Numerical control machining intelligent monitoring method and system based on industrial internet of things

Abstract

The invention discloses a numerical control machining intelligent monitoring method and system based on industrial Internet of Things, and relates to the technical field of industrial Internet of Things and intelligent manufacturing. The method mainly comprises the steps of multi-source data real-time acquisition, edge side data preprocessing, cloud intelligent analysis and diagnosis, digital twin closed-loop optimization and visualization and feedback control. According to the invention, visualization and feedback control are adopted. By introducing the digital twin-driven real-time process optimization and multi-modal data fusion intelligent early warning mechanism, the machining precision and efficiency are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial Internet of Things and intelligent manufacturing, and particularly relates to an intelligent monitoring method and system for numerical control machining based on the industrial Internet of Things. Background Art

[0002] At present, traditional numerical control machining relies on manual inspection and off-line detection, and it is impossible to detect equipment anomalies (such as tool wear, spindle vibration, etc.) in real time, resulting in an increase in the scrap rate or equipment damage; adopting a fixed monitoring frequency, it is impossible to dynamically adjust according to the machining conditions, resulting in missing or redundant data in key processes and there is also a lag in early warning. The anomaly judgment relies on a single threshold and lacks multi-dimensional joint analysis (such as collaborative evaluation of vibration, temperature, and tool wear); the simulated machining data is disconnected from the actual data and cannot be fed back to the digital twin model in real time for dynamic correction.

[0003] In summary, the present invention designs an intelligent monitoring method and system for numerical control machining based on the industrial Internet of Things. Summary of the Invention

[0004] Aiming at the deficiencies in the prior art, the purpose of the present invention is to provide an intelligent monitoring method and system for numerical control machining based on the industrial Internet of Things. By introducing a real-time process optimization driven by digital twin and an intelligent early warning mechanism for multi-modal data fusion, the machining accuracy and efficiency are significantly improved.

[0005] In order to achieve the above purpose, the present invention is realized by the following technical solutions: An intelligent monitoring method for numerical control machining based on the industrial Internet of Things includes the following steps:

[0006] Step 1: Real-time acquisition of multi-source data

[0007] Through high-precision sensors (vibration, temperature, current, acoustic emission) and the built-in interface of the numerical control system (CNC), multi-dimensional data during the machining process (such as spindle load, feed rate, tool temperature) is collected; and an adaptive sampling strategy is adopted: when the prediction error approaches the tolerance threshold, high-frequency monitoring is triggered (such as the laser displacement sensor is increased from 100Hz to 1kHz); for high-risk sub-routes (such as surface machining), acoustic emission signals are introduced for auxiliary monitoring, and micro-cracks are identified in advance through frequency domain features (such as wavelet packet energy entropy); based on the material properties of the tool path, tool wear curve and historical data, a Bayesian network model for the machining error of the sub-route is constructed to dynamically predict the maximum error value;

[0008] Step 2: Edge-side data preprocessing

[0009] Noise filtering (such as wavelet transform), feature extraction (time-frequency domain statistics) and data compression algorithms are deployed at the edge gateway to reduce the cloud transmission load.

[0010] Step 3: Cloud Intelligent Analysis and Diagnosis

[0011] Anomaly detection: Based on the LSTM-AE (Long Short-Term Memory Autoencoder) model, model the time series data, and trigger an alarm when the reconstruction error exceeds the threshold.

[0012] Tool health prediction: Integrate the physical model (wear rate equation) and the data-driven model (random forest regression) to predict the remaining useful life (RUL).

[0013] Process optimization: Dynamically adjust the cutting parameters (such as rotational speed, feed rate) through reinforcement learning (PPO algorithm), with the machining efficiency and quality as the reward function.

[0014] Step 4: Digital Twin Closed-Loop Optimization

[0015] Real-time twin calibration: Real-time feedback the actual machining data (such as spindle load, temperature, vibration spectrum) to the digital twin model, and dynamically optimize the feed speed and cutting depth through reinforcement learning (PPO algorithm). The optimization objectives consider both machining efficiency (material removal rate) and quality (surface roughness).

[0016] Virtual machining verification: Preview the effect of parameter adjustment in the digital twin environment, and verify the feasibility through multi-physics field simulation (cutting force, thermal deformation). Only when the virtual machining results meet the tolerance requirements, the optimized parameters are sent to the physical machine tool for execution to avoid ineffective parameter adjustment.

[0017] Step 5: Visualization and Feedback Control

[0018] Real-time display the device status, alarm information and optimization suggestions through the Web interface, and support the issuance of remote control instructions.

[0019] The alarm levels in Step 3 are divided into three levels:

[0020] Level 1 alarm (emergency stop): The tool position deviation exceeds the threshold (such as ±0.1 mm) or the vibration acceleration suddenly increases (>10 g);

[0021] Level 2 alarm (process adjustment): The temperature gradient is abnormal (ΔT>50℃ / s) or the current fluctuation continuously exceeds the limit (3σ criterion);

[0022] Level 3 alarm (predictive maintenance): Based on the tool wear model (such as Archard equation), predict that the remaining life is less than 10%.

[0023] When the position deviation and the vibration signal are both abnormal, the confidence level is increased to 95% to avoid false alarms.

[0024] An intelligent monitoring system for numerical control machining based on industrial Internet of Things, including a perception layer 1, an edge computing layer 2, a cloud digital twin platform 3 and an application layer 4. The perception layer 1 adopts a multi-source sensor array, including a high-precision grating ruler, an infrared thermal imager linked with a fiber Bragg grating temperature sensor, and an acoustic emission sensor for early damage detection.

[0025] The edge computing layer 2 is a TinyML model deployed on an edge gateway, which realizes real-time FFT analysis of vibration signals and transfer learning classification of tool wear states.

[0026] The cloud digital twin platform 3 receives the actual machining data uploaded by the edge layer, drives the synchronous update of the digital twin model, integrates a reinforcement learning agent (PPO algorithm), generates parameter optimization strategies based on simulation results, and the virtual trial cutting module supports G-code level instruction verification to ensure that the issued parameters are compatible with the machine tool control system.

[0027] The application layer 4 displays a three-dimensional superposition view of tool wear and an optimized path through Microsoft HoloLens.

[0028] Advantages of the present invention:

[0029] 1. Improvement in dynamic monitoring accuracy: Through the error prediction model, the data coverage rate of key processes is increased by 40%;

[0030] 2. Early warning accuracy: Multimodal fusion reduces the false alarm rate to <2% (traditional method >15%);

[0031] 3. Process optimization efficiency: Digital twin closed-loop optimization reduces the number of trial cutting times by 60%. Brief description of the drawings

[0032] The present invention will be described in detail below with reference to the drawings and specific embodiments;

[0033] Figure 1 is the process flow chart of the present invention;

[0034] Figure 2 is the system architecture diagram of the present invention. Specific embodiments

[0035] To make the technical means, creative features, achieved purposes and effects of the present invention easy to understand, the present invention will be further described below in conjunction with specific embodiments.

[0036] Refer to Figure 1-2 , this specific embodiment adopts the following technical solution: An intelligent monitoring method for numerical control machining based on industrial Internet of Things, including the following steps:

[0037] Step 1: Real-time acquisition of multi-source data

[0038] Collect multi-dimensional data (such as spindle load, feed rate, tool temperature) during the machining process through the built-in interfaces of high-precision sensors (vibration, temperature, current, acoustic emission) and the numerical control system (CNC); and adopt an adaptive sampling strategy: when the prediction error approaches the tolerance threshold, trigger high-frequency monitoring (such as the laser displacement sensor is increased from 100Hz to 1kHz); introduce acoustic emission signal-assisted monitoring for high-risk sub-routes (such as surface machining), and identify micro-cracks in advance through frequency domain features (such as wavelet packet energy entropy); based on the material properties of the tool path, tool wear curve and historical data, construct a Bayesian network model for the machining error of the sub-route, and dynamically predict the maximum error value;

[0039] Step 2: Edge-side data preprocessing

[0040] Deploy noise filtering (such as wavelet transform), feature extraction (time-frequency domain statistics) and data compression algorithms at the edge gateway to reduce the cloud transmission load.

[0041] Step 3: Cloud intelligent analysis and diagnosis

[0042] Anomaly detection: Based on the LSTM-AE (Long Short-Term Memory Autoencoder) model, model the time-series data, and trigger an alarm when the reconstruction error exceeds the threshold.

[0043] Tool health prediction: Integrate the physical model (wear rate equation) with the data-driven model (random forest regression) to predict the remaining useful life (RUL).

[0044] Process optimization: Dynamically adjust the cutting parameters (such as speed, feed rate) through reinforcement learning (PPO algorithm), with machining efficiency and quality as the reward function.

[0045] Step 4: Digital twin closed-loop optimization

[0046] Real-time twin calibration: Real-time feedback the actual machining data (such as spindle load, temperature, vibration spectrum) to the digital twin model, and dynamically optimize the feed speed and cutting depth through reinforcement learning (PPO algorithm). The optimization objectives consider both machining efficiency (material removal rate) and quality (surface roughness).

[0047] Virtual trial cut verification: Preview the effect of parameter adjustment in the digital twin environment, and verify the feasibility through multi-physics field simulation (cutting force, thermal deformation). Only when the virtual trial cut results meet the tolerance requirements, the optimized parameters are sent to the physical machine tool for execution, avoiding ineffective parameter adjustment.

[0048] Step 5: Visualization and feedback control

[0049] Real-time display of the device status, alarm information and optimization suggestions through the Web interface, and support the issuance of remote control instructions.

[0050] In this specific implementation, through multimodal data fusion and anomaly tracing mechanism, cross-modal feature correlation analysis and confidence dynamic weighting algorithm are carried out; the cross-modal feature correlation analysis includes: a constructing a cross-modal feature map of vibration-temperature-current, and establishing the correlation weights between parameters through a graph neural network (GNN). b When a certain index is abnormal (such as a sudden increase in vibration), automatically retrieve the abnormal features of the associated modality (such as the current harmonic distortion rate in the same period) to achieve abnormal root cause location.

[0051] The confidence dynamic weighting algorithm: defines the real-time credibility coefficient of each modality data (based on signal-to-noise ratio, sensor health status); when multi-modal joint alarm is triggered, a hierarchical response strategy is triggered:

[0052] High confidence (>90%): directly execute shutdown / parameter adjustment;

[0053] Medium confidence (70%-90%): start digital twin verification;

[0054] Low confidence (<70%): only record logs and prompt manual verification.

[0055] In this specific implementation, a high-fidelity digital twin is constructed in the cloud, integrating the following modules:

[0056] Physical engine (ANSYS): real-time simulation of cutting force-thermal deformation coupling effect;

[0057] Virtual sensor: complement data in the blind area of actual sensors (such as the temperature field of the tool flank).

[0058] When actual machining triggers a secondary alarm, the twin system quickly tests various parameter combinations (such as feed rate ±10%, spindle speed ±5%) in the virtual environment and selects the optimal solution to send to the machine tool.

[0059] In this specific implementation, through the integration of multimodal perception technology and digital twin closed-loop optimization, intelligent monitoring and dynamic regulation of the entire NC machining process are realized. Its core advantages are reflected in:

[0060] 1) Adopting an adaptive sampling strategy and a Bayesian network error prediction model, the monitoring coverage rate of key processes is increased by 40%;

[0061] 2) Innovatively constructing a vibration-temperature-current cross-modal feature map, realizing abnormal root cause location through a graph neural network, and controlling the false alarm rate below 2%;

[0062] 3) Based on the digital twin virtual try-cut verification mechanism, combined with reinforcement learning to dynamically optimize process parameters, reducing the number of try-cuts by 60%, and significantly improving machining accuracy and efficiency;

[0063] 4) The original three - level confidence - graded response strategy realizes an intelligent decision - making closed - loop from emergency shutdown to predictive maintenance.

[0064] This system is particularly applicable to precision machining fields such as aerospace. On the premise of ensuring quality, the scrap rate can be reduced by more than 35%.

[0065] Example 1: Taking the machining of aero - engine blades as an example:

[0066] 1. Abnormality detection: During the finish - machining stage, when a sudden increase in vibration (8g) is detected and the third - harmonic distortion rate of the current exceeds the limit 3 times, the system determines that the tool has chipped (confidence level 92%), triggering a first - level alarm.

[0067] 2. Twin verification: The cloud digital twin simulation shows that reducing the feed rate by 15% can eliminate the vibration abnormality. After actual execution, the surface waviness of the machining is reduced by 60%.

[0068] 3. Dynamic optimization: The optimal parameter combination is obtained through reinforcement learning: the finish - machining speed is increased by 8% while the feed rate is reduced by 5%. On the premise of ensuring Ra ≤ 0.4μm, the efficiency is increased by 12%.

[0069] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above - mentioned embodiments. What is described in the above - mentioned embodiments and the specification only illustrates the principle of the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed. The scope of the present invention claimed is defined by the appended claims and their equivalents.

Claims

1. An intelligent monitoring method for numerical control machining based on the industrial Internet of Things, characterized in that It includes the following steps: Step (1): Real-time acquisition of multi-source data Collect multi-dimensional data during the machining process through high-precision sensors and built-in interfaces of the numerical control system; And adopt an adaptive sampling strategy: when the prediction error approaches the tolerance threshold, trigger high-frequency monitoring; introduce acoustic emission signals to assist in monitoring high-risk sub-routes, and identify micro-cracks in advance through frequency-domain characteristics; Based on the material properties of the tool path, tool wear curve, and historical data, construct a Bayesian network model for the machining error of the sub-route, and dynamically predict the maximum error value; Step (2): Edge-side data preprocessing Deploy noise filtering, feature extraction, and data compression algorithms at the edge gateway to reduce the transmission load to the cloud; Step (3): Cloud intelligent analysis and diagnosis Anomaly detection: Based on the LSTM-AE model, model the time-series data, and trigger an alarm when the reconstruction error exceeds the threshold; Tool health prediction: Integrate physical models and data-driven models to predict the remaining service life; Process optimization: Dynamically adjust cutting parameters through the PPO algorithm, with machining efficiency and quality as the reward function; Step (4): Digital twin closed-loop optimization Real-time twin calibration: Feed the actual machining data back to the digital twin model in real time, and dynamically optimize the feed speed and cutting depth through reinforcement learning. The optimization goal takes both machining efficiency and quality into account; Virtual trial cut verification: Preview the effect of parameter adjustment in the digital twin environment, verify the feasibility through multi-physics field simulation. Only when the virtual trial cut results meet the tolerance requirements, the optimized parameters are sent to the physical machine tool for execution to avoid ineffective parameter adjustment; Step (5): Visualization and feedback control Real-time display of device status, alarm information, and optimization suggestions through a Web interface, and support for issuing remote control instructions.

2. The intelligent monitoring method for numerical control machining based on industrial Internet of Things according to claim 1, characterized in that, The alarm levels in step (3) are divided into three levels: Level 1 alarm: The tool position deviation exceeds the threshold or the vibration acceleration suddenly increases; Level 2 alarm: The temperature gradient is abnormally ΔT>50°C / s or the current fluctuation continuously exceeds the limit; Level 3 alarm: Based on the tool wear model, the predicted remaining life is less than 10%; When the position deviation and vibration signal are both abnormal, the confidence level is increased to 95% to avoid false alarms.

3. An intelligent monitoring system for numerical control machining based on the industrial Internet of Things, characterized in that, It includes a perception layer (1), an edge computing layer (2), a cloud digital twin platform (3), and an application layer (4). The perception layer (1) adopts a multi-source sensor array, including a high-precision grating scale, an infrared thermal imager linked with a fiber Bragg grating temperature sensor, and an acoustic emission sensor for early damage detection.

4. An intelligent monitoring system for numerical control machining based on industrial Internet of Things according to claim 3, characterized in that, The edge computing layer (2) is a TinyML model deployed on the edge gateway, which realizes real-time FFT analysis of vibration signals; transfer learning classification of tool wear states.

5. The intelligent monitoring system for numerical control machining based on industrial Internet of Things according to claim 3, characterized in that, The cloud digital twin platform (3) receives the actual machining data uploaded by the edge layer, drives the digital twin model to update synchronously; integrates a reinforcement learning agent, and generates a parameter optimization strategy based on the simulation results; the virtual trial cut module supports G-code level instruction verification to ensure that the issued parameters are compatible with the machine tool control system.

6. The intelligent monitoring system for numerical control machining based on industrial Internet of Things according to claim 3, wherein, The application layer (4) displays a three-dimensional superposition view of tool wear and an optimization path through Microsoft HoloLens.

Images