Cutting temperature online measurement and temperature field nanoscale resolution analysis method

Abstract

The invention discloses a cutting area temperature on-line measurement method, which comprises the following steps of: S11, manufacturing a cutter by using a high-hardness material with high transmittance in a measurement laser wave band, and fixing a high-precision infrared thermometer on the rear side of the cutter through a precision positioning mechanism, the position and angle of the measuring laser beam are adjusted to ensure that the measuring laser beam axis can pass through the transparent cutter and exit from one, two or all of the rake face, the flank face and the cutting edge of the cutter. And S12, secondly, in the cutting process, the infrared thermometer receives infrared radiation signals which come from the machining area and are transmitted back after penetrating through the tool body. And S13, finally, converting the obtained infrared radiation data into a temperature value through signal analysis inversion processing, and forming a direct temperature measurement image sequence and key point temperature time sequence data for subsequent temperature field reconstruction.

Description

Technical Field

[0001] This invention belongs to the field of precision / ultra-precision machining and intelligent manufacturing technology, specifically relating to a method for online measurement of temperature in the cutting zone and nanometer-level resolution analysis of the temperature field. Background Technology

[0002] During machining, the temperature of the machining zone has a decisive impact on tool life, workpiece surface quality, and machining accuracy. Excessively high cutting temperatures can lead to rapid tool wear, deterioration of machining quality, and even defects such as cracks; conversely, excessively low cutting temperatures may increase cutting resistance and reduce machining efficiency. Therefore, to accurately control the temperature of the cutting zone, guide process optimization, and improve machining quality and efficiency, there is an urgent need for a measurement method that can accurately detect the temperature of the cutting zone in real time.

[0003] Currently, the measurement of cutting temperature mainly relies on thermocouple methods and infrared thermometry. Thermocouple methods are contact measurements, resulting in slow response times and difficulty in precisely positioning the thermocouples in the tool-chip contact area, leading to significant measurement errors. While infrared thermometry is non-contact and offers fast response, it faces significant challenges in practical applications: splattering chips, covering cutting fluid, and the tool itself severely obstruct and interfere with the measurement of infrared radiation in the machining area. Laser infrared thermometers typically have a spot size on the micrometer scale, only able to acquire the average temperature within the spot area, failing to capture the detailed temperature distribution at smaller scales, thus making it impossible to directly obtain the true temperature of the machining area. Existing technologies usually only measure the surface temperature of the workpiece or tool, then indirectly calculate the temperature of the contact area using theoretical models. This method suffers from low accuracy and poor reliability, failing to meet the demands of real-time, precise temperature control in high-precision machining.

[0004] While existing technologies have attempted to predict temperature fields using digital twin models, finite element simulations, or neural networks, none have fundamentally solved the problem of direct and accurate temperature detection in the cutting region. This results in model predictions failing to reflect crucial information such as local temperature gradients and the distribution of minute heat sources at the nanoscale, making it difficult to meet the demands of high-precision machining. Therefore, there is an urgent need for a technical solution that can overcome obstruction interference and acquire the temperature of the core cutting region online during the cutting process. Summary of the Invention

[0005] The purpose of this invention is to solve the above-mentioned problems and provide a technical method that integrates high-precision infrared temperature measurement, finite element simulation and deep learning technology to realize online measurement of the temperature of the workpiece cutting area during the cutting process and nanometer-level resolution analysis of the temperature field of the cutting area.

[0006] To solve the above-mentioned technical problems, the technical solution of the present invention is: an online temperature measurement method for the cutting zone, comprising the following steps:

[0007] S11. A high-hardness material with high transmittance in the measurement laser band is used to make the cutting tool, and a high-precision infrared thermometer is fixed to the back of the cutting tool through a precision positioning mechanism. The position and angle are adjusted to ensure that the axis of the measurement laser beam can pass through the transparent cutting tool and exit from one, two, or all of the cutting face, back face, and cutting edge of the cutting tool.

[0008] S12. Secondly, during the cutting process, the infrared thermometer receives infrared radiation signals from the machining area and after penetrating the tool body.

[0009] S13. Finally, through signal analysis and inversion processing, the acquired infrared radiation data is converted into temperature values, forming a direct temperature measurement image sequence and key point temperature time series data for subsequent temperature field reconstruction.

[0010] Furthermore, the high-hardness transparent material in S11 includes diamond, sapphire, ruby, and glass.

[0011] Furthermore, the target area in S11 is the cutting processing area, including the tool-chip contact area and the tool-workpiece contact area.

[0012] Furthermore, the signal analysis and inversion in S13 specifically includes: the infrared thermometer receives the infrared radiation signal from the processing area and after penetrating the tool body, and by performing photoelectric conversion, noise filtering, intensity calibration and temperature inversion calculation on the radiation signal, the original radiation signal is analyzed into an accurate temperature value.

[0013] Furthermore, the signal analysis and inversion process in S13 is as follows:

[0014] S121, Photoelectric conversion: Converting infrared radiation signals into electrical signals;

[0015] S122. Noise filtering: Eliminate random interference in the signal;

[0016] S123, Intensity Calibration: Based on the standard blackbody source, the correspondence between radiation intensity and temperature is calibrated;

[0017] S124. Temperature inversion calculation: Combining tool material transmittance correction and atmospheric transmission attenuation compensation, the true temperature is inverted by substituting into Planck's formula.

[0018] ;

[0019] in: Let λ be the blackbody monochromatic radiant exitance at wavelength λ (W / (m²·μm)), and h be Planck's constant (6.626 × 10⁻³). 4 J・s), c is the speed of light (3×10⁻⁶ s), and c is the speed of light (3×10⁻⁶ s). 8m / s), λ is the measurement wavelength (μm), k is the Boltzmann constant (1.38×10⁻²³J / K), and T is the absolute temperature (K).

[0020] This invention also discloses a nanometer-resolution analytical method for the cutting temperature field, comprising the following steps:

[0021] S21. First, determine the key machining parameters, such as workpiece material, tool geometry parameters, and cutting parameters. Refine the mesh to the nanometer level. Based on these parameters, establish a cutting machining thermo-mechanical coupled finite element model, perform simulation calculations under multiple sets of parameters, and generate an initial simulation temperature field database.

[0022] S22. Secondly, a hybrid training database is constructed by pairing the average temperature data measured by micron-level light spots with the precise temperature field data simulated at the nanometer level to form a hybrid dataset for training neural networks.

[0023] S23. Then, network models such as encoder-decoder are used for training. The training process is based on a hybrid training database to strengthen the correlation between micron-level temperature data and nano-level simulated temperature. This enables the model to accurately separate and reconstruct the temperature field of the nanoscale processing area from the macro-average data. During the training process, the full-field temperature field distribution under the corresponding conditions is used as the training objective. By minimizing the mean square error or mean absolute error between the predicted result and the true value, the global optimization algorithm is used to iteratively update the network parameters until the model converges.

[0024] S24. Finally, in actual cutting, the real-time acquired local temperature data is input into the trained model, and the corrected and reconstructed complete high-precision temperature field distribution is output in real time.

[0025] Further, in S23, the encoder-decoder network model includes: the encoder part is composed of multiple convolutional layers and pooling layers stacked sequentially, extracting multi-level features of the infrared temperature image through convolution operations, wherein the function of the convolutional layer is: to extract multi-level features of the infrared temperature image through convolution kernels, as shown in the formula:

[0026] ;

[0027] in: Let k be the feature value at position (i,j) in the l-th layer, and k be the kernel size. The weights of the l-th layer convolutional kernel are... For bias terms, This is the activation function.

[0028] The beneficial effects of this invention are:

[0029] 1. The present invention provides an online temperature measurement method for the cutting zone, enabling direct and accurate measurement of the temperature in the core machining area. Thanks to the use of transparent cutting tools with high infrared transmittance (such as single-crystal diamond or optical sapphire), the infrared thermometric beam can directly penetrate the tool body and reach the tool-chip contact area and tool-workpiece contact area without interference, obtaining true in-situ temperature data and distribution images. This method effectively avoids interference from chips and cutting fluid, and provides high-precision local temperature data through signal analysis and inversion, laying a solid foundation for subsequent temperature field reconstruction.

[0030] 2. The present invention provides an online temperature measurement method for the cutting zone, constructing an integrated closed-loop intelligent machining system encompassing measurement, reconstruction, and control. This system integrates a high-precision infrared temperature measurement unit, a finite element simulation calculation unit, a hybrid training database construction unit, an encoder-decoder network module, and a temperature field reconstruction output interface. Based on the reconstructed real-time, accurate full-field temperature distribution, it dynamically and intelligently adjusts feed rate, cutting speed, or coolant parameters. This achieves precise closed-loop control and proactive optimization of machining temperature, shifting from passive monitoring to proactive intelligent control, thereby improving machining efficiency and quality.

[0031] 3. The present invention provides a nanometer-resolution analytical method for cutting temperature fields, achieving high-precision and high-resolution reconstruction from local temperature measurement data to the overall temperature distribution. By fusing finite element simulation data and real temperature measurement data, and utilizing a specially designed encoder-decoder neural network model, a precise mapping from local discrete measurements to a complete and continuous temperature field is achieved. This method not only improves the edge accuracy and spatial consistency of temperature field reconstruction, but also extends to tool temperature field prediction, overcoming the problem of insufficient prediction accuracy caused by the reliance on assumptions in pure simulation models.

[0032] 4. This invention breaks through the limitations of traditional thermometer spot size by fusing nanoscale refined finite element simulation with real temperature measurement data and combining it with an optimized encoder-decoder neural network. It achieves cross-scale mapping "from micrometer-level average temperature to nanometer-level precise temperature field", providing quantifiable and high-resolution temperature field data for nanoscale ultra-precision cutting. Attached Figure Description

[0033] Figure 1 This is a schematic diagram showing the arrangement of the transparent cutting tool and the infrared thermometer of the present invention;

[0034] Figure 2 This is a 3D schematic diagram of the arrangement of the transparent cutting tool and the infrared thermometer of the present invention;

[0035] Figure 3 This is a schematic diagram of the finite element simulation of the present invention;

[0036] Figure 4This is a schematic diagram of the neural network structure of the present invention.

[0037] Explanation of reference numerals in the attached diagram: 1. Infrared thermometer; 2. Infrared focusing lens; 3. Transparent cutting tool; 4. Workpiece; 5. Temperature measurement area; 6. Chips. Detailed Implementation

[0038] The present invention will be further described below with reference to the accompanying drawings and specific embodiments:

[0039] like Figures 1 to 4 As shown, the present invention provides an online temperature measurement method for the cutting zone, comprising the following steps:

[0040] S11. A high-hardness material with high transmittance in the measurement laser band is used to make the cutting tool, and a high-precision infrared thermometer is fixed to the back of the cutting tool through a precision positioning mechanism. Its position and angle are adjusted to ensure that the axis of the measurement laser beam can pass through the transparent cutting tool and exit from one, two, or all of the cutting edge, such as the front cutting edge, the back cutting edge, or the cutting edge.

[0041] The high-hardness transparent materials in step S11 include diamond and sapphire. They have an infrared transmittance of ≥85% and are also high-hardness transparent materials.

[0042] The target area in step S11 is the cutting area, which includes the tool-chip contact area and the tool-workpiece contact area.

[0043] In this embodiment, the mid-infrared thermometer has a measurement resolution of 100µm, a temperature measurement range of 20-600 degrees Celsius, and a response time ≤10ms. The cutting tool is made of single-crystal diamond or optical sapphire with an infrared wavelength of 2-15µm and a transmittance ≥85%. First, an infrared temperature measurement system is constructed. This embodiment uses a transparent cutting tool made of single-crystal diamond, which has high infrared transmittance and high hardness, suitable for cutting polycrystalline copper workpieces. The high-precision infrared thermometer is positioned directly above the back face of the transparent cutting tool, ensuring that the infrared beam axis is perpendicular to the upper surface of the tool and penetrates the tool body to directly irradiate the core heat-generating areas such as the tool-chip contact area and the tool-workpiece contact area, avoiding measurement blind spots caused by beam deviation.

[0044] S12. Secondly, during the cutting process, the infrared thermometer receives infrared radiation signals from the machining area that have penetrated the tool body and are transmitted back.

[0045] During the cutting process, the splashed chips and cutting fluid only cover the workpiece surface in front of the tool or the outside of the tool, while the infrared light path is located inside the tool body, completely isolating it from external interference. The infrared thermometer receives infrared radiation signals directly from the machining area and penetrating the transparent tool in real time. Through a built-in algorithm, the radiation signals are converted into temperature values, forming a high-resolution local temperature image sequence, as well as time-series temperature data for five key points: the center of the tool-chip contact area and the edge of the tool-workpiece contact area. The experimental data obtained in this step accurately reflects the heat source distribution in the machining area, without signal attenuation or distortion caused by chips or cutting fluid, providing high-quality foundational data for subsequent data fusion.

[0046] In this embodiment, the workpiece is polycrystalline copper. Based on the processing requirements of the polycrystalline copper workpiece, the key parameter combination is determined as follows: material properties: polycrystalline copper; shape and size: workpiece width (cutting direction) 12μm, workpiece height 20μm, workpiece length 104μm; cutting parameters: depth of cut 10μm, cutting width 10μm, feed rate 1m / s.

[0047] S13. Finally, through signal analysis and inversion processing, the acquired infrared radiation data is converted into temperature values, forming a direct temperature measurement image sequence and key point temperature time series data for subsequent temperature field reconstruction.

[0048] The signal analysis and inversion in step S13 specifically includes: the infrared thermometer receives the infrared radiation signal from the processing area and after penetrating the tool body, and performs photoelectric conversion, noise filtering, intensity calibration and temperature inversion calculation on the radiation signal to analyze the original radiation signal into an accurate temperature value.

[0049] The signal analytical inversion process in step S13 is as follows:

[0050] S121, Photoelectric Conversion: Converts infrared radiation signals into electrical signals.

[0051] S122. Noise filtering: Removes random interference from the signal.

[0052] S123. Intensity calibration: Based on the standard blackbody source, calibrate the correspondence between radiation intensity and temperature.

[0053] S124. Temperature inversion calculation: Combining tool material transmittance correction and atmospheric transmission attenuation compensation, the true temperature is inverted by substituting into Planck's formula.

[0054] ;

[0055] in: Let λ be the blackbody monochromatic radiant exitance at wavelength λ (W / (m²·μm)), and h be Planck's constant (6.626 × 10⁻³). 4 J・s), c is the speed of light (3×10⁻⁶ s), and c is the speed of light (3×10⁻⁶ s).8 m / s), λ is the measurement wavelength (μm), k is the Boltzmann constant (1.38×10⁻²³J / K), and T is the absolute temperature (K).

[0056] The online temperature measurement method for the cutting zone provided by this invention provides high-precision local temperature data through direct measurement and accurate signal analysis, laying the foundation for temperature field reconstruction based on deep learning.

[0057] This invention also provides a method for resolving cutting temperature fields at nanometer-level resolution, comprising the following steps:

[0058] S21. First, determine the key machining parameters, such as workpiece material, tool geometry parameters, and cutting parameters. Refine the mesh to the nanometer level. Based on these parameters, establish a cutting machining thermo-mechanical coupled finite element model, perform simulation calculations under multiple sets of parameters, and generate an initial simulation temperature field database.

[0059] S22. Secondly, a hybrid training database is constructed by pairing the average temperature data measured by micron-level light spots with the precise temperature field data simulated at the nanometer level to form a hybrid dataset for training neural networks.

[0060] S23. Then, an encoder-decoder network model is used for training. The training process is based on a hybrid training database, strengthening the correlation between micron-level temperature data and nanoscale simulated temperature. This allows the model to accurately separate and reconstruct the temperature field of the nanoscale processing area from the macroscopic average data. During training, the overall temperature field distribution under corresponding conditions is used as the training objective. By minimizing the mean square error or mean absolute error between the predicted result and the true value, a global optimization algorithm is used to iteratively update the network parameters until the model converges. The global optimization algorithm mentioned in this embodiment includes the Adam optimization algorithm.

[0061] In step S23, the encoder-decoder network model includes: the encoder part consists of multiple convolutional layers and pooling layers stacked sequentially, extracting multi-level features of the infrared temperature image through convolution operations. The function of the convolutional layer is to extract multi-level features of the infrared temperature image through convolution kernels, as shown in the formula:

[0062] ;

[0063] in: Let k be the feature value at position (i,j) in the l-th layer, and k be the kernel size. The weights of the l-th layer convolutional kernel are... For bias terms, This is the activation function.

[0064] The decoder consists of multiple transposed convolutional layers or combinations of upsampling and convolutional layers. It progressively restores the feature map size through upsampling operations, ultimately outputting a full-field temperature distribution prediction with the same size as the input image. Feature fusion between the encoder and decoder is achieved through a skip connection structure. This involves concatenating the feature maps from each layer of the encoder with the corresponding layers of the decoder, fusing low-level details with high-level semantic information, such as temperature gradient edges and heat source distribution patterns. This significantly improves the edge accuracy and spatial consistency of the temperature field reconstruction.

[0065] S24. Finally, in actual cutting, the real-time acquired local temperature data is input into the trained model, and the corrected and reconstructed complete high-precision temperature field distribution is output in real time.

[0066] The method of this invention, through a specially designed neural network model, not only achieves temperature field measurement with ultra-high resolution, but also extends to tool temperature field prediction, providing comprehensive data support for machining control.

[0067] This invention provides a deep learning-based method for reconstructing cutting temperature fields. Based on deep learning technology, it achieves accurate reconstruction of the high-resolution temperature field across the entire cutting process domain from local infrared temperature measurement data through finite element simulation and experimental data fusion, and predicts the tool temperature field in contact with the workpiece. This invention specifically selects and designs a convolutional neural network (CNN) with an encoder-decoder structure as the core training model. The weight-sharing characteristic of CNN reduces the number of network parameters and lowers computational complexity. Max pooling is used in the pooling layers to enhance feature translation invariance. This model can effectively process the spatial features of infrared temperature images, achieving end-to-end mapping from local measurement to the global temperature field. The feature map size is then progressively compressed through pooling layers to achieve feature abstraction and dimensionality reduction. The decoder consists of multiple transposed convolutional layers or combinations of upsampling layers and convolutional layers. Upsampling operations progressively restore the feature map size, ultimately outputting a predicted global temperature field distribution consistent with the input image size.

[0068] In this embodiment, the implementation process of the deep learning-based cutting temperature field reconstruction method of the present invention establishes a thermo-mechanical coupled finite element model of the cutting process based on the key parameters of the polycrystalline copper workpiece. Figure 2 This is a schematic diagram of a finite element simulation. The workpiece material properties are defined in the model as follows: density 8960 kg / m³, specific heat capacity 385 J / (kg·K), thermal conductivity 401 W / (m·K); the tool is set as an analytical rigid body; thermal boundary conditions are set as follows: initial workpiece temperature 25℃, convective heat transfer coefficient between the machining zone and air 20 W / (m²·K), and heat transfer coefficient between the cutting fluid and the workpiece 1000 W / (m²·K); force boundary conditions are set as follows: the tool moves along the cutting direction at a feed rate of 1 m / s, constraining the axial and radial displacement of the workpiece.

[0069] Simulations were performed under multiple sets of parameters to simulate the temperature field distribution in the processing area under different working conditions. For polycrystalline copper workpieces, the temperature field data at each time point was obtained through simulation, including the temperature field variation curve of the workpiece surface over time and the temperature distribution cloud map of the entire field. All simulation data were stored according to parameters to form an initial simulation temperature field database.

[0070] The actual temperature data of polycrystalline copper workpieces obtained experimentally in the "Online Temperature Measurement Method for Workpieces in the Cutting Processing Area," such as local temperature image sequences and key point time-series data, are precisely paired and fused with the simulated temperature field data under the corresponding parameters. By ensuring timestamp alignment error ≤1ms, spatial coordinate matching sampling point spacing ≤50μm, and eliminating invalid data with a rejection rate ≤5% (e.g., accidental signal fluctuations in the experiment or data with abnormal boundary conditions in the simulation), a hybrid training database is finally constructed. This database contains sample pairs of "local infrared temperature measurement data input - full-field temperature field distribution output," which are used to train deep learning models.

[0071] This embodiment uses a convolutional neural network (CNN) with an encoder-decoder structure, the structure of which is as follows: Figure 3 As shown, the encoder consists of three convolutional layers and three max-pooling layers, used to extract features from the infrared temperature image and achieve dimensionality reduction, gradually abstracting high-level features of the temperature distribution in the processing area; the decoder consists of three transposed convolutional layers and three convolutional layers, used to map and restore the low-dimensional features output by the encoder to the full-field temperature field distribution; the encoder and decoder are connected by skip connections, which fuse the detailed features of each layer of the encoder with the semantic features of the corresponding layer of the decoder, improving the edge accuracy and spatial consistency of the temperature field reconstruction.

[0072] During training, a hybrid database was used as samples, local infrared thermometry data was used as input, and the overall temperature field distribution was used as the training objective. Mean squared error was used as the loss function, and the network parameters were iteratively updated using the Adam optimizer. After training iterations, the model loss value converged to below 0.005. At this point, the model could reconstruct the complete temperature field from the local thermometry data with high accuracy, and the average absolute error between the reconstructed temperature and the experimentally measured temperature was ≤5℃.

[0073] In the actual machining of polycrystalline copper workpieces, local temperature data collected in real time by an infrared thermometer is input into a trained neural network model. The model completes the calculation within 30ms, outputting the reconstructed high-precision temperature field distribution of the complete machining area, and transmits it to the machining parameter control system in real time via a data interface. In this embodiment, the machining parameter control system refers to the part of the temperature measurement system with intelligent control functions. It can be implemented based on an industrial control computer or a dedicated control module combined with a control algorithm. It is a component of the "multi-unit collaboration" in the technical solution of this invention and is used to complete the closed-loop control function of cutting temperature. It is not a direct application of existing equipment.

[0074] like Figure 1As shown, in this embodiment, the temperature measurement system is an online cutting temperature measurement system, including a high-precision infrared temperature measurement unit, a finite element simulation calculation unit, a hybrid training database construction unit, an encoder-decoder network module, and a temperature field reconstruction output interface. Each functional unit and module of this invention uses an industrial control computer as an integrated carrier, and through hardware connection and software module collaboration, realizes the entire process of online cutting temperature measurement. Specifically, the architecture is as follows: The high-precision infrared temperature measurement unit includes an infrared thermometer, a focusing lens, a transparent cutting tool, a workpiece, and a precision positioning mechanism. The signals generated are transmitted to the industrial control computer via data cables. The finite element simulation calculation unit is a simulation software module running on the industrial control computer, such as ANSYS or ABAQUS, which performs temperature field simulation based on cutting parameters. The hybrid training database construction unit is built using database software such as MySQL or MongoDB on the industrial control computer, used to integrate infrared temperature measurement experimental data and finite element simulation data. The encoder-decoder network module is a deep learning algorithm module running on the industrial control computer, developed based on the TensorFlow and PyTorch frameworks, responsible for training and inference on the data in the database. The temperature field reconstruction output interface is a human-machine interface for industrial control computers, such as a software interface and a data export interface, which can output the reconstructed temperature field in the form of images, values, etc. The high-precision infrared temperature measurement unit, comprising the various functional units and modules of this invention, uses an industrial control computer as an integrated carrier. Through hardware connection and software module collaboration, it realizes the entire process of online temperature measurement in cutting. The precision positioning mechanism included in the high-precision infrared temperature measurement unit adopts the existing combination structure of "electric displacement stage + angle adjustment component" to achieve high-precision adjustment of the infrared thermometer's spatial position and beam angle: the electric displacement stage includes electric slides in the X, Y, and Z directions, driven by stepper motors or servo motors, achieving micron-level position adjustment to ensure accurate positioning of the infrared thermometer in the horizontal and vertical directions. The angle adjustment component is an electric rotary table installed on the top of the Z-axis slide, used to adjust the beam angle of the infrared thermometer, ensuring that the infrared beam axis is perpendicular to the upper surface of the tool. The control module of its control system is integrated into the industrial control computer, which can achieve automated positioning adjustment through programming or manual fine-tuning through the human-machine interface. Figure 1 The arrows in the diagram represent the transmission of the measurement signal and the return of the thermal radiation signal from the cutting area, respectively.

[0075] The system comprises an infrared thermometer 1, an infrared focusing lens 2, a transparent cutting tool 3, and a workpiece 4, arranged sequentially from top to bottom. The end of the transparent cutting tool 3 contacts the workpiece 4, which has a temperature measuring area 5. When the transparent cutting tool 3 contacts the workpiece 4 and performs cutting, it generates chips 6.

[0076] A precision positioning mechanism positions and fixes the infrared thermometer 1 directly above the flank face of the transparent cutting tool 3. By precisely adjusting its position and angle, it ensures that the infrared beam axis is perpendicular to the upper surface of the tool and penetrates the tool body, illuminating the tool-chip contact area. During cutting, the infrared thermometer 1 receives infrared radiation signals from the machining area that have penetrated the tool body. Through a built-in signal analysis and inversion algorithm, it obtains direct temperature measurement images and key point temperature time-series data of the machining area. The finite element simulation calculation unit is used to establish a thermo-mechanical coupled finite element model of the cutting process based on key machining parameters such as workpiece material, tool geometry, and cutting parameters. It simulates and calculates the temperature field distribution of the machining area, generating an initial simulated temperature field database. The hybrid training database construction unit is connected to both the finite element simulation calculation unit and the high-precision infrared thermometer unit. It is used to accurately pair and fuse the simulated temperature field database with real temperature data, forming a hybrid training database for training the neural network. The encoder-decoder network module uses a specially designed encoder-decoder CNN model as its core and is trained based on the hybrid training database to achieve intelligent mapping from local temperature measurement data to the overall temperature distribution. The temperature field reconstruction output interface is connected to the output of the encoder-decoder network module, used to transmit the reconstructed, complete, high-precision temperature field distribution to the machining parameter control system of the cutting equipment in real time. The machining parameter control system compares the received reconstructed temperature field with a preset ideal machining temperature range. If the temperature deviates from the ideal range, it adaptively adjusts the feed rate, cutting speed, or coolant parameters in real time to stabilize the machining temperature within the optimal range. Through the collaborative work of multiple units, the system forms a complete technical closed loop of "measurement-reconstruction-control," realizing online monitoring, high-resolution reconstruction, and intelligent control of the cutting temperature, demonstrating the progressive and organic integration between the two fundamental methods and the final system.

[0077] The cutting online temperature measurement system of the present invention integrates hardware and software components to implement the above-mentioned online temperature measurement method and temperature field reconstruction method, and outputs temperature field data in real time to control machining parameters.

[0078] The online temperature measurement system for cutting in this invention is implemented based on the two core methods described above. The system's core components are adapted to the processing scenario of polycrystalline copper workpieces as follows:

[0079] This unit includes a high-precision infrared thermometer, a three-axis precision adjustment mechanism, and a laser calibrator. The infrared thermometer is fixed above the machine tool column via the adjustment mechanism, and the laser calibrator is used to calibrate the optical path to ensure that the infrared beam penetrates the transparent tool vertically and accurately covers the tool-chip and tool-workpiece contact areas. During the polycrystalline copper cutting process, the infrared radiation signal penetrating the tool is acquired in real time, and local temperature images and key point timing data are output.

[0080] An industrial-grade server is used, and finite element simulation software is installed. A thermo-mechanical coupling model is established based on the parameters of the polycrystalline copper workpiece to complete the simulation calculation of multiple working conditions and generate an initial simulation temperature field database. At the same time, data processing software is provided to realize the format conversion of simulation data and support the construction of a hybrid database.

[0081] A network storage server is used to store the polycrystalline copper experimental data and finite element simulation data acquired by the infrared temperature measurement unit according to the parameters. The built-in data fusion algorithm completes the timestamp alignment, spatial coordinate matching and invalid data removal, automatically generates a hybrid training database, and supports quick query and call according to cutting parameters for neural network training.

[0082] Deployed in an edge computing unit, it loads a pre-trained CNN model based on the PyTorch framework, supports real-time reception of infrared temperature measurement data and output of reconstructed temperature field; it is also equipped with model optimization tools to improve model inference speed and ensure that the real-time requirements of polycrystalline copper workpiece cutting are met.

[0083] The temperature field reconstruction output interface adopts the OPC UA industrial protocol to establish a connection with the CNC machining parameter control system, transmitting the reconstructed full-field temperature field data in real time. The control system compares the received temperature data with the ideal temperature range of 650-750℃ for polycrystalline copper machining. If the temperature is higher than 750℃, the feed rate is automatically reduced and the cutting fluid flow is increased; if the temperature is lower than 650℃, the feed rate is appropriately increased. Through the "measurement-reconstruction-control" closed loop, the machining temperature of the polycrystalline copper workpiece is stably optimized, ultimately improving tool life and reducing workpiece surface roughness.

[0084] This invention integrates high-precision infrared thermometry, finite element simulation, and deep learning technologies to achieve direct measurement of workpiece temperature during machining, high-resolution reconstruction of the global temperature field, and real-time adaptive control of machining parameters. Specifically, this invention includes an online temperature measurement method for the workpiece temperature in the machining area. This method allows the temperature sensor signal to penetrate the transparent tool body and directly illuminate the tool-chip contact area. Temperature data and distribution images, unaffected by chips and cutting fluid, are obtained through signal analysis and inversion. A deep learning-based cutting temperature field reconstruction method is also included. This method constructs a mapping dataset through finite element calculations and utilizes a specially designed encoder-decoder neural network model to accurately reconstruct the global high-resolution temperature field from local temperature measurement data. The temperature measurement system of this invention integrates a high-precision infrared thermometry unit, a finite element simulation calculation unit, a hybrid training database construction unit, an encoder-decoder network module, and a temperature field reconstruction output interface. This system is used for real-time monitoring and reconstruction of the temperature field, which is then fed back to the machining parameter control system. These three parts form a tight technical chain: the online temperature measurement method provides high-quality input data to the system, the temperature field reconstruction method constitutes the intelligent processing core of the system, and the system enables engineering applications and closed-loop control. This invention overcomes the limitations of traditional temperature measurement methods due to interference from chips and cutting fluid, and achieves high-precision, high-resolution online monitoring and control of the temperature in the machining area.

[0085] Those skilled in the art will recognize that the embodiments described herein are intended to help the reader understand the principles of the invention, and should be understood that the scope of protection of the invention is not limited to such specific statements and embodiments. Those skilled in the art can make various other specific modifications and combinations based on the technical teachings disclosed in this invention without departing from the spirit of the invention, and these modifications and combinations are still within the scope of protection of this invention.

Claims

1. A method for online measurement of temperature in the cutting zone, characterized in that, Includes the following steps: S11. A high-hardness material with high transmittance in the measurement laser band is used to make the cutting tool, and a high-precision infrared thermometer is fixed to the back of the cutting tool through a precision positioning mechanism. The position and angle are adjusted to ensure that the axis of the measurement laser beam can pass through the transparent cutting tool and exit from one, two, or all of the cutting face, back face, and cutting edge of the cutting tool. S12. Secondly, during the cutting process, the infrared thermometer receives infrared radiation signals from the machining area and after penetrating the tool body. S13. Finally, through signal analysis and inversion processing, the acquired infrared radiation data is converted into temperature values, forming a direct temperature measurement image sequence and key point temperature time series data for subsequent temperature field reconstruction.

2. The method for online measurement of cutting zone temperature according to claim 1, characterized in that: The high-hardness transparent material in S11 includes diamond, sapphire, ruby, and glass.

3. The method for online measurement of cutting zone temperature according to claim 1, characterized in that: The target area in S11 is the cutting processing area, which includes the tool-chip contact area and the tool-workpiece contact area.

4. The method for online measurement of cutting zone temperature according to claim 1, characterized in that: The signal analysis and inversion in S13 specifically includes: the infrared thermometer receives the infrared radiation signal from the processing area and after penetrating the tool body, and performs photoelectric conversion, noise filtering, intensity calibration and temperature inversion calculation on the radiation signal to analyze the original radiation signal into an accurate temperature value.

5. The method for online measurement of cutting zone temperature according to claim 1, characterized in that: The signal analysis and inversion process in S13 is as follows: S121, Photoelectric conversion: Converting infrared radiation signals into electrical signals; S122. Noise filtering: Eliminate random interference in the signal; S123, Intensity Calibration: Based on the standard blackbody source, the correspondence between radiation intensity and temperature is calibrated; S124. Temperature inversion calculation: Combining tool material transmittance correction and atmospheric transmission attenuation compensation, the true temperature is inverted by substituting into Planck's formula. ； in: Let λ be the blackbody monochromatic radiant exitance at wavelength λ (W / (m²·μm)), and h be Planck's constant (6.626 × 10⁻³). 4 J・s), c is the speed of light (3×10⁻⁶ s), and c is the speed of light 8 m / s), λ is the measurement wavelength (μm), k is the Boltzmann constant (1.38×10⁻²³J / K), and T is the absolute temperature (K).

6. A nanometer-resolution analytical method for cutting temperature fields, characterized in that, Includes the following steps: S21. First, determine the key machining parameters, such as workpiece material, tool geometry parameters, and cutting parameters. Refine the mesh to the nanometer level. Based on these parameters, establish a cutting machining thermo-mechanical coupled finite element model, perform simulation calculations under multiple sets of parameters, and generate an initial simulation temperature field database. S22. Secondly, a hybrid training database is constructed by pairing the average temperature data measured by micron-level light spots with the precise temperature field data simulated at the nanometer level to form a hybrid dataset for training neural networks. S23. Then, network models such as encoder-decoder are used for training. The training process is based on a hybrid training database to strengthen the correlation between micron-level temperature data and nano-level simulated temperature. This enables the model to accurately separate and reconstruct the temperature field of the nanoscale processing area from the macro-average data. During the training process, the full-field temperature field distribution under the corresponding conditions is used as the training objective. By minimizing the mean square error or mean absolute error between the predicted result and the true value, the global optimization algorithm is used to iteratively update the network parameters until the model converges. S24. Finally, in actual cutting, the real-time acquired local temperature data is input into the trained model, and the corrected and reconstructed complete high-precision temperature field distribution is output in real time.

7. The nanometer-resolution analytical method for cutting temperature field according to claim 6, characterized in that: In step S23, the encoder-decoder network model includes: the encoder part is composed of multiple convolutional layers and pooling layers stacked sequentially, extracting multi-level features of the infrared temperature image through convolution operations, wherein the function of the convolutional layer is: to extract multi-level features of the infrared temperature image through convolution kernels, as shown in the formula: ； in: Let k be the feature value at position (i,j) in the l-th layer, and k be the kernel size. The weights of the l-th layer convolutional kernel are... For bias terms, This is the activation function.

Images