Integrated numerical control intelligent machine tool

Abstract

The application discloses an integrated numerical control intelligent machine tool, which comprises a machine tool bed body, a numerical control system, a spindle unit and a feeding transmission unit, and further comprises a multi-source fusion perception module, an embedded intelligent decision unit, an in-situ detection module, a digital twin mapping module and an integrated auxiliary function unit. The integrated numerical control intelligent machine tool disclosed by the application organically combines multi-functional unit deep integration, multi-source fusion perception, embedded edge intelligent decision and digital twin closed-loop iteration, constructs a full-link intelligent processing system of perception, decision, execution, detection and twin correction, and significantly improves the integration degree, processing precision, processing efficiency and autonomous intelligent level of the machine tool.

Description

Technical Field

[0001] This invention belongs to the field of CNC machine tool technology, specifically relating to an integrated intelligent CNC machine tool that integrates multi-source sensing, edge intelligent decision-making, in-situ detection, and digital twin closed loop. Background Technology

[0002] CNC machine tools are the core working machines of high-end equipment manufacturing, and their machining accuracy, operating efficiency, and level of intelligence directly determine the core competitiveness of the precision manufacturing industry. As the requirements for machining accuracy and efficiency of parts continue to increase in fields such as aerospace, automotive molds, and precision instruments, the technical shortcomings of traditional CNC machine tools are becoming increasingly prominent.

[0003] Current CNC machine tools generally adopt an operating mode of "offline programming, preset parameters, open-loop machining, and offline testing," which mainly has the following technical defects: First, the functional units are arranged separately, resulting in low integration. The processing unit, dimensional inspection unit, chip removal unit, temperature control unit, and tool management unit are independent of each other and mostly adopt external or add-on layouts. This results in a large space occupation for the whole machine, and there are delays in data transmission and process connection between units, making it impossible to achieve seamless connection of the processing flow and further compress the processing cycle time.

[0004] Second, the perception dimension is limited and process data is lacking. Traditional CNC machine tools rely solely on linear encoders and grating rulers to achieve semi-closed-loop or full-closed-loop position control. They cannot collect multi-physical field operating condition data such as cutting force, spindle vibration, acoustic emission, and multi-node temperature in real time. They lack the ability to perceive key factors affecting machining accuracy, such as tool wear, cutting chatter, and thermal deformation. Machining errors can often only be detected in post-process inspections, and real-time intervention and dynamic compensation are impossible.

[0005] Third, the process parameters are fixed and lack adaptability. The cutting parameters of existing machine tools mostly rely on offline settings by process engineers, and the machine operates with fixed parameters during the machining process. It is impossible to dynamically optimize according to the fluctuation of workpiece material characteristics, the gradual change of tool wear, and the changes in cutting conditions. It is difficult to simultaneously take into account the machining efficiency and machining stability, and it is easy to have problems such as low efficiency due to conservative cutting parameters, or tool breakage and workpiece scrap due to aggressive parameters.

[0006] Fourth, error compensation is lagging and lacks self-evolution capabilities. Existing in-machine inspection technologies are mostly sampling measurements after the process, used only for pass / fail determination, and do not form a deep closed loop with the machining control system; at the same time, they lack virtual-real mapping and error prediction mechanisms at the digital twin level, resulting in significant lag in error compensation, and the machine tool model cannot be iteratively optimized according to the operating status, making it difficult to guarantee accuracy consistency after long-term operation. Summary of the Invention

[0007] The main objective of this invention is to provide an integrated CNC intelligent machine tool. By organically combining multi-functional unit deep integration, multi-source fusion perception, embedded edge intelligent decision-making, and digital twin closed-loop iteration, a full-link intelligent machining system of perception, decision-making, execution, detection, and twin correction is constructed, which significantly improves the integration, machining accuracy, machining efficiency, and autonomous intelligence level of the machine tool.

[0008] Another objective of this invention is to provide an integrated CNC intelligent machine tool that enables autonomous perception, autonomous decision-making, and autonomous iterative optimization throughout the entire machining process.

[0009] To achieve the above objectives, this invention provides an integrated CNC intelligent machine tool, comprising a machine tool bed, a CNC system, a spindle unit, and a feed transmission unit, and further comprising a multi-source fusion sensing module, an embedded intelligent decision-making unit, an on-machine in-situ detection module, a digital twin mapping module, and an integrated auxiliary function unit, wherein: The multi-source fusion sensing module is deployed in the spindle unit, feed transmission unit and machining area to collect multi-physics field condition data in real time during the machining process. The embedded intelligent decision-making unit has a built-in process parameter self-optimization model and error compensation algorithm, receives operating condition data from the multi-source fusion sensing module and outputs dynamic adjustment instructions; The in-situ detection module is integrated inside the machining compartment of the machine tool bed and is used to perform non-contact dimensional and positional error detection on the workpiece during the machining cycle. The digital twin mapping module communicates bidirectionally with the multi-source fusion sensing module, the in-situ detection module, and the CNC system to construct a digital twin of the machine tool in real time and form a closed-loop feedback link. The integrated auxiliary function unit is integrated inside the machine tool bed and includes an adaptive chip removal unit, a precise temperature control unit, and a tool status monitoring unit.

[0010] As a further preferred embodiment of the above technical solution, the multi-source fusion sensing module includes a triaxial piezoelectric cutting force sensor, a triaxial vibration acceleration sensor, a broadband acoustic emission sensor, a distributed multi-point temperature sensor, and a high-definition industrial vision acquisition unit, wherein: The triaxial piezoelectric cutting force sensor is embedded below the worktable surface of the workpiece clamping table and is used to collect the X / Y / Z three-axis cutting force and cutting torque in real time. The triaxial vibration acceleration sensor is installed on the radial and axial end faces of the front bearing housing of the spindle to collect the time-domain and frequency-domain vibration signals of the spindle during the cutting process; The broadband acoustic emission sensor is attached to the side wall of the machining chamber near the cutting area to capture high-frequency acoustic emission signals generated by tool breakage and workpiece micro-cracks. The distributed multi-point temperature sensors are deployed at multiple key locations; The high-definition industrial vision acquisition unit is installed on the side of the spindle box and is used to monitor the chip accumulation state in the cutting zone and the surface morphology of the workpiece in real time.

[0011] As a further preferred embodiment of the above technical solution, the multi-source fusion sensing module further includes a data preprocessing subunit, which is implemented based on an FPGA hardware acceleration architecture and executes the following sequentially: Based on the clock protocol, the timestamps of multiple data collected by the triaxial piezoelectric cutting force sensor, triaxial vibration acceleration sensor, broadband acoustic emission sensor, distributed multi-point temperature sensor and high-definition industrial vision acquisition unit are aligned. An adaptive wavelet threshold filtering algorithm is used to remove cutting background noise and electromagnetic interference signals; Extract the time-domain, frequency-domain, and time-frequency-domain features of each signal, and perform feature dimensionality reduction and data standardization through principal component analysis; Finally, the processed feature dataset is transmitted to the embedded intelligent decision unit in real time via industrial Ethernet.

[0012] As a further preferred technical solution of the above technical solution, the embedded intelligent decision-making unit adopts an edge computing heterogeneous architecture and has built-in a process parameter self-optimization model and a thermal coupling error prediction model based on deep reinforcement learning. The process parameter self-optimization model uses real-time cutting force, vibration, and temperature characteristics as the state space, the adjustment amount of spindle speed, feed rate, depth of cut, and width of cut as the action space, and the weighted comprehensive value of material removal rate, machining stability, and tool wear rate as the reward function to output the optimal process parameter adjustment command. The thermo-coupling error prediction model takes multi-point temperature data and cutting force data at key locations as input, and predicts the thermal deformation error and stress deformation error of each axis of the machine tool through a trained convolutional neural network. The output of optimal process parameter adjustment instructions and error compensation instructions are then directly connected to the position loop feedforward channel of the CNC system.

[0013] As a further preferred technical solution to the above technical solution, the in-situ detection module includes a high-precision line laser probe and a binocular vision positioning subunit; The line laser probe is installed on the inner side wall of the processing chamber via a pneumatic telescopic protective mechanism and is stored in a sealed protective chamber when not in use. After processing, the workpiece does not need to be disassembled and clamped. The line laser probe moves along an automatically planned scanning path under the drive of the CNC system to acquire three-dimensional point cloud data of the entire surface of the workpiece. The binocular vision positioning subunit is used to identify workpiece clamping posture deviation before processing, automatically calculate the workpiece coordinate system offset and complete the correction; at the same time, it can perform imaging detection on the tool cutting edge to identify the amount of wear and chipping defects on the back face. The in-situ detection module also has a built-in standard gauge block in-situ calibration component, which automatically completes the probe accuracy calibration before each test.

[0014] As a further preferred technical solution to the above technical solution, the digital twin mapping module includes a geometric model, a multiphysics physical property model, and a cutting process model; The geometric model is constructed based on the three-dimensional CAD model of the machine tool, accurately mapping the dimensions, assembly relationships, and kinematic pair constraints of each moving part of the machine tool. The multiphysics model incorporates material thermal conductivity parameters, structural stiffness matrix, and damping characteristics to simulate the temperature field distribution and stress deformation distribution of the machine tool in real time. The cutting process model predicts cutting force, surface roughness, and tool wear rate. The digital twin operates synchronously with the physical machine tool in real time, predicting the error change trend during future processing periods. When the prediction deviation exceeds a set threshold, a pre-compensation command is sent to the embedded intelligent decision-making unit in advance. The measured data from in-situ detection is used to correct the twin model parameters in reverse, realizing continuous iteration of model accuracy.

[0015] As a further preferred embodiment of the above technical solution, the adaptive chip removal unit includes a built-in spiral chip removal mechanism and a zoned high-pressure chip flushing subsystem, wherein: The partitioned high-pressure chip removal subsystem is equipped with nozzles in the processing area and chip removal channel respectively. The chip concentration at the chip removal port is detected in real time by an infrared photoelectric sensor. The chip removal pressure and nozzle opening and closing are automatically adjusted according to the cutting load and chip concentration to achieve chip removal on demand. The precision temperature control unit adopts a multi-loop partitioned independent liquid cooling architecture, which corresponds to three temperature control areas: the spindle box, the three-axis lead screw pair, and the bed base. Each loop is equipped with an independent PID control unit and a flow regulating valve, which suppresses thermal deformation from the source by homogenizing the overall temperature field of the machine tool.

[0016] As a further preferred technical solution to the above technical solution, the tool condition monitoring unit is integrated on the tool changing path on the side of the tool magazine, including a contact trigger probe and a side-light microscopic image acquisition component. During each tool change, the tool passes through the monitoring station in sequence: first, the contact trigger probe completes the high-precision calibration of the tool length and radius, and then the side-light microscopic image acquisition component captures the image of the tool cutting edge; The width of the wear band and the chipping size on the back face of the tool are calculated by image segmentation and edge extraction algorithms, and the tool wear level is automatically classified. Wear and dimensional data are synchronously transmitted to the digital twin mapping module and the embedded intelligent decision-making unit to correct tool compensation values ​​and adjust cutting process parameters; when tool breakage exceeds the safety threshold, an emergency tool change and machining pause command are immediately triggered.

[0017] As a further preferred technical solution of the above technical solution, the CNC system is an open CNC system. After receiving the adjustment instructions from the embedded intelligent decision unit, it adopts a composite compensation strategy of look-ahead pre-interpolation and real-time error superposition to superimpose the thermal deformation compensation amount and the stress deformation compensation amount into the position instructions of the corresponding feed axis, and corrects the motion trajectory of each axis in real time through a high-speed interpolation algorithm. The CNC system supports offline process import and online process parameter iterative optimization, stores process data and inspection data of multiple batches of machining, and automatically updates the experience base of the process parameter optimization model; it also has a machining anomaly graded early warning function, which identifies chatter, tool wear, and chip accumulation anomalies based on multi-source sensing data and issues graded early warnings.

[0018] As a further preferred technical solution to the above technical solution, the machine tool's workflow is as follows: Step S1, Pre-processing datum calibration and twin model initialization: Step S1.1: After the workpiece is clamped, the binocular vision positioning subunit is activated to acquire the workpiece pose image, calculate the offset between the actual coordinate system and the theoretical coordinate system of the workpiece, and automatically correct the workpiece coordinate system in the CNC system. Step S1.2: After the tool is loaded into the spindle, the tool condition monitoring unit performs initial length and radius calibration and edge wear condition detection on the tool, and records the initial parameters of the tool; Step S1.3: Import all machine tool initial state parameters, workpiece parameters, tool parameters and preset process parameters into the digital twin mapping module to construct a reference digital twin that corresponds one-to-one with the physical machine tool and complete the initialization; Step S2: Real-time acquisition and preprocessing of multi-source data during the processing: Step S2.1: After the machining process is started, the multi-source fusion sensing module synchronously collects multiple working condition data such as cutting force, vibration, acoustic emission, temperature and visual images at the set sampling frequency; Step S2.2: The data preprocessing subunit performs timestamp alignment, noise filtering, feature extraction and dimensionality reduction standardization on the multi-channel heterogeneous data to generate a structured feature dataset, which is then transmitted to the embedded intelligent decision unit in real time. Step S3, Embedded intelligent decision-making and dynamic processing control: Step S3.1: The embedded intelligent decision-making unit inputs real-time feature data into the process parameter self-optimization model to determine the current cutting state: if the cutting force fluctuates too much, chattering occurs, or tool wear accelerates, the feed rate and depth of cut are reduced; if the cutting state is stable and the allowance is sufficient, the cutting parameters are gradually increased to improve the material removal rate. Meanwhile, the thermo-coupling error prediction model calculates the thermal deformation and stress deformation error of each feed axis based on real-time temperature and cutting force data, generates position compensation commands and sends them directly to the position loop feedforward channel of the CNC system to achieve real-time dynamic error compensation during the machining process. Step S3.2: The digital twin synchronously maps the physical machine tool's operating status and predicts the error changes in subsequent processing stages in advance. If the prediction deviation exceeds the threshold, the pre-compensation is triggered in advance. Step S4, in-situ inspection and accuracy data acquisition: Step S4.1: After a single processing step is completed or all processing steps are completed, the line laser probe extends out of the protective chamber and performs a full-size scan of the workpiece surface according to the automatically planned scanning path to obtain the workpiece's three-dimensional point cloud data. Step S4.2: Denoise, stitch and fit the point cloud data, calculate the measured values ​​of each dimension feature, form and position tolerance and surface roughness of the workpiece, and compare them with the design tolerance to generate an accuracy inspection report; Step S5, Digital Twin Model Correction and Process Iteration Optimization: Step S5.1: The digital twin mapping module compares the measured accuracy data obtained from the in-situ detection with the predicted data of the twin model and calculates the deviation between the two. Step S5.2: Correct the physical parameters such as thermal conductivity coefficient, structural stiffness coefficient, and cutting force coefficient in the twin model through parameter optimization algorithm to improve the prediction accuracy of subsequent processing; At the same time, the processing condition data, process parameters and test results are stored in the process experience database, and the process parameter self-optimization model is incrementally trained to realize the autonomous iterative evolution of processing capabilities. Repeat steps S1 to S5 to complete continuous self-optimization for multiple batches of processing.

[0019] The beneficial effects of this invention are as follows: 1. The highly integrated architecture of the whole machine significantly improves the efficiency of process connection. This invention deeply integrates multi-functional units such as sensing, detection, chip removal, temperature control, and tool monitoring into the main body of the machine tool bed. It adopts an integrated closed layout, which greatly reduces the space occupied by the whole machine, eliminates the data transmission delay and process transfer time between separate units, realizes the seamless connection of processing, inspection and auxiliary processes, and effectively compresses the cycle time of single-piece processing.

[0020] 2. Multi-source fusion and all-dimensional perception, breaking through the limitations of traditional closed-loop systems. This invention constructs a multi-physics field perception network integrating cutting force, vibration, acoustic emission, temperature, and vision. Combined with a preprocessing subunit accelerated by FPGA hardware, it achieves high-precision time synchronization and feature extraction of heterogeneous data, comprehensively capturing the thermal coupling effect, tool wear state, and cutting stability during the machining process. This provides a rich and accurate data foundation for intelligent decision-making, completely breaking through the perception limitations of traditional CNC machine tools that only have a position closed loop.

[0021] 3. Edge-embedded intelligent decision-making significantly enhances real-time performance and stability. This invention employs an edge computing heterogeneous architecture to deploy a deep reinforcement learning process optimization model and a thermally coupled error prediction model, achieving millisecond-level response without relying on cloud servers. Process parameters can be dynamically optimized based on real-time operating conditions, balancing machining efficiency and tool life; error compensation is directly integrated into the position loop feedforward channel, resulting in a short compensation response cycle and effectively suppressing the impact of cutting deformation and thermal deformation on machining accuracy.

[0022] 4. Digital Twin Virtual-Real Closed Loop, Achieving Autonomous Iterative Evolution. This invention constructs a digital twin integrating geometric, physical, and technological layers. It can not only map the machine tool's operating status in real time but also predict error change trends in advance, achieving proactive error compensation. More importantly, by reverse-calibrating the twin model parameters using measured data from on-site machine inspections, a self-evolving closed loop of virtual-real mapping, prediction compensation, measured correction, and model iteration is formed. As the number of processing batches increases, the model's prediction accuracy and process optimization effect continuously improve, enabling the machine tool to have autonomous learning capabilities.

[0023] 5. A comprehensive intelligent machining system with upgraded overall machining performance. This invention constructs a complete intelligent closed loop, from pre-machining calibration, in-machining perception and compensation, to post-machining detection and model iterative optimization. Compared to traditional open-loop machine tools, it significantly improves machining dimensional consistency, geometric accuracy, and surface quality, greatly reduces the need for manual intervention, and effectively extends tool life and reduces machining energy consumption. It is suitable for high-precision, high-efficiency machining scenarios in high-end manufacturing fields such as aerospace and precision molds. Attached Figure Description

[0024] Figure 1 This is a schematic diagram of the overall structure of the present invention.

[0025] Figure 2 This is a schematic diagram of the workflow of the present invention. Detailed Implementation

[0026] The following description is intended to disclose the present invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art. The basic principles of the invention defined in the following description can be applied to other embodiments, modifications, improvements, equivalents, and other technical solutions that do not depart from the spirit and scope of the invention.

[0027] In the preferred embodiments of the present invention, those skilled in the art should note that the machine tool bed, CNC system, spindle unit, and feed transmission unit involved in the present invention can be regarded as prior art.

[0028] Preferred embodiment.

[0029] like Figure 1 As shown, this invention discloses an integrated CNC intelligent machine tool, including a machine tool bed, a CNC system, a spindle unit, and a feed transmission unit. It also includes a multi-source fusion sensing module, an embedded intelligent decision-making unit, an in-situ detection module, a digital twin mapping module, and an integrated auxiliary function unit, wherein: The multi-source fusion sensing module is deployed in the spindle unit, feed transmission unit and machining area to collect multi-physics field condition data in real time during the machining process. The multi-source fusion sensing module includes a triaxial piezoelectric cutting force sensor, a triaxial vibration acceleration sensor, a broadband acoustic emission sensor, a distributed multi-point temperature sensor, and a high-definition industrial vision acquisition unit, wherein: The triaxial piezoelectric cutting force sensor is embedded below the worktable surface of the workpiece clamping table, with a sampling frequency of not less than 15kHz, and is used to collect the X / Y / Z three-axis cutting force and cutting torque in real time. The triaxial vibration acceleration sensor is installed on the radial and axial end faces of the front bearing housing of the spindle to collect the time-domain and frequency-domain vibration signals of the spindle during the cutting process; The broadband acoustic emission sensor is attached to the side wall of the machining chamber near the cutting area to capture high-frequency acoustic emission signals generated by tool breakage and workpiece micro-cracks. The distributed multi-point temperature sensor has no less than 16 temperature measurement nodes, which are respectively deployed at multiple key locations such as the front and rear bearings of the spindle, the stator, the lead screw and nut pair, the guide rail slider, the bed base and the coolant circuit. The high-definition industrial vision acquisition unit is installed on the side of the spindle box and is used to monitor the chip accumulation state in the cutting zone and the surface morphology of the workpiece in real time.

[0030] The multi-source fusion sensing module also includes a data preprocessing subunit, which is implemented based on an FPGA hardware acceleration architecture and executes the following sequentially: Based on the IEEE 1588 precision clock protocol, timestamps are aligned on multiple data streams acquired by a triaxial piezoelectric cutting force sensor, a triaxial vibration acceleration sensor, a broadband acoustic emission sensor, a distributed multi-point temperature sensor, and a high-definition industrial vision acquisition unit. An adaptive wavelet threshold filtering algorithm is used to remove cutting background noise and electromagnetic interference signals; Extract the time-domain, frequency-domain, and time-frequency-domain features of each signal, and perform feature dimensionality reduction and data standardization through principal component analysis; Finally, the processed feature dataset is transmitted to the embedded intelligent decision unit in real time via industrial Ethernet.

[0031] The embedded intelligent decision-making unit has a built-in process parameter self-optimization model and error compensation algorithm, receives operating condition data from the multi-source fusion sensing module and outputs dynamic adjustment instructions; the embedded intelligent decision-making unit adopts an edge computing heterogeneous architecture and has a built-in process parameter self-optimization model and thermal coupling error prediction model based on deep reinforcement learning. The process parameter self-optimization model adopts an Actor-Critic dual network architecture, with real-time cutting force, vibration, and temperature characteristics as the state space, the adjustment amount of spindle speed, feed rate, depth of cut, and width of cut as the action space, and the weighted comprehensive value of material removal rate, machining stability, and tool wear rate as the reward function, outputting the optimal process parameter adjustment command; The thermo-coupling error prediction model takes multi-point temperature data and cutting force data at key locations as input, and predicts the thermal deformation error and stress deformation error of each axis of the machine tool through a trained convolutional neural network. The output of optimal process parameter adjustment instructions and error compensation instructions are then directly connected to the position loop feedforward channel of the CNC system.

[0032] The in-situ detection module is integrated inside the machining compartment of the machine tool bed and is used to perform non-contact dimensional and positional error detection on the workpiece during the machining cycle. The in-situ detection module includes a high-precision line laser probe and a binocular vision positioning subunit. The line laser probe is installed on the inner side wall of the processing chamber via a pneumatic telescopic protective mechanism and is stored in a sealed protective chamber when not in use. After processing, the workpiece does not need to be disassembled and clamped. The line laser probe moves along an automatically planned scanning path under the drive of the CNC system to acquire three-dimensional point cloud data of the entire surface of the workpiece. The binocular vision positioning subunit is used to identify workpiece clamping posture deviation before processing, automatically calculate the workpiece coordinate system offset and complete the correction; at the same time, it can perform imaging detection on the tool cutting edge to identify the amount of wear and chipping defects on the back face. The in-situ detection module also has a built-in standard gauge block in-situ calibration component, which automatically completes the probe accuracy calibration before each test.

[0033] The digital twin mapping module communicates bidirectionally with the multi-source fusion sensing module, the in-situ detection module, and the CNC system to construct a digital twin of the machine tool in real time and form a closed-loop feedback link; the digital twin mapping module includes a geometric model, a multi-physics field physical property model, and a cutting process model; The geometric model is constructed based on the three-dimensional CAD model of the machine tool, accurately mapping the dimensions, assembly relationships, and kinematic pair constraints of each moving part of the machine tool. The multiphysics model incorporates material thermal conductivity parameters, structural stiffness matrix, and damping characteristics to simulate the temperature field distribution and stress deformation distribution of the machine tool in real time. The cutting process model integrates an improved Johnson-Cook cutting constitutive model, which can predict cutting force, surface roughness, and tool wear rate. The digital twin operates synchronously with the physical machine tool in real time, predicting the error change trend during future processing periods. When the prediction deviation exceeds a set threshold, a pre-compensation command is sent to the embedded intelligent decision-making unit in advance. The measured data from in-situ detection is used to correct the twin model parameters in reverse, realizing continuous iteration of model accuracy.

[0034] The integrated auxiliary function unit is integrated inside the machine tool bed and includes an adaptive chip removal unit, a precision temperature control unit, and a tool condition monitoring unit, wherein: The adaptive chip removal unit includes a built-in spiral chip removal mechanism and a zoned high-pressure chip flushing subsystem, wherein: The partitioned high-pressure chip removal subsystem is equipped with nozzles in the processing area and chip removal channel respectively. The chip concentration at the chip removal port is detected in real time by an infrared photoelectric sensor. The chip removal pressure and nozzle opening and closing are automatically adjusted according to the cutting load and chip concentration to achieve chip removal on demand. The precision temperature control unit adopts a multi-loop partitioned independent liquid cooling architecture, which corresponds to three temperature control areas: the spindle box, the three-axis lead screw pair, and the bed base. Each loop is equipped with an independent PID control unit and a flow regulating valve, which suppresses thermal deformation from the source by homogenizing the overall temperature field of the machine tool.

[0035] The tool condition monitoring unit is integrated on the tool change path on the side of the tool magazine and includes a contact trigger probe and a side-light microscopic image acquisition component. During each tool change, the tool passes through the monitoring station in sequence: first, the contact trigger probe completes the high-precision calibration of the tool length and radius, and then the side-light microscopic image acquisition component captures the image of the tool cutting edge; The width of the wear band and the chipping size on the back face of the tool are calculated by image segmentation and edge extraction algorithms, and the tool wear level is automatically classified. Wear and dimensional data are synchronously transmitted to the digital twin mapping module and the embedded intelligent decision-making unit to correct tool compensation values ​​and adjust cutting process parameters; when tool breakage exceeds the safety threshold, an emergency tool change and machining pause command are immediately triggered.

[0036] It is worth mentioning that the CNC system is an open CNC system. After receiving the adjustment instructions from the embedded intelligent decision unit, it adopts a composite compensation strategy of look-ahead pre-interpolation and real-time error superposition. The thermal deformation compensation amount and the stress deformation compensation amount are superimposed on the position instructions of the corresponding feed axis respectively, and the motion trajectory of each axis is corrected in real time through high-speed interpolation algorithm. The CNC system supports offline process import and online process parameter iterative optimization, stores process data and inspection data of multiple batches of machining, and automatically updates the experience base of the process parameter optimization model; it also has a machining anomaly graded early warning function, which identifies chatter, tool wear, and chip accumulation anomalies based on multi-source sensing data and issues graded early warnings.

[0037] like Figure 2 As shown, the working process of the machine tool of the present invention is as follows: Step S1, Pre-processing datum calibration and twin model initialization: Step S1.1: After the workpiece is clamped, the binocular vision positioning subunit is activated to acquire the workpiece pose image, calculate the offset between the actual coordinate system and the theoretical coordinate system of the workpiece, and automatically correct the workpiece coordinate system in the CNC system. Step S1.2: After the tool is loaded into the spindle, the tool condition monitoring unit performs initial length and radius calibration and edge wear condition detection on the tool, and records the initial parameters of the tool; Step S1.3: Import all machine tool initial state parameters, workpiece parameters, tool parameters and preset process parameters into the digital twin mapping module to construct a reference digital twin that corresponds one-to-one with the physical machine tool and complete the initialization; Step S2: Real-time acquisition and preprocessing of multi-source data during the processing: Step S2.1: After the machining process is started, the multi-source fusion sensing module synchronously collects multiple working condition data such as cutting force, vibration, acoustic emission, temperature and visual images at the set sampling frequency; Step S2.2: The data preprocessing subunit performs timestamp alignment, noise filtering, feature extraction and dimensionality reduction standardization on the multi-channel heterogeneous data to generate a structured feature dataset, which is then transmitted to the embedded intelligent decision unit in real time. Step S3, Embedded intelligent decision-making and dynamic processing control: Step S3.1: The embedded intelligent decision-making unit inputs real-time feature data into the process parameter self-optimization model to determine the current cutting state: if the cutting force fluctuates too much, chattering occurs, or tool wear accelerates, the feed rate and depth of cut are reduced; if the cutting state is stable and the allowance is sufficient, the cutting parameters are gradually increased to improve the material removal rate. Meanwhile, the thermo-coupling error prediction model calculates the thermal deformation and stress deformation error of each feed axis based on real-time temperature and cutting force data, generates position compensation commands and sends them directly to the position loop feedforward channel of the CNC system to achieve real-time dynamic error compensation during the machining process. Step S3.2: The digital twin synchronously maps the physical machine tool's operating status and predicts the error changes in subsequent processing stages in advance. If the prediction deviation exceeds the threshold, the pre-compensation is triggered in advance. Step S4, in-situ inspection and accuracy data acquisition: Step S4.1: After a single processing step is completed or all processing steps are completed, the line laser probe extends out of the protective chamber and performs a full-size scan of the workpiece surface according to the automatically planned scanning path to obtain the workpiece's three-dimensional point cloud data. Step S4.2: Denoise, stitch and fit the point cloud data, calculate the measured values ​​of each dimension feature, form and position tolerance and surface roughness of the workpiece, and compare them with the design tolerance to generate an accuracy inspection report; Step S5, Digital Twin Model Correction and Process Iteration Optimization: Step S5.1: The digital twin mapping module compares the measured accuracy data obtained from the in-situ detection with the predicted data of the twin model and calculates the deviation between the two. Step S5.2: Correct the physical parameters such as thermal conductivity coefficient, structural stiffness coefficient, and cutting force coefficient in the twin model through parameter optimization algorithm to improve the prediction accuracy of subsequent processing; At the same time, the processing condition data, process parameters and test results are stored in the process experience database, and the process parameter self-optimization model is incrementally trained to realize the autonomous iterative evolution of processing capabilities. Repeat steps S1 to S5 to complete continuous self-optimization for multiple batches of processing.

[0038] It is worth mentioning that the technical features of the machine tool bed body, CNC system, spindle unit and feed transmission unit involved in this patent application should be regarded as prior art. The specific structure, working principle and possible control method and spatial arrangement of these technical features can be adopted by conventional choices in the field and should not be regarded as the inventive point of this patent. This patent will not be further elaborated in detail.

[0039] For those skilled in the art, modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this invention should be included within the protection scope of this invention.

Claims

1. An integrated CNC intelligent machine tool, comprising a machine tool bed, a CNC system, a spindle unit, and a feed transmission unit, characterized in that, It also includes a multi-source fusion sensing module, an embedded intelligent decision-making unit, an in-situ detection module, a digital twin mapping module, and an integrated auxiliary function unit, among which: The multi-source fusion sensing module is deployed in the spindle unit, feed transmission unit and machining area to collect multi-physics field condition data in real time during the machining process. The embedded intelligent decision-making unit has a built-in process parameter self-optimization model and error compensation algorithm, receives operating condition data from the multi-source fusion sensing module and outputs dynamic adjustment instructions; The in-situ detection module is integrated inside the machining compartment of the machine tool bed and is used to perform non-contact dimensional and positional error detection on the workpiece during the machining cycle. The digital twin mapping module communicates bidirectionally with the multi-source fusion sensing module, the in-situ detection module, and the CNC system to construct a digital twin of the machine tool in real time and form a closed-loop feedback link. The integrated auxiliary function unit is integrated inside the machine tool bed and includes an adaptive chip removal unit, a precise temperature control unit, and a tool status monitoring unit.

2. The integrated CNC intelligent machine tool according to claim 1, characterized in that, The multi-source fusion sensing module includes a triaxial piezoelectric cutting force sensor, a triaxial vibration acceleration sensor, a broadband acoustic emission sensor, a distributed multi-point temperature sensor, and a high-definition industrial vision acquisition unit, wherein: The triaxial piezoelectric cutting force sensor is embedded below the worktable surface of the workpiece clamping table and is used to collect the X / Y / Z three-axis cutting force and cutting torque in real time. The triaxial vibration acceleration sensor is installed on the radial and axial end faces of the front bearing housing of the spindle to collect the time-domain and frequency-domain vibration signals of the spindle during the cutting process; The broadband acoustic emission sensor is attached to the side wall of the machining chamber near the cutting area to capture high-frequency acoustic emission signals generated by tool breakage and workpiece micro-cracks. The distributed multi-point temperature sensors are deployed at multiple key locations; The high-definition industrial vision acquisition unit is installed on the side of the spindle box and is used to monitor the chip accumulation state in the cutting zone and the surface morphology of the workpiece in real time.

3. An integrated CNC intelligent machine tool according to claim 2, characterized in that, The multi-source fusion sensing module also includes a data preprocessing subunit, which is implemented based on an FPGA hardware acceleration architecture and executes the following sequentially: Based on the clock protocol, the timestamps of multiple data collected by the triaxial piezoelectric cutting force sensor, triaxial vibration acceleration sensor, broadband acoustic emission sensor, distributed multi-point temperature sensor and high-definition industrial vision acquisition unit are aligned. An adaptive wavelet threshold filtering algorithm is used to remove cutting background noise and electromagnetic interference signals; Extract the time-domain, frequency-domain, and time-frequency-domain features of each signal, and perform feature dimensionality reduction and data standardization through principal component analysis; Finally, the processed feature dataset is transmitted to the embedded intelligent decision unit in real time via industrial Ethernet.

4. An integrated CNC intelligent machine tool according to claim 1, characterized in that, The embedded intelligent decision-making unit adopts an edge computing heterogeneous architecture and incorporates a process parameter self-optimization model and a thermal coupling error prediction model based on deep reinforcement learning. The process parameter self-optimization model uses real-time cutting force, vibration, and temperature characteristics as the state space, the adjustment amount of spindle speed, feed rate, depth of cut, and width of cut as the action space, and the weighted comprehensive value of material removal rate, machining stability, and tool wear rate as the reward function to output the optimal process parameter adjustment command. The thermo-coupling error prediction model takes multi-point temperature data and cutting force data at key locations as input, and predicts the thermal deformation error and stress deformation error of each axis of the machine tool through a trained convolutional neural network. The output of optimal process parameter adjustment instructions and error compensation instructions are then directly connected to the position loop feedforward channel of the CNC system.

5. An integrated CNC intelligent machine tool according to claim 1, characterized in that, The in-situ detection module includes a high-precision line laser probe and a binocular vision positioning subunit; The line laser probe is installed on the inner side wall of the processing chamber via a pneumatic telescopic protective mechanism and is stored in a sealed protective chamber when not in use. After processing, the workpiece does not need to be disassembled and clamped. The line laser probe moves along an automatically planned scanning path under the drive of the CNC system to acquire three-dimensional point cloud data of the entire surface of the workpiece. The binocular vision positioning subunit is used to identify workpiece clamping posture deviation before processing, automatically calculate the workpiece coordinate system offset and complete the correction; at the same time, it can perform imaging detection on the tool cutting edge to identify the amount of wear and chipping defects on the back face. The in-situ detection module also has a built-in standard gauge block in-situ calibration component, which automatically completes the probe accuracy calibration before each test.

6. An integrated CNC intelligent machine tool according to claim 1, characterized in that, The digital twin mapping module includes a geometric model, a multiphysics physical property model, and a cutting process model; The geometric model is constructed based on the three-dimensional CAD model of the machine tool, accurately mapping the dimensions, assembly relationships, and kinematic pair constraints of each moving part of the machine tool. The multiphysics model incorporates material thermal conductivity parameters, structural stiffness matrix, and damping characteristics to simulate the temperature field distribution and stress deformation distribution of the machine tool in real time. The cutting process model predicts cutting force, surface roughness, and tool wear rate. The digital twin operates synchronously with the physical machine tool in real time, predicting the error change trend during future processing periods. When the prediction deviation exceeds a set threshold, a pre-compensation command is sent to the embedded intelligent decision-making unit in advance. The measured data from in-situ detection is used to correct the twin model parameters in reverse, realizing continuous iteration of model accuracy.

7. An integrated CNC intelligent machine tool according to claim 1, characterized in that, The adaptive chip removal unit includes a built-in spiral chip removal mechanism and a zoned high-pressure chip flushing subsystem, wherein: The partitioned high-pressure chip removal subsystem is equipped with nozzles in the processing area and chip removal channel respectively. The chip concentration at the chip removal port is detected in real time by an infrared photoelectric sensor. The chip removal pressure and nozzle opening and closing are automatically adjusted according to the cutting load and chip concentration to achieve chip removal on demand. The precision temperature control unit adopts a multi-loop partitioned independent liquid cooling architecture, which corresponds to three temperature control areas: the spindle box, the three-axis lead screw pair, and the bed base. Each loop is equipped with an independent PID control unit and a flow regulating valve, which suppresses thermal deformation from the source by homogenizing the overall temperature field of the machine tool.

8. An integrated CNC intelligent machine tool according to claim 7, characterized in that, The tool condition monitoring unit is integrated on the tool change path on the side of the tool magazine and includes a contact trigger probe and a side-light microscopic image acquisition component. During each tool change, the tool passes through the monitoring station in sequence: first, the contact trigger probe completes the high-precision calibration of the tool length and radius, and then the side-light microscopic image acquisition component captures the image of the tool cutting edge; The width of the wear band and the chipping size on the back face of the tool are calculated by image segmentation and edge extraction algorithms, and the tool wear level is automatically classified. Wear and dimensional data are synchronously transmitted to the digital twin mapping module and the embedded intelligent decision-making unit to correct tool compensation values ​​and adjust cutting process parameters; when tool breakage exceeds the safety threshold, an emergency tool change and machining pause command are immediately triggered.

9. An integrated CNC intelligent machine tool according to claim 1, characterized in that, The CNC system is an open CNC system. After receiving the adjustment instructions from the embedded intelligent decision unit, it adopts a composite compensation strategy of look-ahead pre-interpolation and real-time error superposition. The thermal deformation compensation amount and the stress deformation compensation amount are superimposed on the position instructions of the corresponding feed axes respectively. The motion trajectory of each axis is corrected in real time through a high-speed interpolation algorithm. The CNC system supports offline process import and online process parameter iterative optimization, stores process data and inspection data of multiple batches of machining, and automatically updates the experience base of the process parameter optimization model; it also has a machining anomaly graded early warning function, which identifies chatter, tool wear, and chip accumulation anomalies based on multi-source sensing data and issues graded early warnings.

10. An integrated CNC intelligent machine tool according to claim 1, characterized in that, The working process of the machine tool is as follows: Step S1, Pre-processing datum calibration and twin model initialization: Step S1.1: After the workpiece is clamped, the binocular vision positioning subunit is activated to acquire the workpiece pose image, calculate the offset between the actual coordinate system and the theoretical coordinate system of the workpiece, and automatically correct the workpiece coordinate system in the CNC system. Step S1.2: After the tool is loaded into the spindle, the tool condition monitoring unit performs initial length and radius calibration and edge wear condition detection on the tool, and records the initial parameters of the tool; Step S1.3: Import all machine tool initial state parameters, workpiece parameters, tool parameters and preset process parameters into the digital twin mapping module to construct a reference digital twin that corresponds one-to-one with the physical machine tool and complete the initialization; Step S2: Real-time acquisition and preprocessing of multi-source data during the processing: Step S2.1: After the machining process is started, the multi-source fusion sensing module synchronously collects multiple working condition data such as cutting force, vibration, acoustic emission, temperature and visual images at the set sampling frequency; Step S2.2: The data preprocessing subunit performs timestamp alignment, noise filtering, feature extraction and dimensionality reduction standardization on the multi-channel heterogeneous data to generate a structured feature dataset, which is then transmitted to the embedded intelligent decision unit in real time. Step S3, Embedded intelligent decision-making and dynamic processing control: Step S3.1: The embedded intelligent decision-making unit inputs real-time feature data into the process parameter self-optimization model to determine the current cutting state: if the cutting force fluctuates too much, chattering occurs, or tool wear accelerates, the feed rate and depth of cut are reduced; if the cutting state is stable and the allowance is sufficient, the cutting parameters are gradually increased to improve the material removal rate. Meanwhile, the thermo-coupling error prediction model calculates the thermal deformation and stress deformation error of each feed axis based on real-time temperature and cutting force data, generates position compensation commands and sends them directly to the position loop feedforward channel of the CNC system to achieve real-time dynamic error compensation during the machining process. Step S3.2: The digital twin synchronously maps the physical machine tool's operating status and predicts the error changes in subsequent processing stages in advance. If the prediction deviation exceeds the threshold, the pre-compensation is triggered in advance. Step S4, in-situ inspection and accuracy data acquisition: Step S4.1: After a single processing step is completed or all processing steps are completed, the line laser probe extends out of the protective chamber and performs a full-size scan of the workpiece surface according to the automatically planned scanning path to obtain the workpiece's three-dimensional point cloud data. Step S4.2: Denoise, stitch and fit the point cloud data, calculate the measured values ​​of each dimension feature, form and position tolerance and surface roughness of the workpiece, and compare them with the design tolerance to generate an accuracy inspection report; Step S5, Digital Twin Model Correction and Process Iteration Optimization: Step S5.1: The digital twin mapping module compares the measured accuracy data obtained from the in-situ detection with the predicted data of the twin model and calculates the deviation between the two. Step S5.2: Correct the physical parameters such as thermal conductivity coefficient, structural stiffness coefficient, and cutting force coefficient in the twin model through parameter optimization algorithm to improve the prediction accuracy of subsequent processing; At the same time, the processing condition data, process parameters and test results are stored in the process experience database, and the process parameter self-optimization model is incrementally trained to realize the autonomous iterative evolution of processing capabilities. Repeat steps S1 to S5 to complete continuous self-optimization for multiple batches of processing.

Images