process the feature data at each time step; the convolutional operation can automatically extract local correlations and spatial patterns between features; and reduce dimensionality to enhance the model's feature representation capabilities. The high-dimensional feature sequences extracted by the convolutional layer are fed into the LSTM layer. The LSTM layer, with its gating mechanism, captures and memorizes the long-term dependencies in the time series, and learns the dynamic laws and trends of the carbon skateboard degradation process. The final hidden state output of the LSTM layer is passed to the fully connected layer, which maps the learned high-level spatiotemporal fusion features to a predicted remaining lifespan of a carbon skateboard; confidence intervals for the predicted remaining lifespan are output by introducing probabilistic outputs or Monte Carlo Dropout techniques into the fully connected layer. Automated grinding settings are achieved by combining the predicted remaining service life with a 3D digital model of the carbon slide plate. This includes: selecting a grinding strategy based on the predicted remaining service life that meets a confidence threshold; performing mesh analysis based on the selected grinding strategy and the 3D digital model to calculate the cutting depth of the grinding tool along the planned path; inputting the cutting depth into a pre-calibrated carbon slide plate material removal model to calculate a constant desired cutting force setting; and establishing a quantitative relationship between cutting force, feed rate, cutting depth, and material removal rate using the carbon slide plate material removal model. Carbon slide plate cutting is performed according to the grinding strategy, cutting path, depth, and expected cutting force settings determined by the grinding settings.

2. The carbon skateboard full life cycle polishing method based on life prediction according to claim 1, characterized in that, The mean absolute error, root mean square error, and coefficient of determination of the predicted remaining useful life are used to quantitatively evaluate and validate the remaining useful life prediction model.

3. The carbon skateboard full life cycle polishing method based on life prediction according to claim 1, characterized in that, Grinding strategy selection based on predicted remaining service life that meets confidence thresholds includes: Select the target remaining useful life whose confidence interval meets the confidence threshold; If the predicted remaining service life of the target is much greater than the next planned maintenance cycle, a "conservative polishing" strategy will be adopted. The goal of the "conservative polishing" strategy is only to remove the surface oxide layer, microcracks, fatigue layer and the most uneven wear peaks, so as to maximize the preservation of the carbon skateboard's healthy materials. If the predicted remaining service life of the target is close to or less than the next planned maintenance cycle, the "restorative polishing" strategy will be initiated. The goal of the "restorative polishing" strategy is to restore the working profile of the carbon slide to a near-standard geometry as much as possible to ensure that it has excellent flow collection performance during its remaining service life.

4. The carbon skateboard full life cycle polishing method based on life prediction according to claim 1, characterized in that, The carbon slide plate is cut according to the determined grinding strategy, cutting path, depth, and desired cutting force setting, including: The grinding path, cutting depth, and calculated desired cutting force setting value are sent to the PLC control cabinet of the cutting execution equipment. After parsing the instructions, the PLC control cabinet controls the grinding equipment to execute them; The grinding equipment feeds back the actual cutting force to the PLC control cabinet; In each control cycle, the PLC calculates the force deviation between the desired cutting force setpoint and the actual cutting force in real time. The force deviation is immediately input into an optimized PID controller, or an adaptive controller or fuzzy PID controller. The controller calculates the corresponding control compensation signal based on the magnitude, historical accumulation and trend of the force deviation. The compensation signal is then applied to the grinding execution equipment in real time.

5. The carbon skateboard full life cycle polishing method based on life prediction according to claim 4, characterized in that, When the force deviation value is greater than 0, it indicates that the actual cutting force is too small. The controller will output a positive compensation signal to increase the feed rate at the end of the execution, thereby increasing the instantaneous cutting amount and causing the actual cutting force to rise back to near the desired cutting force setting value. When the force deviation value is less than 0, it indicates that the actual cutting force is too large. The controller will output a negative compensation signal to reduce the feed rate or perform an instantaneous tool lifting action to reduce the cutting depth, prevent overcutting and protect the tool, and cause the actual cutting force to fall back to the desired cutting force setting value.

6. The carbon skateboard full life cycle polishing method based on life prediction according to claim 1, characterized in that, The actual wear data after each carbon skateboard grinding operation is collected as a new training dataset sample, which is used to perform regular incremental training and fine-tuning of the carbon skateboard remaining service life model. This allows the carbon skateboard remaining service life model to adapt to individual differences and changes in operating conditions of different carbon skateboards, thereby continuously improving the prediction accuracy.

7. A life-cycle polishing system for carbon skateboards based on life prediction for implementing the method of any one of claims 1-6, characterized in that, include: The data processing module is responsible for the visual and full lifecycle historical data acquisition and collaborative aggregation of carbon skateboards; Personalized data cleaning and preprocessing are performed on each carbon skateboard based on internal standard scores; The cleaned high-quality feature parameters and operating condition data are aligned and reconstructed in strict chronological order, and the actual lifespan is configured for the corresponding time points to construct a regular training dataset suitable for the input of time series analysis models. The model training module uses the training dataset to train the carbon skateboard's remaining service life model; the remaining service life prediction module uses the carbon skateboard's remaining service life model to predict the remaining service life. The adaptive grinding execution module, comprising a decision unit, PLC control cabinet, six-dimensional force sensor, servo drive system, and grinding execution mechanism, is used to automatically set grinding parameters based on the predicted remaining service life and a three-dimensional digital model of the carbon slide plate. This includes: selecting a grinding strategy based on the predicted remaining service life that meets a confidence threshold; performing mesh analysis based on the selected grinding strategy and the three-dimensional digital model to calculate the cutting depth of the grinding tool on the planned path; inputting the cutting depth into a pre-calibrated carbon slide plate material removal model to calculate a constant desired cutting force setting; establishing a quantitative relationship between cutting force, feed rate, cutting depth, and material removal rate in the carbon slide plate material removal model; and performing carbon slide plate cutting according to the determined grinding strategy, cutting path, depth, and desired cutting force setting; thus achieving adaptive force-controlled grinding. The full-cycle data management module is used to store all associated data for any carbon skateboard and supports incremental learning of the carbon skateboard's remaining service life model.

Images