Real-time monitoring device for tool wear of numerically controlled lathe

Abstract

The invention discloses a numerical control lathe tool abrasion real-time monitoring device, and relates to the technical field of lathe tool abrasion monitoring, the numerical control lathe tool abrasion real-time monitoring device comprises a lathe body and a fixing frame, a three-axis rack is installed on the upper portion of the lathe body, a tool driving end is installed at the adjusting end of the three-axis rack, and a tool is installed at the bottom of the tool driving end; a vibration sensor and a magnetoelastic six-dimensional force sensor are installed on the surface of the vibration sensor, the magnetoelastic six-dimensional force sensor is located at the top of the tool, a fixing frame is fixed to the surface of the three-axis rack, and an industrial camera is fixedly connected to the surface of the fixing frame. According to the invention, in the data acquisition system, a vibration sensor module, a force sensor module and a visual sensor module respectively acquire data in real time at sampling frequencies adapted to the characteristics of each sensor, so that the synchronism and integrity of multi-source signals are ensured; the filtering module carries out targeted filtering according to noise characteristics of the vibration signals and the force signals, characteristic signals in the cutting process are effectively reserved, and the influence of power frequency interference and non-cutting vibration noise is reduced.

Description

Technical Field

[0001] This invention relates to the field of lathe tool wear monitoring technology, specifically a real-time monitoring device for CNC lathe tool wear. Background Technology

[0002] The real-time tool wear monitoring device for CNC lathes is an electromechanical system that integrates sensor technology, signal processing algorithms, and intelligent decision-making mechanisms. Its core function is to acquire tool wear status information in real time during the machining process and realize tool life prediction, adaptive adjustment of machining parameters, and fault early warning through data-driven methods.

[0003] However, in existing technologies, sensor design does not fully consider the impact of complex machining environments, such as high temperature, high vibration, high dust, and cutting fluid impact, on their performance. The principle of a single sensor has inherent limitations, and the data fusion of multiple sensors does not effectively handle the coupling relationship between signals and common-mode noise. At the same time, the monitoring model lacks the ability to adapt to changes in working conditions. In addition, the hardware system has engineering defects such as interface incompatibility and complex installation and maintenance in its integrated design. These problems lead to low monitoring accuracy and high false alarm rate in actual applications, and the inability to identify tool wear status in a timely and accurate manner. This results in fluctuations in machining quality, decreased production efficiency, and increased maintenance costs, which seriously restricts the improvement of the intelligence and reliability of CNC machining. Summary of the Invention

[0004] The purpose of this invention is to provide a real-time monitoring device for tool wear on CNC lathes, so as to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a real-time monitoring device for tool wear on a CNC lathe, comprising a bed and a fixed frame. A three-axis frame is mounted on the upper part of the bed. A tool drive end is mounted on the adjustment end of the three-axis frame. A tool is mounted at the bottom of the tool drive end. A vibration sensor and a magnetoelastic six-dimensional force sensor are respectively mounted on the surface of the vibration sensor. The magnetoelastic six-dimensional force sensor is located on the top of the tool. The fixed frame is fixed to the surface of the three-axis frame, and an industrial camera is fixedly connected to the surface of the fixed frame. A monitoring system is uniformly connected to the vibration sensor, the magnetoelastic six-dimensional force sensor, and the industrial camera. The vibration sensor is used to collect the vibration amplitude, vibration magnitude, and vibration phase of the tool drive end. The magnetoelastic six-dimensional force sensor is used to collect the force values ​​and torque values ​​of the tool in the X, Y, and Z axis directions. The industrial camera is used to collect the geometry and surface morphology of the tool. The monitoring system includes a data acquisition system, a data preprocessing system, a feature extraction system, a feature fusion analysis system, a threshold judgment and early warning system, and a maintenance suggestion and self-learning system. The data acquisition system is used to collect monitoring data from vibration sensors, magnetoelastic six-dimensional force sensors, and industrial cameras in real time. The data preprocessing system is used to remove noise and interference from the collected raw data. The feature extraction system is used to extract feature parameters that can reflect the tool wear state from the preprocessed data and reduce the data dimensionality. The feature fusion analysis system is used to fuse the features extracted by vibration sensors, magnetoelastic six-dimensional force sensors, and industrial cameras. The threshold judgment and early warning system is used to determine whether the tool wear exceeds the safe range based on the set multi-level dynamic thresholds. If it exceeds the safe range, the corresponding early warning is triggered. The maintenance suggestion and self-learning system is used to generate maintenance suggestions based on the tool wear level and update the knowledge base and wear case library.

[0006] Preferably, the data acquisition system includes a vibration sensor module, a force sensor module, and a vision sensor module. The vibration sensor module is used to detect the vibration of the tool during the cutting process based on the monitoring data of the vibration sensor. The force sensor module is used to detect the force on the tool during the cutting process based on the monitoring data of the magnetoelastic six-dimensional force sensor. The vision sensor module is used to detect the geometry and surface morphology of the tool based on the monitoring data of the industrial camera.

[0007] Preferably, the data preprocessing system includes a filtering module and a noise reduction module. The filtering module is used to remove high-frequency noise and low-frequency interference from the data, and the noise reduction module is used to reduce the noise level in the data.

[0008] Preferably, the feature extraction system includes a vibration data feature extraction module, a force data feature extraction module, and a visual data feature extraction module. The vibration data feature extraction module is used to extract specific frequency components and amplitude features from the vibration data. The force data feature extraction module is used to extract features such as the rate of change of cutting force, mean, and variance from the force data of the tool. The visual data feature extraction module is used to extract geometric shape changes and surface morphology features from the tool image.

[0009] Preferably, the feature fusion and analysis system includes a feature fusion module, a comparative analysis module, and an inference prediction module. Specifically, the feature fusion module is used to fuse different types of features such as force, vibration, and vision to form a comprehensive feature vector. The comparative analysis module is used to compare the fused features with a historical wear feature database to find similar wear patterns. The inference prediction module uses a Bayesian network to perform uncertain inference and a deep learning model to predict the wear state of the tool.

[0010] Preferably, the threshold judgment and early warning system includes a dynamic threshold setting module, a wear level judgment module, and an early warning module. The dynamic threshold setting module dynamically adjusts the early warning threshold according to the processing material, tool type, and cutting parameters. The wear level judgment module is used to judge the wear level of the tool. The early warning module is used to trigger an early warning when the wear level of the tool exceeds the threshold.

[0011] Preferably, the maintenance suggestion and self-learning system includes a maintenance suggestion generation module, a knowledge base update module, and a self-learning optimization module. The maintenance suggestion generation module generates corresponding maintenance suggestions based on the wear level of the tool. The knowledge base update module updates the knowledge base with new wear cases and feature information. The self-learning optimization module optimizes the judgment model through reinforcement learning algorithms.

[0012] Preferably, a rinsing pipe is fixedly connected to the surface of the mounting bracket. The rinsing pipe is located above the industrial camera, and multiple nozzles are fixedly connected to the bottom of the rinsing pipe. A water supply device is connected to the outside of the rinsing pipe.

[0013] Preferably, a cleaning brush is fixedly connected to the upper part of the bed, which is used to scrub the surface of the industrial camera, and a clamping assembly is installed on the upper part of the bed.

[0014] Compared with the prior art, the beneficial effects of the present invention are: 1. In this invention, in the data acquisition system, the vibration sensor module, force sensor module, and vision sensor module acquire data in real time at sampling frequencies adapted to the characteristics of each sensor, ensuring the synchronization and integrity of multi-source signals; the filtering module performs targeted filtering on the noise characteristics of vibration and force signals, effectively preserving the characteristic signals in the cutting process and reducing the impact of power frequency interference and non-cutting vibration noise.

[0015] 2. In this invention, the noise reduction module removes cutting fluid splash noise from industrial camera images through wavelet transform, improving image quality and creating conditions for the accurate extraction of subsequent wear features. After preprocessing, vibration, force, and visual data are extracted by each feature extraction module from the corresponding signals, forming multi-dimensional feature information. This provides a comprehensive and reliable input basis for subsequent multi-sensor data fusion and intelligent judgment of wear status, effectively solving the limitations of single sensor monitoring and improving the system's adaptability to complex processing environments and the accuracy of tool wear monitoring.

[0016] 3. In this invention, the vibration features, force features, and visual features are normalized by the feature fusion module, and a confidence matrix is ​​established using DS evidence theory. This effectively integrates the complementary information of multiple sensors to form a high-confidence judgment on states such as "moderate wear," avoiding the one-sidedness of a single sensor. The comparison and analysis module calculates the cosine similarity between the fused features and the historical wear feature database, and combines it with the LSTM neural network prediction of the inference prediction module to achieve intelligent identification of wear levels, improving the reliability of wear state judgment under complex working conditions. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the data acquisition system in a real-time monitoring device for CNC lathe tool wear according to the present invention; Figure 2 This is a flowchart of a real-time monitoring device for CNC lathe tool wear according to the present invention; Figure 3 This is a first three-dimensional structural schematic diagram of a real-time monitoring device for CNC lathe tool wear according to the present invention; Figure 4 This is a schematic diagram of the second three-dimensional structure of a real-time monitoring device for CNC lathe tool wear according to the present invention; Figure 5 This is a schematic diagram of the third three-dimensional structure of a real-time monitoring device for CNC lathe tool wear according to the present invention; Figure 6 This is a three-dimensional structural diagram of the flushing pipe in a real-time monitoring device for CNC lathe tool wear according to the present invention.

[0018] In the diagram: 1. Bed; 2. Three-axis frame; 3. Tool drive end; 4. Vibration sensor; 5. Magnetoelastic six-dimensional force sensor; 6. Fixture; 7. Industrial camera; 8. Cleaning brush; 9. Fixture assembly; 10. Flushing pipe; 11. Nozzle. Detailed Implementation

[0019] 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, and 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.

[0020] Example 1: Refer to Figures 1-6As shown: A real-time monitoring device for tool wear on a CNC lathe includes a bed 1 and a fixed frame 6. A three-axis frame 2 is mounted on the upper part of the bed 1. A tool drive end 3 is mounted on the adjusting end of the three-axis frame 2. A tool is mounted on the bottom of the tool drive end 3. A vibration sensor 4 and a magnetoelastic six-dimensional force sensor 5 are respectively mounted on the surface of the vibration sensor 4. The magnetoelastic six-dimensional force sensor 5 is located on the top of the tool. The fixed frame 6 is fixed to the surface of the three-axis frame 2, and an industrial camera 7 is fixedly connected to the surface of the fixed frame 6. The vibration sensor 4, the magnetoelastic six-dimensional force sensor 5, and the industrial camera 7 are uniformly connected to a monitoring system. The vibration sensor 4 is used to collect the vibration amplitude, vibration amplitude, and vibration phase of the tool drive end 3. The magnetoelastic six-dimensional force sensor 5 is used to collect the force values ​​and torque values ​​of the tool in the X, Y, and Z axis directions. The industrial camera 7 is used to collect the geometry and surface morphology of the tool. The monitoring system includes a data acquisition system, a data preprocessing system, a feature extraction system, a feature fusion analysis system, a threshold judgment and early warning system, and a maintenance suggestion and self-learning system. The data acquisition system is used to collect monitoring data from vibration sensor 4, magnetoelastic six-dimensional force sensor 5, and industrial camera 7 in real time. The data preprocessing system is used to remove noise and interference from the collected raw data. The feature extraction system is used to extract feature parameters that can reflect the tool wear state from the preprocessed data to reduce the data dimensionality. The feature fusion analysis system is used to fuse the features extracted by vibration sensor 4, magnetoelastic six-dimensional force sensor 5, and industrial camera 7. The threshold judgment and early warning system is used to judge whether the tool wear degree exceeds the safe range according to the set multi-level dynamic threshold. If it exceeds the safe range, the corresponding early warning is triggered. The maintenance suggestion and self-learning system is used to generate maintenance suggestions based on the tool wear level and update the knowledge base and wear case library at the same time.

[0021] The data acquisition system includes a vibration sensor module, a force sensor module, and a vision sensor module. The vibration sensor module is used to detect the vibration of the tool during the cutting process based on the monitoring data of the vibration sensor 4. The force sensor module is used to detect the force on the tool during the cutting process based on the monitoring data of the magnetoelastic six-dimensional force sensor 5. The vision sensor module is used to detect the geometry and surface morphology of the tool based on the monitoring data of the industrial camera 7.

[0022] The data preprocessing system includes a filtering module and a noise reduction module. The filtering module is used to remove high-frequency noise and low-frequency interference from the data, while the noise reduction module is used to reduce the noise level in the data.

[0023] The feature extraction system includes a vibration data feature extraction module, a force data feature extraction module, and a visual data feature extraction module. The vibration data feature extraction module is used to extract specific frequency components and amplitude features from vibration data. The force data feature extraction module is used to extract features such as the rate of change of cutting force, mean, and variance from the force data of the tool. The visual data feature extraction module is used to extract geometric shape changes and surface morphology features from the tool image.

[0024] In this embodiment, a three-axis frame 2 is installed on the upper part of the bed 1, and the tool drive end 3 is fixed at its adjustment end. The tool is installed and fixed to the tool drive end 3. The axial force and torque in multiple directions are collected in real time by a magnetoelastic six-dimensional force sensor 5. The vibration sensor 4 is attached to the side of the tool drive end 3 to collect the vibration amplitude, frequency and phase. The fixing frame 6 is fixed to the front end of the three-axis frame 2 by bolts. The lens of the industrial camera 7 is vertically aligned with the cutting edge of the tool to collect the tool image in real time.

[0025] The data acquisition system reads the acceleration signal of vibration sensor 4 in real time at a sampling rate of 10kHz through the vibration sensor module, and the force sensor module simultaneously acquires the six-dimensional force data of magnetoelastic six-dimensional force sensor 5 at a sampling rate of 5kHz. The vision sensor module captures a frame of tool image every 20ms through industrial camera 7. At the same time, the filtering module uses a 50Hz low-pass filter to remove power frequency interference from the vibration signal and a 10-500Hz band-pass filter to retain the cutting characteristic frequency from the force signal. The noise reduction module uses wavelet transform to denoise the image of industrial camera 7 and remove the salt-and-pepper noise generated by cutting fluid splashing.

[0026] Subsequently, the vibration data feature extraction module calculates the root mean square value and the proportion of the dominant frequency energy of the vibration signal; the force data feature extraction module calculates the mean and variance of the cutting force; and the visual data feature extraction module measures the width VB of the flank wear band and the radius R of the tool tip arc using an edge detection algorithm.

[0027] The magnetoelastic six-dimensional force sensor 5 collects axial forces and torques in multiple directions in real time, which can comprehensively capture the stress state of the tool during the cutting process and provide multi-dimensional data support for analyzing the stress anomalies caused by tool wear. The vibration sensor 4 is attached to the side of the tool drive end 3 to collect vibration amplitude, frequency and phase, which can sensitively identify the vibration characteristic changes caused by tool wear, such as abnormal vibration signals during chatter or chipping. The mounting bracket 6 is fixed to the front end of the three-axis frame 2 by bolts, so that the lens of the industrial camera 7 is perpendicular to the cutting edge of the tool, ensuring that the real-time acquired tool image clearly reflects the details of the cutting edge wear and avoiding monitoring blind spots caused by viewing angle deviation.

[0028] In the data acquisition system, the vibration sensor module, force sensor module, and vision sensor module acquire data in real time at sampling frequencies adapted to the characteristics of each sensor, ensuring the synchronization and integrity of multi-source signals. The filtering module performs targeted filtering on the noise characteristics of vibration and force signals, effectively preserving characteristic signals during the cutting process and reducing the impact of power frequency interference and non-cutting vibration noise. The noise reduction module removes cutting fluid splash noise from the industrial camera 7 images through wavelet transform, improving image quality and creating conditions for the accurate extraction of subsequent wear features. After preprocessing, the vibration, force, and vision data are used by each feature extraction module to extract key parameters reflecting the tool wear state from the corresponding signals, forming multi-dimensional feature information. This provides a comprehensive and reliable input basis for subsequent multi-sensor data fusion and intelligent wear state judgment, effectively solving the limitations of single-sensor monitoring and improving the system's adaptability to complex machining environments and the accuracy of tool wear monitoring.

[0029] Example 2: According to Figures 1-6 As shown, the feature fusion and analysis system includes a feature fusion module, a comparative analysis module, and an inference and prediction module. Specifically, the feature fusion module is used to fuse different types of features such as force, vibration, and vision to form a comprehensive feature vector. The comparative analysis module is used to compare the fused features with the historical wear feature database to find similar wear patterns. The inference and prediction module uses a Bayesian network to perform uncertain inference and a deep learning model to predict the wear state of the tool. The threshold judgment and early warning system includes a dynamic threshold setting module, a wear level judgment module, and an early warning module. The dynamic threshold setting module dynamically adjusts the early warning threshold according to the processing material, tool type, and cutting parameters. The wear level judgment module is used to judge the wear level of the tool. The early warning module is used to trigger an early warning when the wear level of the tool exceeds the threshold. The maintenance suggestion and self-learning system includes a maintenance suggestion generation module, a knowledge base update module, and a self-learning optimization module. The maintenance suggestion generation module generates corresponding maintenance suggestions based on the wear level of the tool. The knowledge base update module updates the knowledge base with new wear cases and feature information. The self-learning optimization module optimizes the judgment model through reinforcement learning algorithms. A flushing pipe 10 is fixedly connected to the surface of the mounting bracket 6. The flushing pipe 10 is located above the industrial camera 7, and multiple nozzles 11 are fixedly connected to the bottom of the flushing pipe 10. A water supply device is connected to the flushing pipe 10. A cleaning brush 8 is fixedly connected to the upper part of the bed 1. The cleaning brush 8 is used to brush the surface of the industrial camera 7. A clamping assembly 9 is installed on the upper part of the bed 1.

[0030] In this embodiment, the vibration feature RMS, dominant frequency, force feature Fz_avg, σ_Fx, visual feature VB, and R are normalized into feature vectors ranging from 0 to 1 through the feature fusion module, and a sensor confidence matrix is ​​established using DS evidence theory. For example, when vibration RMS > 0.8g, Fz_avg > 400N, and VB > 0.2mm, the fused evidence supports a 92% confidence level for "moderate wear".

[0031] The comparison and analysis module then calculates the cosine similarity between the fused features and the historical wear feature database, with a threshold of 0.75. The inference and prediction module uses an LSTM neural network to predict the wear level by inputting the feature vector. The training set accuracy is 95%, and the test set generalization error is ≤8%.

[0032] An industrial camera 7 mounted on the surface of the mounting bracket 6 captures real-time images of the tool's back face. A cleaning brush 8 automatically cleans the lens surface every 5 minutes to ensure image clarity. A fixture assembly 9 secures the workpiece and works in conjunction with the three-axis frame 2 to achieve precise positioning.

[0033] The dynamic threshold setting module calculates the warning threshold in real time based on the processing parameters. For example, the wear level judgment module divides the VB value into three levels: slight VB<0.3mm, moderate 0.3mm≤VB<0.6mm, and severe VB≥0.6mm, corresponding to three levels of warning yellow light, buzzer alarm, and machine tool emergency stop.

[0034] The rinsing pipe 10 is located above the industrial camera 7, connected to an external water supply device with a water pressure of 0.3 MPa. The nozzle 11 sprays deionized water every 10 minutes to rinse away metal dust from the lens surface. The industrial camera 7 moves along the X-axis of the three-axis frame 2 and is cleaned by the cleaning brush 8, which removes stubborn stains. When the system determines that the wear is "moderate," the maintenance suggestion generation module automatically sends a command to the CNC to reduce the feed rate by 15%. If a "severe wear" warning is triggered, an emergency stop signal is sent within 0.2 seconds, a tool change order is generated, and the wear cases of the tool in the knowledge base are updated simultaneously.

[0035] The feature fusion module normalizes vibration features (RMS, dominant frequency), force features (Fz_avg, σ_Fx), and visual features (VB, R), and establishes a confidence matrix using DS evidence theory. This effectively integrates complementary information from multiple sensors, forming a high-confidence judgment on states such as "moderate wear," avoiding the bias of a single sensor. The comparison and analysis module calculates the cosine similarity between the fused features and the historical wear feature database, and combines this with the LSTM neural network prediction of the inference prediction module to achieve intelligent identification of wear levels, improving the reliability of wear state judgment under complex working conditions. The industrial camera 7 on the surface of the mounting bracket 6, together with the cleaning brush 8 and the nozzle 11, forms a self-cleaning system. By automatically brushing the lens surface every 5 minutes and spraying deionized water every 10 minutes, it effectively removes metal dust and stubborn stains, ensuring that the tool back face image captured by the industrial camera 7 in real time is clear, providing high-quality data for visual feature extraction. The fixture assembly 9 and the three-axis frame 2 work together to achieve precise positioning, ensuring the relative positional accuracy of the tool and the workpiece, laying the foundation for the stability of the cutting process and the accuracy of monitoring data. The dynamic threshold setting module adjusts the warning threshold in real time based on machining parameters, while the wear level judgment module, combined with VB values, divides the warning into three levels and corresponds to different response measures (yellow light, buzzer alarm, and machine tool emergency stop), enabling the system to make intelligent decisions based on the degree of wear. When the wear level is judged as "moderate wear," the feed rate is automatically adjusted; when the wear level is "severe," the system quickly stops and generates a tool change order, while simultaneously updating the knowledge base. This achieves closed-loop optimization from wear monitoring to machining control and maintenance management, effectively improving the safety, stability, and intelligent level of tool life management in the machining process. It also solves the problems of poor adaptability to working conditions, delayed warnings, and high maintenance costs in existing technologies.

[0036] The method of use and working principle of this device: When using this invention, a three-axis frame 2 is installed on the upper part of the bed 1, and the tool drive end 3 is fixed at its adjusting end. The tool is installed and fixed to the tool drive end 3. The magnetoelastic six-dimensional force sensor 5 collects axial force and torque in multiple directions in real time. The vibration sensor 4 is attached to the side of the tool drive end 3 to collect vibration amplitude, frequency and phase. The fixing bracket 6 is fixed to the front end of the three-axis frame 2 by bolts. The lens of the industrial camera 7 is vertically aligned with the cutting edge of the tool to collect tool images in real time.

[0037] The data acquisition system reads the acceleration signal from vibration sensor 4 in real time at a sampling rate of 10kHz via the vibration sensor module, and the force sensor module simultaneously acquires the six-dimensional force data from magnetoelastic six-dimensional force sensor 5 at a sampling rate of 5kHz. The vision sensor module captures one frame of the tool image every 20ms via industrial camera 7. Meanwhile, the filtering module uses a 50Hz low-pass filter to remove power frequency interference from the vibration signal and a 10-500Hz band-pass filter to retain the cutting characteristic frequency from the force signal; the noise reduction module uses wavelet transform to denoise the images from the industrial camera 7 and remove the salt-and-pepper noise generated by the splashing of cutting fluid.

[0038] Subsequently, the vibration data feature extraction module calculates the root mean square value and the proportion of the dominant frequency energy of the vibration signal; the force data feature extraction module calculates the mean and variance of the cutting force; and the visual data feature extraction module measures the width VB of the flank wear band and the radius R of the tool tip arc using an edge detection algorithm.

[0039] The vibration features RMS, dominant frequency, force features Fz_avg, σ_Fx, visual features VB, and R are normalized into feature vectors in the range of 0-1 through the feature fusion module, and the sensor confidence matrix is ​​established using DS evidence theory.

[0040] For example, when the vibration RMS > 0.8g, Fz_avg > 400N, and VB > 0.2mm, the fused evidence supports "moderate wear" with a confidence level of 92%.

[0041] The comparison and analysis module then calculates the cosine similarity between the fused features and the historical wear feature database, with a threshold set to 0.75. The inference prediction module uses an LSTM neural network to predict wear level by inputting feature vectors. The accuracy on the training set is 95%, and the generalization error on the test set is ≤8%.

[0042] The dynamic threshold setting module calculates the warning threshold in real time based on the processing parameters. For example, the wear level judgment module divides the VB value into three levels: slight VB<0.3mm, moderate 0.3mm≤VB<0.6mm, and severe VB≥0.6mm, corresponding to three levels of warning yellow light, buzzer alarm, and machine tool emergency stop.

[0043] The rinsing pipe 10 is located above the industrial camera 7, and is connected to an external water supply device with a water pressure of 0.3MPa. The nozzle 11 sprays deionized water every 10 minutes to rinse the metal dust on the lens surface. The industrial camera 7 moves along the X-axis of the three-axis frame 2 and is cleaned by the cleaning brush 8. The cleaning brush 8 cleans the industrial camera 7 to remove stubborn stains.

[0044] When the system determines that the wear is "moderate", the maintenance suggestion generation module automatically sends an instruction to the CNC to reduce the feed rate by 15%; if the "severe wear" warning is triggered, an emergency stop signal is sent within 0.2 seconds and a tool change work order is generated, and the wear cases of the tool in the knowledge base are updated simultaneously.

[0045] Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A real-time monitoring device for tool wear on a CNC lathe, comprising a lathe bed (1) and a mounting bracket (6), wherein a three-axis frame (2) is mounted on the upper part of the lathe bed (1), a tool drive end (3) is mounted on the adjusting end of the three-axis frame (2), a tool is mounted on the bottom of the tool drive end (3), and a vibration sensor (4) and a magnetoelastic six-dimensional force sensor (5) are respectively mounted on the surface of a vibration sensor (4), the magnetoelastic six-dimensional force sensor (5) being located on the top of the tool, the mounting bracket (6) being fixed on the surface of the three-axis frame (2), and an industrial camera (7) being fixedly connected to the surface of the mounting bracket (6), characterized in that: The vibration sensor (4), the magnetoelastic six-dimensional force sensor (5), and the industrial camera (7) are uniformly connected to a monitoring system. The vibration sensor (4) is used to collect the vibration amplitude, vibration magnitude, and vibration phase of the tool drive end (3). The magnetoelastic six-dimensional force sensor (5) is used to collect the force values ​​of the tool in the X, Y, and Z axis directions and the torque values ​​in the X, Y, and Z axis directions. The industrial camera (7) is used to collect the geometry of the tool and the surface morphology of the tool. The monitoring system includes a data acquisition system, a data preprocessing system, a feature extraction system, a feature fusion analysis system, a threshold judgment and early warning system, and a maintenance suggestion and self-learning system. The data acquisition system is used to collect monitoring data from the vibration sensor (4), the magnetoelastic six-dimensional force sensor (5), and the industrial camera (7) in real time. The data preprocessing system is used to remove noise and interference from the collected raw data. The feature extraction system is used to extract feature parameters that can reflect the wear state of the tool from the preprocessed data and reduce the data dimension. The feature fusion analysis system is used to fuse the features extracted by the vibration sensor (4), the magnetoelastic six-dimensional force sensor (5), and the industrial camera (7). The threshold judgment and early warning system is used to judge whether the wear degree of the tool exceeds the safe range according to the set multi-level dynamic threshold. If it exceeds the safe range, the corresponding early warning will be triggered. The maintenance suggestion and self-learning system is used to generate maintenance suggestions according to the wear level of the tool and update the knowledge base and wear case library at the same time.

2. The real-time monitoring device for CNC lathe tool wear according to claim 1, characterized in that: The data acquisition system includes a vibration sensor module, a force sensor module and a vision sensor module. The vibration sensor module is used to detect the vibration of the tool during the cutting process based on the monitoring data of the vibration sensor (4). The force sensor module is used to detect the force on the tool during the cutting process based on the monitoring data of the magnetoelastic six-dimensional force sensor (5). The vision sensor module is used to detect the geometry and surface morphology of the tool based on the monitoring data of the industrial camera (7).

3. The real-time monitoring device for CNC lathe tool wear according to claim 1, characterized in that: The data preprocessing system includes a filtering module and a noise reduction module. The filtering module is used to remove high-frequency noise and low-frequency interference from the data, while the noise reduction module is used to reduce the noise level in the data.

4. The real-time monitoring device for CNC lathe tool wear according to claim 1, characterized in that: The feature extraction system includes a vibration data feature extraction module, a force data feature extraction module, and a visual data feature extraction module. The vibration data feature extraction module is used to extract specific frequency components and amplitude features from vibration data. The force data feature extraction module is used to extract features such as the rate of change of cutting force, mean, and variance from the force data of the tool. The visual data feature extraction module is used to extract geometric shape changes and surface morphology features from the tool image.

5. The real-time monitoring device for CNC lathe tool wear according to claim 1, characterized in that: The feature fusion and analysis system includes a feature fusion module, a comparative analysis module, and an inference and prediction module. The feature fusion module is used to fuse different types of features such as force, vibration, and vision to form a comprehensive feature vector. The comparative analysis module is used to compare the fused features with the historical wear feature database to find similar wear patterns. The inference and prediction module uses a Bayesian network to perform uncertain inference and a deep learning model to predict the wear state of the tool.

6. The real-time monitoring device for CNC lathe tool wear according to claim 1, characterized in that: The threshold judgment and early warning system includes a dynamic threshold setting module, a wear level judgment module, and an early warning module. The dynamic threshold setting module dynamically adjusts the early warning threshold according to the machining material, tool type, and cutting parameters. The wear level judgment module is used to judge the wear level of the tool. The early warning module is used to trigger an early warning when the wear level of the tool exceeds the threshold.

7. The real-time monitoring device for CNC lathe tool wear according to claim 1, characterized in that: The maintenance suggestion and self-learning system includes a maintenance suggestion generation module, a knowledge base update module, and a self-learning optimization module. The maintenance suggestion generation module generates corresponding maintenance suggestions based on the wear level of the tool. The knowledge base update module updates the knowledge base with new wear cases and feature information. The self-learning optimization module optimizes the judgment model through reinforcement learning algorithms.

8. The real-time monitoring device for CNC lathe tool wear according to claim 1, characterized in that: A flushing pipe (10) is fixedly connected to the surface of the mounting bracket (6). The flushing pipe (10) is located above the industrial camera (7), and multiple nozzles (11) are fixedly connected to the bottom of the flushing pipe (10). A water supply device is connected to the outside of the flushing pipe (10).

9. The real-time monitoring device for CNC lathe tool wear according to claim 1, characterized in that: A cleaning brush (8) is fixedly connected to the upper part of the bed (1). The cleaning brush (8) is used to clean the surface of the industrial camera (7). A clamping assembly (9) is installed on the upper part of the bed (1).

Images