Piston pin polishing monitoring system

Abstract

The invention belongs to the technical field of precision machining, and particularly relates to a piston pin polishing monitoring system which is characterized in that data such as a temperature field, stress and surface morphology are acquired through a multi-dimensional sensing array, and thermal deformation deviation is corrected in combination with a multi-physics field coupling model; a self-adaptive decision framework is constructed by adopting deep reinforcement learning DRL, and a polishing strategy is optimized through digital twinborn verification; lIBS material identification and multi-process linkage are integrated, and rapid adaptation of 12 types of materials is achieved; real-time control and global optimization are balanced based on an edge-cloud collaborative architecture; environment-friendly safety monitoring is added, and the green manufacturing standard is met. The system solves the problems that a traditional system is low in precision, poor in adaptability and high in cost, the surface roughness Ra of the piston pin is smaller than or equal to 0.8 micrometer, the percent of pass is increased to 99%, and the system is suitable for batch production of high-precision piston pins.

Description

Technical Field

[0001] This invention belongs to the field of precision machining technology, specifically relating to a piston pin grinding monitoring system. Background Technology

[0002] As a critical force-transmitting component in an engine, the piston pin's surface roughness must be ≤0.8μm, its dimensional accuracy tolerance ≤±0.001mm, and its mechanical properties directly affect the engine's operating efficiency and lifespan. Existing piston pin grinding and monitoring technologies have the following key shortcomings: 1. Lack of multiphysics coupling analysis Traditional systems only monitor single parameters such as temperature and vibration independently, without considering the dynamic coupling relationship between the local high temperature field (up to 150℃) and the contact stress field (up to 500MPa) and the surface morphology during the grinding process. This results in thermal deformation error ≥0.002mm or workpiece cracking rate of up to 3%-5% caused by stress overload.

[0003] 2. Adaptive strategies are rigid and have weak generalization ability. Relying on preset thresholds or fixed algorithms such as PID control, it has poor adaptability to piston pins made of different materials such as aluminum alloy and ceramics, and the pass rate of special material workpieces is only 60%-70%; moreover, it cannot optimize parameters independently through process iteration, and the response lag is ≥100ms when facing abnormal working conditions such as sudden wear of grinding wheels.

[0004] 3. Insufficient multi-process collaboration and material identification It only covers a single grinding wheel process and does not link subsequent polishing and honing processes. The process conversion error is ≥0.001mm. It lacks a material intelligent recognition mechanism, requires manual parameter switching, and the changeover time is ≥30 minutes, which is difficult to meet the needs of flexible production lines.

[0005] 4. Imbalance between cost and synergy High-precision systems rely on imported sensors, such as laser scanners, which account for 60% of the cost, with a single system costing ≥ 200,000 yuan. Furthermore, they employ local closed-loop control, which cannot achieve global optimization of process parameters across multiple devices, resulting in batch processing consistency deviations ≥ 0.003 mm.

[0006] 5. Gap in environmental and safety monitoring It does not address environmental issues such as dust concentrations exceeding 5mg / m³ and noise levels ≥90dB during the grinding process, nor does it address safety hazards caused by workpiece clamping loosening force deviations ≥20%, thus failing to meet the requirements of standards such as GBZ2.1-2019 and ISO12100.

[0007] Therefore, there is an urgent need to build a new generation of piston pin grinding monitoring system that integrates multi-dimensional perception, intelligent decision-making, flexible adaptation and green safety features. Summary of the Invention

[0008] This invention aims to address the problems of existing piston pin grinding monitoring systems, such as insufficient consideration of multi-physics coupling effects, weak generalization ability of adaptive strategies, poor multi-process coordination, high cost, and lack of environmental protection and safety. It provides a high-precision, highly adaptable, low-cost intelligent monitoring system that meets environmental protection and safety standards.

[0009] This invention discloses a piston pin grinding monitoring system, including a data acquisition module, a data analysis and intelligent decision-making module, a control execution module, a human-computer interaction module, and an edge-cloud collaborative platform. These modules work together to achieve precise control over the entire piston pin grinding process. The specific structure is as follows: 1. Data Acquisition Module This module achieves comprehensive perception of the polishing process through a multi-dimensional sensor array, including: Basic parameter sensing unit Displacement sensor: A Keyence LK-G80 laser displacement sensor with a resolution of 0.01μm and a sampling frequency of 1kHz is used. It is installed at the end of the grinding wheel feed shaft to monitor the feed displacement accuracy of the grinding wheel relative to the piston pin in real time, which is ±0.0005mm, and to calculate the grinding depth. Temperature sensor: Uses type K thermocouple, with a temperature measurement range of -50~300℃ and an accuracy of ±0.5℃. It is evenly distributed in 3 points and attached to the piston pin surface with an axial spacing of 5mm to collect instantaneous temperature data. Vibration sensor: A PCB356A16 accelerometer with a range of ±50g and a frequency response of 0.5-10kHz is used. It is installed on the bottom of the workbench to collect the vibration acceleration signal of the equipment.

[0010] Multiphysics sensing unit Infrared thermal imager: The FLIRA655sc is used with a resolution of 640×512 and a thermal sensitivity of ≤30mK. It is installed 300mm above the grinding station and generates 10 frames of temperature field distribution images per second. The temperature gradient resolution is ±0.1℃ / mm. Micro-strain sensor: The foil strain gauge has an accuracy of ±1με and a response time of ≤1ms. It is attached to the non-machined area of ​​the piston pin and monitors the strain value in real time during the grinding process, ranging from -200 to +200με. Laser scanner: The domestically produced Huake Precision HK-3D02 scanner has an accuracy of ±0.002mm and a scanning speed of 2000 points / second, acquiring three-dimensional point cloud data of the piston pin surface with a point cloud density of ≥100 points / mm².

[0011] Material and process sensing unit Laser-induced breakdown spectroscopy (LIBS): Employs the ProGenius system from Ocean Optics, with a wavelength range of 200-1000nm, a resolution of 0.1nm, and is mounted 100mm above the workpiece. It analyzes surface elemental composition with a recognition time of ≤2 seconds and an accuracy of 98%. Roughness sensor: The Taylor Hopson Surtronic S-100 is used, with a measurement range of 0.02-10 μm Ra and an accuracy of ±5%, to collect surface roughness data in real time.

[0012] Environmental safety sensing unit Dust concentration sensor: The Panten G7 sensor has a detection range of 0-10 mg / m³ and an accuracy of ±0.1 mg / m³. It is installed 500 mm to the side of the grinding station. Noise sensor: AWA5636 with a range of 30-130dB and an accuracy of ±1dB is used to monitor the noise level at the workstation. Piezoelectric sensor: PiezoSystems PSV-100 with a range of 0-1000N and an accuracy of ±2%, integrated into the workpiece fixture to detect clamping force.

[0013] 2. Data Analysis and Intelligent Decision-Making Module This module performs fusion processing and intelligent analysis on multi-source data, and outputs optimized control strategies, including: Multiphysics Coupling Analysis Submodule Data preprocessing: Gaussian filtering with σ=1.5 was applied to the temperature field image to remove noise; the strain signal was subjected to wavelet transform with db4 wavelet basis and 5-level decomposition to extract feature frequencies. Coupled Model Construction: A coupled model of "temperature-stress-surface morphology" was established based on the finite element analysis (FEA) software Abaqus, with the following formula: σ(x,y,t)=E×[ε(x,y,t+α×ΔT(x,y,t)] Where σ is the stress in MPa, E is the elastic modulus of the material (69 GPa for aluminum alloy), ε is the strain, and α is the coefficient of thermal expansion (23 × 10⁻⁶ for aluminum alloy). -6 / ℃, where ΔT is the temperature difference at point (x,y) at time t; Deviation Correction: When the calculated temperature gradient is greater than 5℃ / mm or the stress exceeds 10% of the material's yield strength (e.g., the yield strength of 45 steel is 355MPa, and the threshold is set to 390MPa), the calculated surface curvature value will be automatically corrected by a correction amount ≤0.0005mm. -1 .

[0014] Reinforcement Learning Adaptive Decision Submodule DRL Framework: Employs the Deep Deterministic Policy Gradient (DDPG) algorithm to construct an Actor-Critic dual network structure. Actor network: Input a 12-dimensional state vector including temperature, stress, roughness, etc., and output a continuous motion space grinding force of 0.1-1MPa and a rotation speed of 1000-5000r / min; Critic network: The reward function is "surface accuracy compliance rate weight 0.6 + energy consumption reduction rate 0.2 + grinding wheel life extension rate 0.2" to evaluate the value of actions; Strategy iteration: An iteration is triggered every 100 workpieces processed. The experience replay pool stores 5000 sets of samples. The target network soft update coefficient is set to 0.001. Virtual verification: Based on Unity3D, a digital twin environment is built. The new strategy requires a pass rate of >95% in 1000 virtual polishing tests, and the average trial and error cost is reduced by 70%.

[0015] Material identification and process matching submodule Material classification: After dimensionality reduction by principal component analysis (PCA) of LIBS spectral data, it was input into a support vector machine (SVM) classifier, which identified 12 types of materials, including ceramics and composite materials, with an accuracy of 98%. Process library matching: 12 preset process parameter templates, such as "diamond grinding wheel + 0.2MPa force + 2000r / min" for ceramic materials, can complete parameter self-configuration within 1 minute, including grinding wheel type, feed speed, cooling flow rate, etc.

[0016] Multi-process collaborative analysis submodule Process switching judgment: When the real-time roughness Ra > 0.4μm, the polishing process switching signal is triggered; when Ra ≤ 0.4μm and the roundness error > 0.0005mm, the honing machine is linked for finishing. Error compensation: The process conversion error is calculated based on laser scanning data to be ≤0.0005mm, and corrected by the feed axis compensation amount of ±0.0003mm.

[0017] Environmental safety analysis submodule Threshold determination: When the dust concentration is >2mg / m³, the high air volume mode of the dust removal equipment is triggered, and the air volume is increased by 50%; when the noise is >85dB, the noise is controlled by reducing the grinding wheel speed by ≤20%, while ensuring that the surface roughness does not deteriorate. Safety warning: When the clamping force is less than 80% of the design value (e.g., if the design value is 1000N and the threshold is 800N), a stop signal will be output immediately with a response time of ≤10ms; when the grinding wheel vibration spectrum deviates from the normal range by ±10Hz, the risk of cracking will be predicted and an alarm will be triggered.

[0018] 3. Control Execution Module This module drives the actuators based on the decision results to achieve precise control, including: Basic parameter adjustment unit Grinding wheel replacement: When the wear rate is >0.01mm / min, the automatic grinding wheel changing mechanism should have a tool changing time of ≤15 seconds. Temperature control: When the temperature is >100℃, adjust the cooling system flow rate to increase by 30% and reduce the feed rate by 20%; Vibration suppression: When the vibration spectrum is abnormal, the spindle frequency is adjusted by ±5Hz through the frequency converter to avoid resonance.

[0019] Adaptive Execution Unit Piezoelectric drive feed: The PIE-509 piezoelectric controller is used to achieve fine adjustment of the grinding wheel feed with a resolution of 0.01μm and a response time of 20ms. Multi-axis linkage: 6-axis robotic arm ABBIRB1200, repeatability ±0.01mm, adjustable grinding angle range 0-90°, accuracy ±0.5°, adaptable to complex structures such as piston pin end fillet.

[0020] Multi-process linkage unit Equipment switching: The PLC-linked grinding wheel machine and ultrasonic polishing machine operate at a frequency of 20-40kHz, with a process switching time of ≤5 seconds; Parameter coordination: During the polishing stage, the pressure is automatically reduced by 0.1-0.3MPa, the amplitude is increased by 5-10μm, and Ra≤0.2μm is ensured.

[0021] Environmental safety implementation unit Dust control: When dust levels exceed the standard, the air volume is increased from 1000m³ / h to 1500m³ / h using a variable frequency fan; Emergency stop: When the clamping force is abnormal, the servo motor brake is triggered for a time of ≤50ms and the warning light is illuminated.

[0022] 4. Human-Computer Interaction Module This module provides a visual user interface and a manual intervention channel, including: Real-time monitoring interface: Displays 20 key parameters such as temperature field distribution thermal map, stress cloud map, and surface roughness curve, with a refresh rate of 10Hz; Parameter configuration module: Supports manual setting of material thresholds such as yield strength, coefficient of thermal expansion, and grinding strategy weights such as precision priority / energy consumption priority mode; Alarm and Log Unit: Records equipment anomalies such as sensor failures and parameter overruns, stores one year of historical data, supports Excel export, and allows traceability of the processing parameters for each batch of piston pins.

[0023] 5. Edge-Cloud Collaboration Platform Edge nodes: Employ Advantech IPC-610 industrial PC with Intel i5 processor, responsible for real-time control with a response time ≤20ms, and running DRL inference models and multiphysics coupled calculations; Cloud platform: Deployed on Alibaba Cloud ECS server, it aggregates data from 100+ devices and trains a global model using the FedAvg algorithm, iterating once every 24 hours; Global optimization: When the diameter deviation of a batch of workpieces is greater than 0.002mm, a unified feed rate correction scheme of ±0.0005mm is pushed, improving batch consistency by 30%.

[0024] Monitoring methods and procedures Initialization phase Sensor calibration: The laser scanner is calibrated with a standard gauge block accuracy of ±0.0001mm, and the temperature measurement error is corrected to ±0.5℃; System configuration: Import workpiece parameters: diameter 3-50mm, length 20-100mm, material type; initialize DRL model parameters. Network synchronization: The clock is calibrated between the edge and the cloud via 5G Time-Sensitive Networking (TSN), with a synchronization error of ≤0.1ms.

[0025] Data acquisition and preprocessing stage Multi-source data synchronous acquisition: Basic parameters such as displacement and temperature are synchronized with infrared thermal imaging and LIBS spectral data via TSN, with a sampling frequency of 1kHz; Data cleaning: Point cloud data is denoised by removing points outside the mean 3σ using statistical filtering, and vibration signals are cleaned by removing 50Hz power frequency interference.

[0026] Intelligent analysis and decision-making stage Multiphysics coupling calculation: outputs stress field and temperature field distribution, corrects surface curvature values ​​with an accuracy of ±0.0005mm. - ¹; DRL strategy generation: Outputs optimal grinding parameters based on current temperature, stress, and roughness. Material identification and process matching: After classifying LIBS data, the corresponding process template is called; Environmental safety assessment: Real-time monitoring of dust, noise, and clamping force, and output of early warning or control signals.

[0027] Control Execution Phase Priority control: Safety signals such as insufficient clamping force are executed first, with a response time ≤10ms, followed by process parameter adjustments; Multi-process linkage: Switching between grinding / polishing processes based on roughness data, and synchronously compensating for conversion errors; Edge-cloud collaboration: Edge nodes perform real-time control, while the cloud records data and optimizes the global model.

[0028] Feedback and Iteration Phase Effect evaluation: After polishing, laser scanning detection showed that the surface roughness Ra ≤ 0.8 μm and the dimensional tolerance ± 0.001 mm; Model update: Feed the actual processing data error value and energy consumption back to the DRL model, and iterate the strategy once every 100 workpieces; Log storage: Process parameters and test results are saved in the cloud, supporting quality traceability and a retention period of 3 years.

[0029] This process achieves high precision, high stability, and continuous optimization in piston pin grinding through a closed-loop control system across all stages and an intelligent iteration mechanism. During the initialization phase, precise calibration and synchronization errors ≤0.1ms lay the data foundation for the entire process. The data acquisition and analysis phase combines multiphysics coupling to calculate curvature accuracy of ±0.0005mm-¹, DRL strategy generation, and material matching to ensure the scientific nature of decision-making. During the control execution phase, safety is prioritized and a response time of ≤10ms is achieved. Multiple processes are linked and errors are compensated to ensure machining accuracy. During the feedback and iteration phase, the strategy is continuously optimized through effect evaluation and model updates every 100 pieces. Combined with cloud-based traceability of 3 years of data, stable control of surface roughness Ra≤0.8μm and dimensional tolerance±0.001mm is finally achieved, while improving the traceability and long-term adaptability of the process.

[0030] Compared with the prior art, the beneficial effects of the present invention are: 1. Significantly improved machining accuracy and quality stability Based on the multiphysics coupling model σ=E×(ε+α×ΔT), thermal deformation deviation is corrected in real time, and the surface curvature measurement accuracy reaches ±0.0005mm. - ¹ Dimensional tolerances are controlled within ±0.0005mm, a 50% improvement over traditional systems; The scrap rate caused by thermal deformation and stress overload has been reduced from 3%-5% to below 0.5%, and the surface roughness of piston pins made of special materials such as ceramic coatings has been stably controlled at Ra≤0.3μm.

[0031] 2. Significantly enhanced adaptability and process generalization. The deep reinforcement learning (DRL) framework autonomously optimizes grinding parameters through 100 policy iterations per cycle. For example, it generates the optimal combination of "0.2MPa + 2000r / min" for ceramic materials, increasing the pass rate of workpieces of different materials from 60%-70% to 99%. Digital twin virtual verification reduces trial and error costs by 70%, shortens the response time for abnormal operating conditions from ≥100ms to ≤20ms, and reduces the occurrence rate of over-polishing by 80%.

[0032] 3. Improved efficiency and flexible production capacity through multi-process collaboration The "grinding-polishing" process switching time is ≤5 seconds, and the conversion error is ≤0.0005mm. The precision connection of the entire process is achieved through feed axis compensation. The Laser-Induced Breakdown Spectrometer (LIBS) achieves 98% accuracy in intelligent identification of 12 types of materials, reducing changeover time from ≥30 minutes to 1 minute and improving the adaptability of flexible production lines by 30 times.

[0033] 4. Global coordination and economic optimization The edge-cloud collaborative architecture optimizes global parameters through federated learning, reducing batch processing consistency deviation from ≥0.003mm to ≤0.0008mm, and improving production stability by 375%. By adopting domestically produced sensors such as Huake Precision Laser Scanner and lightweight model design, the system cost has been reduced from ≥200,000 yuan to 50,000-80,000 yuan, and the adoption rate among small and medium-sized enterprises has increased by 40%. Energy consumption is reduced by 15%-25%, grinding wheel life is extended by 20%-23%, and the unit workpiece processing cost is reduced by 30%.

[0034] 5. Environmental protection and production safety assurance Dust concentration is controlled at ≤1.2mg / m³, which is better than the 2mg / m³ standard, and noise is ≤80dB, which is lower than the 85dB limit, meeting the requirements of green manufacturing; The response time to abnormal clamping force is ≤10ms, the accuracy rate of early warning of grinding wheel cracks is ≥95%, the safety accident rate is 0, and the production safety is significantly improved.

[0035] In summary, this invention, through multi-physics coupling analysis, intelligent decision-making, and global collaboration, comprehensively breaks through the bottlenecks in accuracy, efficiency, and economy of traditional systems, providing core technical support for the mass production of high-precision piston pins. Attached Figure Description

[0036] Figure 1 This is a block diagram of the overall architecture of the piston pin grinding monitoring system of the present invention; Figure 2 This is a detailed flowchart of the data acquisition module in this invention; Figure 3 This is a flowchart of the data analysis and intelligent decision-making process in this invention; Figure 4 This is a flowchart of the control execution module adjustment process in this invention; Figure 5 This is a flowchart of the edge-cloud collaboration and human-computer interaction process in this invention. Detailed Implementation

[0037] The technical solution of the present invention will now be described with reference to the accompanying drawings and embodiments.

[0038] Please see Figure 1-5 This embodiment provides the following technical solution: A piston pin grinding monitoring system includes a data acquisition module, a data analysis and intelligent decision-making module, a control execution module, a human-machine interaction module, and an edge-cloud collaborative platform. The data acquisition module is used to synchronously collect basic parameters, multi-physics field parameters, material characteristic parameters, and environmental safety parameters during the grinding process. The data analysis and intelligent decision-making module integrates and processes the collected data to generate optimized control strategies. The control execution module drives the actuators according to the strategies. The human-machine interaction module provides a visual operation and intervention interface. The edge-cloud collaborative platform realizes the collaboration between local real-time control and global process optimization.

[0039] This piston pin grinding monitoring system, through a multi-module collaborative architecture, achieves intelligent control of the entire process from data perception to global optimization, with significant technical effects: the data acquisition module covers dimensions such as basic parameters, multiple physical fields, material characteristics, and environmental safety, breaking through the limitations of traditional single-parameter monitoring and providing comprehensive data support for accurate decision-making; the data analysis and intelligent decision-making module integrates and processes multi-source data, generating optimization strategies that can adapt to the real-time working conditions of a single workpiece while also taking into account the consistency requirements of batch production; the control execution module links with the edge-cloud collaborative platform, ensuring the surface roughness Ra≤0.4μm and dimensional tolerance ±0.0005mm of a single workpiece through local real-time control response speed ≤20ms, while improving the stability and consistency deviation of batch production ≤0.0008mm through cloud-based global optimization; the human-machine interaction module realizes parameter visualization and manual intervention, balancing intelligence and operational flexibility.

[0040] Overall, the system upgrades piston pin grinding from "passive monitoring" to "active optimization," increasing the pass rate of special material workpieces from 60%-70% to 99%, reducing energy consumption by 15%-25%, and meeting the requirements of green manufacturing and safe production, thus comprehensively improving the quality, efficiency, and economy of precision piston pin processing.

[0041] Specifically, the data acquisition module includes a basic parameter sensing unit and a multi-physics sensing unit. The basic parameter sensing unit includes a displacement sensor with a resolution of 0.01μm, a temperature sensor with an accuracy of ±0.5℃, and a vibration sensor with a frequency response of 0.5-10kHz. The multi-physics sensing unit includes an infrared thermal imager with a resolution of 640×512, a micro-strain sensor with an accuracy of ±1με, and a laser scanner with an accuracy of ±0.002mm, used to acquire temperature field distribution, stress and strain, and surface three-dimensional point cloud data.

[0042] The multi-physics sensing unit simultaneously acquires temperature field distribution, stress and strain, and surface three-dimensional morphology through infrared thermal imaging with a resolution of 640×512 and micro-strain accuracy of ±1με and laser scanning accuracy of ±0.002mm, filling the gap in the perception of thermo-mechanical-shape coupling effects in traditional systems. The combination of the two provides multi-dimensional and high-precision data support for subsequent intelligent decision-making, improving the comprehensiveness and accuracy of grinding process monitoring from the source, and laying the foundation for precision control and anomaly early warning.

[0043] Specifically, the data acquisition module also includes a material and process sensing unit, which includes a laser-induced breakdown spectrometer (LIBS) with a wavelength range of 200-1000nm and a roughness sensor with a measurement range of 0.02-10μmRa. The LIBS is used to analyze the elemental composition of the workpiece surface with an accuracy of 98%, and the roughness sensor is used to collect the surface roughness data of the machined surface in real time.

[0044] Laser-induced breakdown spectroscopy (LIBS) with a wavelength range of 200-1000nm and a recognition accuracy of 98% can quickly pinpoint the elemental composition of workpiece surfaces, providing direct evidence for matching specific grinding strategies to different materials such as ceramics and aluminum alloys. The surface roughness sensor with a range of 0.02-10μmRa monitors the quality of the processed surface in real time, ensuring that the grinding accuracy is always within a controllable range. The combination of the two fills the gap in the perception of material differences and real-time process effects in traditional systems, providing key data support for adaptive adjustment of grinding parameters and ensuring consistency in batch processing, effectively improving the system's adaptability to workpieces of multiple materials and its ability to control process quality.

[0045] Specifically, the data acquisition module also includes an environmental safety sensing unit, which includes a dust concentration sensor with an accuracy of 0.1 mg / m³, a noise sensor with a range of 30-130 dB, and a piezoelectric sensor with an accuracy of ±2%. The dust and noise sensors are linked to the dust removal equipment, and the piezoelectric sensor is used to detect the workpiece clamping force.

[0046] The dust concentration sensor with an accuracy of 0.1 mg / m³ and the noise sensor with a range of 30-130 dB monitor environmental parameters in real time, linking with dust removal equipment for dynamic control to ensure that dust and noise levels meet standards and comply with green manufacturing requirements. The piezoelectric sensor with ±2% accuracy accurately detects clamping force, promptly identifying the risk of workpiece loosening and triggering protective measures. The combination of these three sensors fills the gap in environmental protection and safety monitoring in traditional systems, ensuring compliance with operating environment regulations and preventing safety accidents caused by clamping abnormalities, thus achieving synergy between high-precision machining and safety and environmental protection.

[0047] Specifically, the data analysis and intelligent decision-making module includes a multiphysics coupling analysis submodule. This submodule constructs a coupled model of temperature field, stress field, and surface morphology based on the formula σ=E×(ε+α×ΔT), where σ is stress, E is elastic modulus, ε is strain, α is thermal expansion coefficient, and ΔT is temperature difference. When the stress exceeds 10% of the material's yield strength or the temperature gradient is >5℃ / mm, the surface curvature calculation deviation is automatically corrected, with a correction amount ≤0.0005mm-1.

[0048] The accurate calculation of the thermo-mechanical-shape coupling effect is achieved based on the formula σ=E×(ε+α×ΔT). When the stress or temperature gradient exceeds the limit, such as a temperature gradient > 5℃ / mm, the surface curvature deviation is automatically corrected to ≤0.0005mm. - ¹ This eliminates the interference of thermal deformation on measurement accuracy; it fundamentally solves the problem of misjudgment of surface accuracy caused by local high temperature or stress overload, making the curvature calculation more consistent with the actual shape of the workpiece, providing accurate morphological basis for subsequent grinding parameter optimization, and further improving the stability of processing accuracy.

[0049] Specifically, the data analysis and intelligent decision-making module includes a reinforcement learning adaptive decision-making sub-module. This sub-module uses the deep deterministic policy gradient (DDPG) algorithm to construct an Actor-Critic network. The comprehensive reward function is based on a surface accuracy achievement rate of ≥99%, an energy consumption reduction rate of ≥15%, and a grinding wheel life extension rate of ≥20%. The strategy is iterated once every 100 workpieces are processed. The new strategy can only be applied after it has been verified in a digital twin virtual environment with a pass rate of >95%.

[0050] The DDPG algorithm focuses on surface accuracy, energy consumption, and grinding wheel life. It iterates the strategy every 100 workpieces, enabling the system to autonomously learn the optimal parameter combination, such as the force and speed for material matching. The digital twin virtual verification pass rate of >95% ensures the reliability of the new strategy, avoiding the cost of physical trial and error. Through continuous iteration, it improves the surface accuracy compliance rate to ≥99%, reduces energy consumption by ≥15%, and extends grinding wheel life by ≥20%, significantly enhancing the system's adaptability to complex working conditions and its long-term operational economy.

[0051] Specifically, the data analysis and intelligent decision-making module includes a material identification and process matching sub-module. This sub-module identifies 12 types of piston pin materials, including ceramics and composite materials, through LIBS data classification, and calls a preset process library to achieve parameter self-configuration within 1 minute. The process library includes parameter templates such as grinding wheel type, grinding force 0.1-1MPa, and rotation speed 1000-5000r / min.

[0052] Based on LIBS data, it can accurately identify 12 types of materials, including ceramics and composite materials. Combined with the preset process library covering key parameters such as grinding wheel type, force, and speed, the parameter self-configuration can be completed within 1 minute. It completely solves the problems of traditional manual changeover taking ≥30 minutes and low parameter adaptation accuracy, making the processing and switching of piston pins of different materials more efficient and the matching of process parameters more accurate, significantly improving the adaptability and consistency of multi-variety batch production.

[0053] Specifically, the control execution module includes a basic parameter adjustment unit and an adaptive execution unit; the basic parameter adjustment unit links the automatic grinding wheel changing mechanism to change the tool in ≤15 seconds when the grinding wheel wear rate is >0.01mm / min, and adjusts the cooling flow rate to increase by 30% and reduce the feed rate by 20% when the temperature is >100℃; the adaptive execution unit uses a piezoelectric controller to achieve feed adjustment with a resolution of 0.01μm and a response time of ≤20ms.

[0054] The basic parameter adjustment unit automatically triggers wheel replacement within 15 seconds and adjusts cooling and feed for typical working conditions such as grinding wheel wear and temperature exceeding limits, ensuring machining stability. The adaptive execution unit achieves a 0.01μm-level micro-adjustment response of ≤20ms with the help of a piezoelectric controller, accurately adapting to changes in the micro-morphology of the piston pin surface. The combination of the two solves the standardized handling of abnormal conventional parameters and meets the micro-adjustment requirements of high-precision machining, thus ensuring both grinding accuracy and efficiency.

[0055] Specifically, the control execution module also includes a multi-process linkage unit, which links the grinding wheel and ultrasonic polishing machine at a frequency of 20-40kHz. When the real-time roughness Ra>0.4μm, it automatically switches to the polishing process with a process switching error ≤0.0005mm, and the deviation is compensated and corrected by the feed axis.

[0056] Based on real-time roughness Ra>0.4μm, automatic process switching is triggered, seamlessly linking grinding wheel grinding and ultrasonic polishing at 20-40kHz without manual intervention; process switching error ≤0.0005mm, and deviation is dynamically compensated and corrected through the feed axis to ensure the continuity of machining accuracy across processes; it breaks through the precision loss bottleneck of traditional segmented machining, enabling the surface roughness of piston pins to be stably controlled at Ra≤0.2μm, significantly improving the integrated machining capability and quality stability of complex surfaces.

[0057] Specifically, the edge-cloud collaborative platform achieves clock synchronization error ≤0.1ms through TSN, the industrial PC at the edge node is responsible for real-time control response time ≤20ms, and the cloud platform gathers data from 100+ devices through federated learning FedAvg algorithm to optimize global parameters. When the diameter deviation of batch workpieces is >0.002mm, a unified feed correction scheme of ±0.0005mm is pushed.

[0058] TSN clock synchronization error ≤0.1ms ensures spatiotemporal alignment of multi-source data, and edge nodes complete real-time control within 20ms, ensuring the machining accuracy of single workpieces; The cloud aggregates data from 100+ devices through federated learning, and pushes a unified correction scheme of ±0.0005mm for batch diameter deviations >0.002mm, improving batch production consistency. The combination of the two not only meets the real-time requirements of high-precision processing, but also solves the consistency problem of batch production of multiple devices, improving the overall processing stability by more than 30%.

[0059] Monitoring methods and procedures Initialization phase Sensor calibration: The laser scanner is calibrated with a standard gauge block accuracy of ±0.0001mm, and the temperature measurement error is corrected to ±0.5℃; System configuration: Import workpiece parameters: diameter 3-50mm, length 20-100mm, material type; initialize DRL model parameters. Network synchronization: The clock is calibrated between the edge and the cloud via 5G Time-Sensitive Networking (TSN), with a synchronization error of ≤0.1ms.

[0060] Data acquisition and preprocessing stage Multi-source data synchronous acquisition: Basic parameters such as displacement and temperature are synchronized with infrared thermal imaging and LIBS spectral data via TSN, with a sampling frequency of 1kHz; Data cleaning: Point cloud data is denoised by removing points outside the mean 3σ using statistical filtering, and vibration signals are cleaned by removing 50Hz power frequency interference.

[0061] Intelligent analysis and decision-making stage Multiphysics coupling calculation: outputs stress field and temperature field distribution, corrects surface curvature values ​​with an accuracy of ±0.0005mm. - ¹; DRL strategy generation: Outputs optimal grinding parameters based on current temperature, stress, and roughness. Material identification and process matching: After classifying LIBS data, the corresponding process template is called; Environmental safety assessment: Real-time monitoring of dust, noise, and clamping force, and output of early warning or control signals.

[0062] Control Execution Phase Priority control: Safety signals such as insufficient clamping force are executed first, with a response time ≤10ms, followed by process parameter adjustments; Multi-process linkage: Switching between grinding / polishing processes based on roughness data, and synchronously compensating for conversion errors; Edge-cloud collaboration: Edge nodes perform real-time control, while the cloud records data and optimizes the global model.

[0063] Feedback and Iteration Phase Effect evaluation: After polishing, laser scanning detection showed that the surface roughness Ra ≤ 0.8 μm and the dimensional tolerance ± 0.001 mm; Model update: Feed the actual processing data error value and energy consumption back to the DRL model, and iterate the strategy once every 100 workpieces; Log storage: Process parameters and test results are saved in the cloud, supporting quality traceability and a retention period of 3 years.

[0064] This process achieves high precision, high stability, and continuous optimization in piston pin grinding through a closed-loop control system across all stages and an intelligent iteration mechanism. During the initialization phase, precise calibration and synchronization errors ≤0.1ms lay the data foundation for the entire process. During the data acquisition and analysis phase, curvature calculations using multiphysics coupling achieve an accuracy of ±0.0005 mm. - ¹ DRL strategy generation and material matching ensure scientific decision-making; During the control execution phase, safety is prioritized and a response time of ≤10ms is achieved. Multiple processes are linked and errors are compensated to ensure machining accuracy. During the feedback and iteration phase, the strategy is continuously optimized through effect evaluation and model updates every 100 pieces. Combined with cloud-based traceability of 3 years of data, stable control of surface roughness Ra≤0.8μm and dimensional tolerance±0.001mm is finally achieved, while improving the traceability and long-term adaptability of the process.

[0065] Specific Implementation Cases I. Basic Information of the Implementation Targets This implementation case focuses on the grinding and machining of key parameters of ceramic-coated piston pins for a certain type of automotive engine, verifying the system's high precision, multi-material compatibility, and environmental and safety performance. Workpiece material: 45 steel substrate + Al2O3 ceramic coating, thickness 0.1-0.2mm Dimensions: Diameter 10mm, Length 50mm, Corner radius R1.5mm at both ends Accuracy requirements: Surface roughness Ra≤0.4μm, dimensional tolerance±0.001mm, roundness error≤0.0005mm Process requirements: The outer surface must be ground and the end corners rounded to prevent the ceramic coating from peeling off. II. System Configuration and Initialization Parameters 1. Hardware Configuration Details

[0066] 2. Initialize parameter configuration

[0067] III. Data Implementation of the Entire Polishing Process 1. Duration of data acquisition and preprocessing phase: 120s

[0068] 2. Duration of intelligent analysis and decision-making phase: 10 seconds / batch

[0069] 3. Duration of the control execution phase: 180s

[0070] 4. Duration of feedback and iteration phase: 30 seconds

[0071] IV. Summary Table of Implementation Results

[0072] V. Implementation Conclusions This system successfully achieves high-precision grinding of ceramic-coated piston pins through technologies such as multi-physics coupling analysis to correct thermal deformation deviations, reinforcement learning to adaptively optimize parameters, and multi-process linkage control. All indicators are better than the design requirements, especially in terms of surface accuracy Ra=0.3μm, energy consumption reduction of 20.8%, and one-minute changeover for multiple materials. This verifies the practicality and advanced nature of the system in the mass production of precision piston pins.

[0073] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A piston pin grinding monitoring system, characterized in that, It includes a data acquisition module, a data analysis and intelligent decision-making module, a control execution module, a human-machine interaction module, and an edge-cloud collaborative platform. The data acquisition module is used to synchronously collect basic parameters, multi-physics field parameters, material characteristic parameters, and environmental safety parameters during the polishing process. The data analysis and intelligent decision-making module integrates and processes the collected data to generate optimized control strategies. The control execution module drives the actuators according to the strategies. The human-machine interaction module provides a visual operation and intervention interface. The edge-cloud collaborative platform realizes the collaboration between local real-time control and global process optimization.

2. The piston pin grinding monitoring system according to claim 1, characterized in that: The data acquisition module includes a basic parameter sensing unit and a multi-physics sensing unit. The basic parameter sensing unit includes a displacement sensor with a resolution of 0.01 μm, a temperature sensor with an accuracy of ±0.5℃, and a vibration sensor with a frequency response of 0.5-10kHz. The multi-physics sensing unit includes an infrared thermal imager with a resolution of 640×512, a micro-strain sensor with an accuracy of ±1με, and a laser scanner with an accuracy of ±0.002mm, used to acquire temperature field distribution, stress and strain, and surface three-dimensional point cloud data.

3. The piston pin grinding monitoring system according to claim 1, characterized in that: The data acquisition module also includes a material and process sensing unit, which includes a laser-induced breakdown spectrometer (LIBS) with a wavelength range of 200-1000 nm and a roughness sensor with a measurement range of 0.02-10 μm Ra. The LIBS is used to analyze the elemental composition of the workpiece surface with an accuracy of 98%, and the roughness sensor is used to collect the surface roughness data of the processed surface in real time.

4. The piston pin grinding monitoring system according to claim 1, characterized in that: The data acquisition module also includes an environmental safety sensing unit, which includes a dust concentration sensor with an accuracy of 0.1 mg / m³, a noise sensor with a range of 30-130 dB, and a piezoelectric sensor with an accuracy of ±2%. The dust and noise sensors are linked to a dust removal device, and the piezoelectric sensor is used to detect the workpiece clamping force.

5. The piston pin grinding monitoring system according to claim 1, characterized in that: The data analysis and intelligent decision-making module includes a multiphysics coupling analysis submodule. This submodule constructs a coupled model of temperature field, stress field, and surface morphology based on the formula σ=E×(ε+α×ΔT), where σ is stress, E is the elastic modulus, ε is strain, α is the coefficient of thermal expansion, and ΔT is the temperature difference. When the stress exceeds 10% of the material's yield strength or the temperature gradient is >5℃ / mm, the surface curvature calculation deviation is automatically corrected, with a correction amount ≤0.0005mm. -1 .

6. The piston pin grinding monitoring system according to claim 1, characterized in that, The data analysis and intelligent decision-making module includes a reinforcement learning adaptive decision-making submodule. This submodule uses the deep deterministic policy gradient (DDPG) algorithm to construct an Actor-Critic network. The comprehensive reward function is based on a surface accuracy achievement rate of ≥99%, an energy consumption reduction rate of ≥15%, and a grinding wheel life extension rate of ≥20%. The strategy is iterated once every 100 workpieces are processed. The new strategy can only be applied after the pass rate of verification in the digital twin virtual environment is >95%.

7. The piston pin grinding monitoring system according to claim 1, characterized in that, The data analysis and intelligent decision-making module includes a material identification and process matching submodule. This submodule identifies 12 types of piston pin materials, including ceramics and composite materials, through LIBS data classification, and calls a preset process library to achieve parameter self-configuration within 1 minute. The process library includes parameter templates such as grinding wheel type, grinding force 0.1-1MPa, and rotation speed 1000-5000r / min.

8. A piston pin grinding monitoring system according to claim 1, characterized in that: The control execution module includes a basic parameter adjustment unit and an adaptive execution unit. The basic parameter adjustment unit activates the automatic grinding wheel changing mechanism with a tool change time of ≤15 seconds when the grinding wheel wear rate is >0.01mm / min, and adjusts the cooling flow rate by 30% and reduces the feed rate by 20% when the temperature is >100℃. The adaptive execution unit uses a piezoelectric controller to achieve feed adjustment with a resolution of 0.01μm and a response time of ≤20ms.

9. A piston pin grinding monitoring system according to claim 1, characterized in that: The control execution module also includes a multi-process linkage unit, which links the grinding wheel and ultrasonic polishing machine at a frequency of 20-40kHz. When the real-time roughness Ra>0.4μm, it automatically switches to the polishing process with a process switching error ≤0.0005mm, and the deviation is compensated and corrected by the feed axis.

10. A piston pin grinding monitoring system according to claim 1, characterized in that: The edge-cloud collaborative platform achieves clock synchronization error ≤0.1ms through TSN, the edge node industrial PC is responsible for real-time control response time ≤20ms, and the cloud platform gathers data from 100+ devices to optimize global parameters through federated learning FedAvg algorithm. When the diameter deviation of batch workpieces is >0.002mm, a unified feed correction scheme of ±0.0005mm is pushed.

Images